Many biological molecules exist as “diastereomers” — molecules that have the same chemical structure but different spatial arrangements of their atoms. In some cases, these slight structural differences can lead to significant changes in the molecules’ functions or chemical properties.

As one example, the cancer drug doxorubicin can have heart-damaging side effects in a small percentage of patients. However, a diastereomer of the drug, known as epirubicin, which has a single alcohol group that points in a different direction, is much less toxic to heart cells.

“There are a lot of examples like that in medicinal chemistry where something that seems small, such as the position of a single atom in space, may actually be really profound,” says Alison Wendlandt, an associate professor of chemistry at MIT.

Wendlandt’s lab is focused on designing new tools that can convert these molecules into different forms.  Her group is also working on similar tools that can change a molecule into a different constitutional isomer — a molecule that has an atom or chemical group located in a different spot, even though it has the same chemical formula as the original.

“If you have a target molecule and you needed to make it without such a tool, you would have to go back to the beginning and make the whole molecule again to get to the final structure that you wanted,” Wendlandt says.

These tools can also lend themselves to creating entirely new molecules that might be difficult or even impossible to build using traditional chemical synthesis techniques.

“We’re focused on a broad suite of selective transformations, the goal being to make the biggest impact on how you might envision making a molecule,” she says. “If you are able to open up access to the interconversion of molecular structures, you can then think completely differently about how you would make a molecule.”

From math to chemistry

As the daughter of two geologists, Wendlandt found herself immersed in science from a young age. Both of her parents worked at the Colorado School of Mines, and family vacations often involved trips to interesting geological formations.

In high school, she found math more appealing than chemistry, and she headed to the University of Chicago with plans to major in mathematics. However, she soon had second thoughts, after encountering abstract math.

“I was good at calculus and the kind of math you need for engineering, but when I got to college and I encountered topology and N-dimensional geometry, I realized I don’t actually have the skills for abstract math. At that point I became a little bit more open-minded about what I wanted to study,” she says.

Though she didn’t think she liked chemistry, an organic chemistry course in her sophomore year changed her mind.

“I loved the problem-solving aspect of it. I have a very, very bad memory, and I couldn’t memorize my way through the class, so I had to just learn it, and that was just so fun,” she says.

As a chemistry major, she began working in a lab focused on “total synthesis,” a research area that involves developing strategies to synthesize a complex molecule, often a natural compound, from scratch.

Although she loved organic chemistry, a lab accident — an explosion that injured a student in her lab and led to temporary hearing loss for Wendlandt — made her hesitant to pursue it further. When she applied to graduate schools, she decided to go into a different branch of chemistry — chemical biology. She studied at Yale University for a couple of years, but she realized that she didn’t enjoy that type of chemistry and left after receiving a master’s degree.

She worked in a lab at the University of Kentucky for a few years, then applied to graduate school again, this time at the University of Wisconsin. There, she worked in an organic chemistry lab, studying oxidation reactions that could be used to generate pharmaceuticals or other useful compounds from petrochemicals.

After finishing her PhD in 2015, Wendlandt went to Harvard University for a postdoc, working with chemistry professor Eric Jacobsen. There, she became interested in selective chemical reactions that generate a particular isomer, and began studying catalysts that could perform glycosylation — the addition of sugar molecules to other molecules — at specific sites.

Editing molecules

Since joining the MIT faculty in 2018, Wendlandt has worked on developing catalysts that can convert a molecule into its mirror image or an isomer of the original.

In 2022, she and her students developed a tool called a stereo-editor, which can alter the arrangement of chemical groups around a central atom known as a stereocenter. This editor consists of two catalysts that work together to first add enough energy to remove an atom from a stereocenter, then replace it with an atom that has the opposite orientation. That energy input comes from a photocatalyst, which converts captured light into energy.

“If you have a molecule with an existing stereocenter, and you need the other enantiomer, typically you would have to start over and make the other enantiomer. But this new method tries to interconvert them directly, so it gives you a way of thinking about molecules as dynamic,” Wendlandt says. “You could generate any sort of three-dimensional structure of that molecule, and then in an independent step later, you could completely reorganize the 3D structure.”

She has also developed tools that can convert common sugars such as glucose into other isomers, including allose and other sugars that are difficult to isolate from natural sources, and tools that can create new isomers of steroids and alcohols. She is now working on ways to convert six-membered carbon rings to seven or eight-membered rings, and to add, subtract, or replace some of the chemical groups attached to the rings.

“I’m interested in creating general tools that will allow us to interconvert static structures. So, that may be taking a certain functional group and moving it to another part of the molecule entirely, or taking large rings and making them small rings,” she says. “Instead of thinking of molecules that we assemble as static, we’re thinking about them now as potentially dynamic structures, which could change how we think about making organic molecules.”

This approach also opens up the possibility of creating brand new molecules that haven’t been seen before, Wendlandt says. This could be useful, for example, to create drug molecules that interact with a target enzyme in just the right way.

“There’s a huge amount of chemical space that’s still unknown, bizarre chemical space that just has not been made. That’s in part because maybe no one has been interested in it, or because it’s just too hard to make that specific thing,” she says. “These kinds of tools give you access to isomers that are maybe not easily made.”

Anders Sejr Hansen, Class of 1943 Career Development Professor in the Department of Biological Engineering, has been named as the recipient of the 2024-25 Harold E. Edgerton Faculty Achievement Award.

The annual award was established in fall 1982 as a permanent tribute to Institute Professor Emeritus Harold E. Edgerton for his great and enduring support for younger faculty members over the years. The purpose of the award is to recognize exceptional distinction in teaching, in research, and in service.

Hansen is the principal investigator of the Hansen Lab, which develops new methods to resolve 3D genome structure at high spatiotemporal resolution to understand how DNA looping and 3D folding regulates gene expression in health and disease. His areas of research include cancer biology, computational systems biology, instrumentation and measurement, and synthetic biology.

“My research focuses on how the expression of our genes is regulated,” says Hansen. “All the cells in our body have the same DNA and the same genes. Thus, the software or applications to each cell are the same. What’s different between a neuron and a blood cell is what genes they choose to express. My research focuses on understanding how this regulation takes place.”

Those who nominated Anders for the award emphasized his remarkable productivity, mentioning his two “highly cited, paradigm-shifting research articles in Science and Nature Genetics,” and his research presentations at 50 invited talks, including two keynotes, at universities and conferences worldwide. They also highlighted his passion for mentorship and career development for the 20 current members of his laboratory.

“Anders is an outstanding role model and ambassador of biological engineering, combining a powerful research program, run as a caring mentor, and innovative undergraduate education,” says Christopher Voigt, the Daniel I.C. Wang Professor in Biological Engineering and head of the Department of Biological Engineering.

Adds Laurie Boyer, a professor of biology and biological engineering, “His work reveals new insights into how we think about the dynamics of gene regulation that would not otherwise be possible. The Hansen Lab’s work provides a unified framework rapidly adopted by the field to learn how conserved regulators provide exquisite spatial and temporal control of gene expression in the context of 3D genome architecture.”

During the nomination process, students praised Hansen’s passion for his work, along with his ability to prepare them to apply their education outside the classroom.

“He always strives to guide each lab member towards both short-term scientific success and long-term career planning through regular one-on-one meetings, facilitating collaborations and access to scientific resources, and sharing his own experiences,” says Jin Yang, a graduate student in biological engineering and member of the Hansen Lab.

“Dr. Hansen's infectious excitement for the course material made it very enjoyable to come to class and envision potential applications of the fundamental topics he taught,” adds another one of his students. “Excellent lecturer!”

Hansen obtained his undergraduate and master’s degree in chemistry at Oxford University. He received his PhD in chemistry and chemical biology from Harvard University, where he applied systems biology approaches to understand how cells can encode and transmit information in the dynamics of transcription factor activation. For his postdoc at the University of California at Berkeley, Hansen developed new imaging approaches for dissecting the dynamics of architectural proteins with single-molecule resolution in living cells. Hansen joined MIT as an assistant professor of biological engineering in early 2020.

His recognitions include an NIH K99 Pathway to Independence Award (2019), NIH Director’s New Innovator Award (2020), a Pew-Stewart Scholar for Cancer Research Award (2021), an NSF CAREER Award (2024), and an NIH Director’s Transformative Research Award (2024).

Hansen has served on several committees at MIT, including the MIT Biological Engineering Graduate Program Admissions Committee, the MIT Computational and Systems Biology Graduate Admissions Committee, and the MIT Biological Engineering Graduate Recruiting Committee, of which he has been chair since 2023.

“I have known about the Edgerton Award since I started at MIT, and I think the broad focus on both research, teaching, and service really captures what makes MIT such a unique and wonderful place,” says Hansen. “I was therefore absolutely thrilled to receive the news that I would receive the Edgerton Award this year, and I am very grateful to all the wonderful colleagues here at MIT who have supported me over the years, and all the exceptional people in my lab whose work is being recognized.”

Johann Wolfgang von Goethe (1749-1832), the German polymath whose life and work embodied the connections between the arts and sciences, is said to have described architecture as “frozen music.” 

When the new Edward and Joyce Linde Music Building at MIT had its public opening earlier this year, the temperature outside may have been below freezing but the performances inside were a warm-up for the inaugural concert that took place in the evening. During the afternoon, visitors were invited to workshops in Balinese gamelan and Senegalese drumming, alongside performances by the MIT Chamber Music Society, MIT Festival Jazz Ensemble, and the MIT Laptop Ensemble (FaMLE), demonstrating the synergy between global music traditions and contemporary innovation in music technology. The building was filled with visitors from the MIT community and the Boston area, keen to be among the first to enter the new building and discover what MIT Music had planned for the opening occasion.

The evening’s landmark concert, Sonic Jubilance, celebrated the building’s completion and the pivotal role of MIT Music and Theater Arts (MTA) at the center of life on campus. The program was distinguished by five world premieres by MIT composers: “Summit and Mates,” by assistant professor in jazz Miguel Zenón; “Grace,” by senior lecturer in music Charles Shadle; “Two Noble Kinsmen,” by professor emeritus in music John Harbison; and “Madrigal,” by Keeril Makan, the Michael (1949) and Sonja Koerner Music Composition Professor. 

The premieres were interwoven through the program with performances by MIT ensembles demonstrating the breadth and depth of the conservatory-level music program — from the European classical tradition to Brazilian beats to Boston jazz (the full list of participating ensembles can be found below). 

Each performance demonstrated the different ways the space could be used to create new relationships between musicians and audiences. Designed in the round by the architecture firm SANAA, the Thomas Tull Concert Hall allows sound to resonate from the circular stage or from the aisles above the tiered seating; performers might be positioned below, above, or even in the midst of the audience.

“Music has been a part of MIT's curriculum and culture from the beginning,” said Chancellor Melissa Nobles in her opening address. “Arriving at this magnificent space has taken the collective efforts of past presidents, provosts, deans, faculty, alumni, and students, all working to get us here this evening.” 

Jay Scheib, the Class of 1949 Professor and MIT MTA section head, emphasized the vital role of Music at MIT as a source of cohesion and creativity for students, faculty, and the wider MIT community. 

“The new building is an extraordinary home for us. As a destination to convene communities around world musics and cultures, to engage in emerging music technologies, and to experience concerts and premieres featuring our extraordinary students and our internationally renowned faculty — the Edward and Joyce Linde Music Building is truly a transformational thing." 

The concert was also the launch event of Artfinity, MIT’s largest public festival of the arts since 2011, featuring more than 80 free performing and visual arts events. The concert hall will host performances throughout the spring, ranging from classical to jazz to rap, and more.

Institute Professor Marcus Thompson — the faculty co-lead for Artfinity alongside Azra Akšamija, director and associate professor of the Art, Culture, and Technology Program (ACT) at MIT — shared thoughts on the Edward and Joyce Linde Music Building as a point of orientation for the festival. 

“Our building offers the opportunity to point to the presence and importance of other art forms, media, practices, and experiences that can bring us together as practitioners and audiences, lifting our spirits and our sights,” Thompson reflected. “An ensemble of any kind is a community as well as a metaphor for what connects us, applying different talents to create more than we can do alone.”

The new compositions by the four faculty members were a case in point. The program opened with “Summit,” a brass fanfare projected from the top of the hall with ceremonial zeal. “The piece was specifically written as an opener for the concert,” Zenón explained. “My aim was to compose something that would make a statement straight away, while also using the idea of the ‘groove’ as a driving force. The title has two meanings. The first is a mountaintop, or the top of a structure — which is where the ensemble will be placed for the performance. The second is a gathering of great minds and great leaders, which is what MIT feels like for me.” Later in the program, Zenón premiered a jazz contrafact, “Mates,” playing on Benny Golson’s Stablemates, a tribute to Herb Pomeroy, founder of MIT’s jazz program. “The idea here is to use something connected to the jazz tradition — and to Boston’s history — and approach it from a more personal perspective,” said Zenón.

“Two Noble Kinsmen,” by Harbison, was composed as a benediction for the new home of MIT Music. “In choosing to set Shakespeare’s final words in this new piece for choir and strings, I wanted to convey the sense of an invocation, an introduction, an address to unseen forces,” said Harbison. “In this case, I wanted to leave the musical structure as plain as possible so that we understand why these words are chosen. I hoped to capture the stoic balance of these lines — they are in themselves a kind of verbal music.”

In setting the words of the poem “Grace,” by the Chickasaw poet Linda Hogan, Shadle — a composer of Choctaw heritage — envisioned a “sonic extension” of the MIT Land Acknowledgement. “‘Grace’ intended to speak to the Indigenous presence at the Institute and to open the new building with a reminder of the balm music that can bring to a troubled world,” said Shadle. “I hope that I have composed music that links Indigenous and Western traditions in ways that are compelling and thoughtful and that, while recognizing the ‘pieces of hurt,’ still makes a place for grace.”

Before the concert’s euphoric finale — a performance by Rambax Senegalese Drum Ensemble directed by Lamine Touré — “Madrigal” (the evening’s fourth world premiere) served to demonstrate the spatial dimensions of sound made possible by the design of the concert hall. 

Makan’s composition was performed by four student violinists positioned at the top of each aisle and a fifth, Professor Natalie Lin Douglas, at the center of the stage, simultaneously showcasing the geometry of the hall and referencing the ever-shifting perspectives of the sculpture that stands at the north entrance of the building — “Madrigal (2024),” by Sanford Biggers.

“My piece aims to capture the multifaceted quality of Sanford Biggers’ sculpture. From whichever vantage point we might look at it, we see the same patterns in new relationships with one another. In other words, there is no one point of view that is privileged over another.”

As faculty lead for the building project, Makan developed a friendship with Joyce Linde, who provided the principal gift that led to the building. “Joyce and I were on the selection committee to choose an artist to create a site-specific sculpture outside the building. She was very excited about the process, and very engaged with Sanford,” said Makan. “Joyce passed away before she was able to see the building’s completion, and I wanted to honor her legacy by writing an original piece of music in her memory.”

That sense of relationship, pattern-making, and new beginnings was articulated by Frederick Harris, director and senior lecturer in music and the co-producer of the concert, alongside Andy Wilds, program manager in music. “The hall is an instrument; we’re communing with this incredible space and getting to know it,” said Harris. “It’s a relationship. The circular form of the hall is very welcoming, not only to immersive experiences but also to shared experiences.”

The role of music in cultivating community will ensure that the building will become an integral part of MIT life. The work taking place in rehearsal rooms matches the innovation of the Institute’s labs — proving that the arts are a necessary counterpart to science and technology, continuous with the human instinct to express and invent. Sonic Jubilance sets the tone of what’s to come. 

MIT Music ensembles (in order of concert appearance):

• MIT Concert Choir

• MIT Chamber Chorus

• MIT Chamber Music Society

• MIT Vocal Jazz Ensemble

• MIT Jazz Advanced Music Performance Ensemble

• MIT Axiom Ensemble

• MIT Wind Ensemble

• MIT Gamelan Galak Tika

• Rambax MIT
   

When Michael Benjamin, principal research scientist in the MIT Center for Ocean Engineering, arrived at MIT 25 years ago, only professors and postdocs were allowed to touch the department’s underwater vehicles. The vehicles were expensive, he explains, and required extensive training to operate.

“People were scared to death about losing or damaging them, [and] there was no education pipeline to teach students,” he says, adding that the introduction of class 2.680 (Marine Autonomy, Sensing, and Communication) changed this a lot, by creating a class where undergraduate and graduate students could learn to write autonomy code, and run their software on robots on the Charles River. The addition of class 2.S01 (Introduction to Autonomous Underwater Vehicles) last year took the hands-on learning opportunities even further.

“2.S01 is a return to our roots: underwater vehicles. We wanted to create a learning environment where every student handles a robot, and no one is afraid about losing one,” he says. Each student is sent home with an electronics kit, which Benjamin calls the heart of the robot. “They can experiment all they want in their dorm room, and we’ll give them another kit if they break it.”

The AUVs and student test kits in 2.S01 were designed and built by Supun Randeni, a research scientist in mechanical engineering and the primary lecturer and content creator of 2.S01, and Captain Michael Sacarny from MIT Sea Grant. “Dr. Randeni and Captain Sacarny are the geniuses behind the class,” says Benjamin. Together, Randeni and Sacarny run the hands-on lab instruction.

The goal is to expand education and research opportunities to include a larger and younger group of students. “It’s the exact opposite of 25 years ago, when only a privileged few people were allowed to get inside the robot,” says Benjamin. “Student growth and interest is directly related to the degree they have ‘ownership’ of their robot. Physical possession, but also responsibility for its safe operation and return.”

2.S01 provides students with an in-depth insight into autonomous underwater vehicles (AUVs), by introducing theoretical and practical aspects of the AUV design process. This includes fundamentals of naval architecture, electrical systems design, mechanical design, and software design. Students assemble their own AUVs by using a kit of parts and guidance from instructors, beginning with core electronics and building out a full vehicle for deployment in the Charles River on the MIT campus in the final weeks.

Among the activities, students engage in waterproofing vacuum tests, pre-launch sub-system tests, and dockside tests for ballasting, all followed by in-water low-level control tuning runs. Students also construct autonomy missions — first in simulation, followed by in-water autonomous missions to conduct an environmental survey in the Charles River. The course’s final labs include group competitions involving in-water challenges. For the second iteration of the course, which starts in late March, the instructors plan to add more labs that allow the students to explore the intricacies of the electronic, more simulations options, and more water time.

Adowyn Bryne, a second-year mechanical engineering (MechE) student, took the course last year as a member of the first cohort, but this wasn’t her first experience with underwater vehicles. She’d participated in a SeaPerch program in high school. “I chose 2.S01 because I wanted to learn about more complex underwater vehicles,” says Bryne. “I didn’t find out until later in the semester that SeaPerch was actually started at MIT Sea Grant!”

Benjamin says he hopes there are a few things that first-year students take away from participating in 2.S01: first, an understanding that marine robotics is a very cross-disciplinary effort, involving mechanical engineering, electrical engineering, control theory, computer science and ocean science; and, second, the opportunity to view the effort as a gateway to exploring and understanding the ocean. Students says it’s that, and so much more.

Isabella Yeung, a third-year Course 12 student, took the class during her sophomore year after participating in an MIT Undergraduate Research Opportunities Program (UROP) in the MIT Sea Grant Bio Lab with Carolina Bastidas. Bastidas is a research scientist in MIT Sea Grant's Marine Advisory Services group.

“While UROP-ing, I’d seen many AUVs and other projects being developed at MIT Sea Grant,” Yeung says. “I was curious to learn more and have a deeper insight into what they were doing. This class was a prime opportunity to jump into the world of marine robotics without having any background in Course 2.”

She called the course “easily one of the most hands-on (and downright fun) classes” she’s ever taken, adding that she appreciated having the opportunity to assemble and deploy the AUV.

“As someone who enjoys tinkering, I appreciated the opportunity to get my hands dirty — quite literally, with grease and Charles [River] water,” says Yeung. “I looked forward to all of the classes, especially the deployment sessions. Nothing quite matched the sheer rush of launching our program, rushing to drop the AUV into the Charles, and engaging in a boat chase, hoping it hadn’t gone rogue.”

Bryne advises students considering the course to not worry too much if the class lines up with a particular career path they’re considering. “Your first year is about exploring. If you’re interested in the class, take it! You might find a new area of interest. Regardless of whether you want to keep learning about AUVs, you’ll get valuable transferable skills and have a lot of fun.”

Bryne, herself, says the experience is helping to set the stage for exploring future interests and opportunities. “Every time I’ve gotten to do something with robots, I’ve loved it,” she says, “but I’m also very passionate about women’s health. I want to design medical technology specifically for women, but I definitely think there’s room to incorporate robotics into that. It’s great that MechE is such a broad field, and that the curriculum at MIT allows me to explore so many potential areas of study.”

Many diseases are caused by dysfunctional gene expression that leads to too much or too little of a given protein. Efforts to cure those diseases include everything from editing genes to inserting new genetic snippets into cells to injecting the missing proteins directly into patients.

CAMP4 is taking a different approach. The company is targeting a lesser-known player in the regulation of gene expression known as regulatory RNA. CAMP4 co-founder and MIT Professor Richard Young has shown that by interacting with molecules called transcription factors, regulatory RNA plays an important role in controlling how genes are expressed. CAMP4’s therapeutics target regulatory RNA to increase the production of proteins and put patients’ levels back into healthy ranges.

The company’s approach holds promise for treating diseases caused by defects in gene expression, such as metabolic diseases, heart conditions, and neurological disorders. Targeting regulatory RNAs as opposed to genes could also offer more precise treatments than existing approaches.

“If I just want to fix a single gene’s defective protein output, I don’t want to introduce something that makes that protein at high, uncontrolled amounts,” says Young, who is also a core member of the Whitehead Institute. “That’s a huge advantage of our approach: It’s more like a correction than sledgehammer.”

CAMP4’s lead drug candidate targets urea cycle disorders (UCDs), a class of chronic conditions caused by a genetic defect that limits the body’s ability to metabolize and excrete ammonia. A phase 1 clinical trial has shown CAMP4’s treatment is safe and tolerable for humans, and in preclinical studies the company has shown its approach can be used to target specific regulatory RNA in the cells of humans with UCDs to restore gene expression to healthy levels.

“This has the potential to treat very severe symptoms associated with UCDs,” says Young, who co-founded CAMP4 with cancer genetics expert Leonard Zon, a professor at Harvard Medical School. “These diseases can be very damaging to tissues and causes a lot of pain and distress. Even a small effect in gene expression could have a huge benefit to patients, who are generally young.”

Mapping out new therapeutics

Young, who has been a professor at MIT since 1984, has spent decades studying how genes are regulated. It’s long been known that molecules called transcription factors, which orchestrate gene expression, bind to DNA and proteins. Research published in Young’s lab uncovered a previously unknown way in which transcription factors can also bind to RNA. The finding indicated RNA plays an underappreciated role in controlling gene expression.

CAMP4 was founded in 2016 with the initial idea of mapping out the signaling pathways that govern the expression of genes linked to various diseases. But as Young’s lab discovered and then began to characterize the role of regulatory RNA in gene expression around 2020, the company pivoted to focus on targeting regulatory RNA using therapeutic molecules known as antisense oligonucleotides (ASOs), which have been used for years to target specific messenger RNA sequences.

CAMP4 began mapping the active regulatory RNAs associated with the expression of every protein-coding gene and built a database, which it calls its RAP Platform, that helps it quickly identify regulatory RNAs to target  specific diseases and select ASOs that will most effectively bind to those RNAs.

Today, CAMP4 is using its platform to develop therapeutic candidates it believes can restore healthy protein levels to patients.

“The company has always been focused on modulating gene expression,” says CAMP4 Chief Financial Officer Kelly Gold MBA ’09. “At the simplest level, the foundation of many diseases is too much or too little of something being produced by the body. That is what our approach aims to correct.”

Accelerating impact

CAMP4 is starting by going after diseases of the liver and the central nervous system, where the safety and efficacy of ASOs has already been proven. Young believes correcting genetic expression without modulating the genes themselves will be a powerful approach to treating a range of complex diseases.

“Genetics is a powerful indicator of where a deficiency lies and how you might reverse that problem,” Young says. “There are many syndromes where we don’t have a complete understanding of the underlying mechanism of disease. But when a mutation clearly affects the output of a gene, you can now make a drug that can treat the disease without that complete understanding.”

As the company continues mapping the regulatory RNAs associated with every gene, Gold hopes CAMP4 can eventually minimize its reliance on wet-lab work and lean more heavily on machine learning to leverage its growing database and quickly identify regRNA targets for every disease it wants to treat.

In addition to its trials in urea cycle disorders, the company plans to launch key preclinical safety studies for a candidate targeting seizure disorders with a genetic basis, this year. And as the company continues exploring drug development efforts around the thousands of genetic diseases where increasing protein levels are can have a meaningful impact, it’s also considering collaborating with others to accelerate its impact.

“I can conceive of companies using a platform like this to go after many targets, where partners fund the clinical trials and use CAMP4 as an engine to target any disease where there’s a suspicion that gene upregulation or downregulation is the way to go,” Young says.

It’s considered a scientific landmark: A 1975 meeting at the Asilomar Conference Center in Pacific Grove, California, shaped a new safety regime for recombinant DNA, ensuring that researchers would apply caution to gene splicing. Those ideas have been so useful that in the decades since, when new topics in scientific safety arise, there are still calls for Asilomar-type conferences to craft good ground rules.

There’s something missing from this narrative, though: It took more than the Asilomar conference to set today’s standards. The Asilomar concepts were created with academic research in mind — but the biotechnology industry also makes products, and standards for that were formulated after Asilomar.

“The Asilomar meeting and Asilomar principles did not settle the question of the safety of genetic engineering,” says MIT scholar Robin Scheffler, author of a newly published research paper on the subject.

Instead, as Scheffler documents in the paper, Asilomar helped generate further debate, but those industry principles were set down later in the 1970s — first in Cambridge, Massachusetts, where politicians and concerned citizens wanted local biotech firms to be good neighbors. In response, the city passed safety laws for the emerging industry. And rather than heading off to places with zero regulations, local firms — including a fledgling Biogen — stayed put. Over the decades, the Boston area became the world leader in biotech.

Why stay? In essence, regulations gave biotech firms the certainty they needed to grow — and build. Lenders and real-estate developers needed signals that long-term investment in labs and facilities made sense. Generally, as Scheffler notes, even though “the idea that regulations can be anchoring for business does not have a lot of pull” in economic theory, in this case, regulations did matter.

“The trajectory of the industry in Cambridge, including biotechnology companies deciding to accommodate regulation, is remarkable,” says Scheffler. “It’s hard to imagine the American biotechnology industry without this dense cluster in Boston and Cambridge. These things that happened on a very local scale had huge echoes.”

Scheffler’s article, “Asilomar Goes Underground: The Long Legacy of Recombinant DNA Hazard Debates for the Greater Boston Area Biotechnology Industry,” appears in the latest issue of the Journal of the History of Biology. Scheffler is an associate professor in MIT’s Program in Science, Technology, and Society.

Business: Banking on certainty

To be clear, the Asilomar conference of 1975 did produce real results. Asilomar led to a system that helped evaluate projects’ potential risk and determine appropriate safety measures. The U.S. federal government subsequently adopted Asilomar-like principles for research it funded.

But in 1976, debate over the subject arose again in Cambridge, especially following a cover story in a local newspaper, the Boston Phoenix. Residents became concerned that recombinant DNA projects would lead to, hypothetically, new microorganisms that could damage public health.

“Scientists had not considered urban public health,” Scheffler says. “The Cambridge recombinant DNA debate in the 1970s made it a matter of what your neighbors think.”

After several months of hearings, research, and public debate (sometimes involving MIT faculty) stretching into early 1977, Cambridge adopted a somewhat stricter framework than the federal government had proposed for the handling of materials used in recombinant DNA work.

“Asilomar took on a new life in local regulations,” says Scheffler, whose research included government archives, news accounts, industry records, and more.

But a funny thing happened after Cambridge passed its recombinant DNA rules: The nascent biotech industry took root, and other area towns passed their own versions of the Cambridge rules.

“Not only did cities create more safety regulations,” Scheffler observes, “but the people asking for them switched from being left-wing activists or populist mayors to the Massachusetts Biotechnology Council and real estate development concerns.”

Indeed, he adds, “What’s interesting is how quickly safety concerns about recombinant DNA evaporated. Many people against recombinant DNA came to change their thinking.” And while some local residents continued to express concerns about the environmental impact of labs, “those are questions people ask when they no longer worry about the safety of the core work itself.”

Unlike federal regulations, these local laws applied to not only lab research but also products, and as such they let firms know they could work in a stable business environment with regulatory certainty. That mattered financially, and in a specific way: It helped companies build the buildings they needed to produce the products they had invented.

“The venture capital cycle for biotechnology companies was very focused on the research and exciting intellectual ideas, but then you have the bricks and mortar,” Scheffler says, referring to biotech production facilities. “The bricks and mortar is actually the harder problem for a lot of startup biotechnology companies.”

After all, he notes, “Venture capital will throw money after big discoveries, but a banker issuing a construction loan has very different priorities and is much more sensitive to things like factory permits and access to sewers 10 years from now. That’s why all these towns around Massachusetts passed regulations, as a way of assuring that.”

To grow globally, act locally

Of course, one additional reason biotech firms decided to land in the Boston area was the intellectual capital: With so many local universities, there was a lot of industry talent in the region. Local faculty co-founded some of the high-flying firms.

“The defining trait of the Cambridge-Boston biotechnology cluster is its density, right around the universities,” Scheffler says. “That’s a unique feature local regulations encouraged.”

It’s also the case, Scheffler notes, that some biotech firms did engage in venue-shopping to avoid regulations at first, although that was more the case in California, another state where the industry emerged. Still, the Boston-area regulations seemed to assuage both industry and public worries about the subject.

The foundations of biotechnology regulation in Massachusetts contain some additional historical quirks, including the time in the late 1970s when the city of Cambridge mistakenly omitted the recombinant DNA safety rules from its annually published bylaws, meaning the regulations were inactive. Officials at Biogen sent them a reminder to restore the laws to the books.

Half a century on from Asilomar, its broad downstream effects are not just a set of research principles — but also, refracted through the Cambridge episode, key ideas about public discussion and input; reducing uncertainty for business, the particular financing needs of industries; the impact of local and regional regulation; and the openness of startups to recognizing what might help them thrive.

“It’s a different way to think about the legacy of Asilomar,” Scheffler says. “And it’s a real contrast with what some people might expect from following scientists alone.” 

When some commuter trains arrive at the end of the line, they must travel to a switching platform to be turned around so they can depart the station later, often from a different platform than the one at which they arrived.

Engineers use software programs called algorithmic solvers to plan these movements, but at a station with thousands of weekly arrivals and departures, the problem becomes too complex for a traditional solver to unravel all at once.

Using machine learning, MIT researchers have developed an improved planning system that reduces the solve time by up to 50 percent and produces a solution that better meets a user’s objective, such as on-time train departures. The new method could also be used for efficiently solving other complex logistical problems, such as scheduling hospital staff, assigning airline crews, or allotting tasks to factory machines.

Engineers often break these kinds of problems down into a sequence of overlapping subproblems that can each be solved in a feasible amount of time. But the overlaps cause many decisions to be needlessly recomputed, so it takes the solver much longer to reach an optimal solution.

The new, artificial intelligence-enhanced approach learns which parts of each subproblem should remain unchanged, freezing those variables to avoid redundant computations. Then a traditional algorithmic solver tackles the remaining variables.

“Often, a dedicated team could spend months or even years designing an algorithm to solve just one of these combinatorial problems. Modern deep learning gives us an opportunity to use new advances to help streamline the design of these algorithms. We can take what we know works well, and use AI to accelerate it,” says Cathy Wu, the Thomas D. and Virginia W. Cabot Career Development Associate Professor in Civil and Environmental Engineering (CEE) and the Institute for Data, Systems, and Society (IDSS) at MIT, and a member of the Laboratory for Information and Decision Systems (LIDS).

She is joined on the paper by lead author Sirui Li, an IDSS graduate student; Wenbin Ouyang, a CEE graduate student; and Yining Ma, a LIDS postdoc. The research will be presented at the International Conference on Learning Representations.

Eliminating redundance

One motivation for this research is a practical problem identified by a master’s student Devin Camille Wilkins in Wu’s entry-level transportation course. The student wanted to apply reinforcement learning to a real train-dispatch problem at Boston’s North Station. The transit organization needs to assign many trains to a limited number of platforms where they can be turned around well in advance of their arrival at the station.

This turns out to be a very complex combinatorial scheduling problem — the exact type of problem Wu’s lab has spent the past few years working on.

When faced with a long-term problem that involves assigning a limited set of resources, like factory tasks, to a group of machines, planners often frame the problem as Flexible Job Shop Scheduling.

In Flexible Job Shop Scheduling, each task needs a different amount of time to complete, but tasks can be assigned to any machine. At the same time, each task is composed of operations that must be performed in the correct order.

Such problems quickly become too large and unwieldy for traditional solvers, so users can employ rolling horizon optimization (RHO) to break the problem into manageable chunks that can be solved faster.

With RHO, a user assigns an initial few tasks to machines in a fixed planning horizon, perhaps a four-hour time window. Then, they execute the first task in that sequence and shift the four-hour planning horizon forward to add the next task, repeating the process until the entire problem is solved and the final schedule of task-machine assignments is created.

A planning horizon should be longer than any one task’s duration, since the solution will be better if the algorithm also considers tasks that will be coming up.

But when the planning horizon advances, this creates some overlap with operations in the previous planning horizon. The algorithm already came up with preliminary solutions to these overlapping operations.

“Maybe these preliminary solutions are good and don’t need to be computed again, but maybe they aren’t good. This is where machine learning comes in,” Wu explains.

For their technique, which they call learning-guided rolling horizon optimization (L-RHO), the researchers teach a machine-learning model to predict which operations, or variables, should be recomputed when the planning horizon rolls forward.

L-RHO requires data to train the model, so the researchers solve a set of subproblems using a classical algorithmic solver. They took the best solutions — the ones with the most operations that don’t need to be recomputed — and used these as training data.

Once trained, the machine-learning model receives a new subproblem it hasn’t seen before and predicts which operations should not be recomputed. The remaining operations are fed back into the algorithmic solver, which executes the task, recomputes these operations, and moves the planning horizon forward. Then the loop starts all over again.

“If, in hindsight, we didn’t need to reoptimize them, then we can remove those variables from the problem. Because these problems grow exponentially in size, it can be quite advantageous if we can drop some of those variables,” she adds.

An adaptable, scalable approach

To test their approach, the researchers compared L-RHO to several base algorithmic solvers, specialized solvers, and approaches that only use machine learning. It outperformed them all, reducing solve time by 54 percent and improving solution quality by up to 21 percent.

In addition, their method continued to outperform all baselines when they tested it on more complex variants of the problem, such as when factory machines break down or when there is extra train congestion. It even outperformed additional baselines the researchers created to challenge their solver.

“Our approach can be applied without modification to all these different variants, which is really what we set out to do with this line of research,” she says.

L-RHO can also adapt if the objectives change, automatically generating a new algorithm to solve the problem — all it needs is a new training dataset.

In the future, the researchers want to better understand the logic behind their model’s decision to freeze some variables, but not others. They also want to integrate their approach into other types of complex optimization problems like inventory management or vehicle routing.

This work was supported, in part, by the National Science Foundation, MIT’s Research Support Committee, an Amazon Robotics PhD Fellowship, and MathWorks.

On board Intuitive Machines’ Athena spacecraft, which made a moon landing on March 6, were cutting-edge MIT payloads: a depth-mapping camera and a mini-rover called “AstroAnt.” Also on that craft were the words and voices of people from around the world speaking in dozens of languages. These were etched on a 2-inch silicon wafer computationally designed by Professor Craig Carter of the MIT Department of Materials Science and Engineering and mounted on the mission’s Lunar Outpost MAPP Rover.

Dubbed the Humanity United with MIT Art and Nanotechnology in Space (HUMANS), the project is a collaboration of art and science, bringing together experts from across MIT — with technical expertise from the departments of Aeronautics and Astronautics, Mechanical Engineering, and Electrical Engineering and Computer Science; nano-etching and testing from MIT.nano; audio processing from the MIT Media Lab’s Opera of the Future and the Music and Theater Arts Section; and lunar mission support from the Media Lab’s Space Exploration Initiative.

While a 6-inch HUMANS wafer flew on the Axiom-2 mission to the International Space Station in 2023, the 2-inch wafer was a part of the IM-2 mission to the lunar south polar region, linked to the MIT Media Lab’s To the Moon to Stay program, which reimagines humankind’s return to the moon. IM-2 ended prematurely after the Athena spacecraft tipped onto its side shortly after landing in March, but the HUMANS wafer fulfilled its mission by successfully reaching the lunar surface.

“If you ask a person on the street: ‘What does MIT do?’ Well, that person might say they’re a bunch of STEM nerds who make devices and create apps. But that’s not the entire MIT. It’s more multifaceted than that,” Carter says. “This project embodies that. It says, ‘We’re not just one-trick ponies.’”

A message etched in silicon

The HUMANS project, initially conceived of by MIT students, was inspired by the Golden Record, a pair of gold-plated phonograph records launched in 1977 aboard the Voyager 1 and 2 spacecraft, with human voices, music, and images. Designed to explore the outer solar system, the Voyagers have since traveled into interstellar space, beyond the sun’s heliosphere. But while the earlier project was intended to introduce humanity to an extraterrestrial audience, the HUMANS message is directed at fellow human beings — reminding us that space belongs to all.

Maya Nasr PhD ’23, now a researcher at Harvard University, has led the project since 2020, when she was a graduate student in the MIT Department of Aeronautics and Astronautics. She co-founded it with Lihui Zhang SM ’21, from the MIT Technology and Policy Program. The team invited people to share what space means to them, in writing or audio, to create a “symbol of unity that promotes global representation in space.”

When Nasr and Zhang sought an expert to translate their vision into a physical artifact, they turned to Carter, who had previously created the designs and algorithms for many art projects and, most recently, for One.MIT, a series of mosaics composed of the names of MIT faculty, students, and staff. Carter quickly agreed.

“I love figuring out how to turn equations into code, into artifacts,” Carter says. “Whether they’re art or not is a difficult question. They’re definitely artful. They’re definitely artisanal.”

Carter played a pivotal role in the computational design and fabrication of the silicon wafer now on the surface of the moon. He first translated the submitted phrases, in 64 languages, into numerical representations that could be turned into fonts. He also reverse-engineered a typesetting language to “kern” the text — adjusting the spacing between letters for visual clarity.

“Kerning is important for the aesthetics of written text. You’d want a Y to be not-too-close to a neighboring T, but farther from a W,” Carter said. “All of the phrases were sequences of words like D-O-G, and it’s not as simple as, put a D, put an O, put a G. It’s put a D, figure out where the O should be, put the O, figure out where the G should be, put the G.”

After refining the text placement, Carter designed an algorithm that geometrically transformed both the text and the audio messages’ digital waveforms — graphical representations of sound — into spirals on the wafer. The design pays homage to the Voyagers’ Golden Records, which featured spiral grooves, much like a vinyl record.

In the center of the disc is an image of a globe, or map projection — Carter found publicly available geospatial coordinates and mapped them into the design.

“I took those coordinates and then created something like an image from the coordinates. It had to be geometry, not pixels,” he says.

Once the spirals and globe imagery were in place, Carter handed the data for the design to MIT.nano, which has specialized instruments for high-precision etching and fabrication.

Human voices, lunar surface

“I hope people on Earth feel a deep sense of connection and belonging — that their voices, stories, and dreams are now part of this new chapter in lunar exploration,” Nasr says. “When we look at the moon, we can feel an even deeper connection, knowing that our words — in all their diversity — are now part of its surface, carrying the spirit of humanity forward.”

For Carter, the project conveys the human capacity for wonder and a shared sense of what’s possible. “In many cases, looking outward forces you to look inward at the same time to put the wonder in some kind of personal context,” Carter says. “So if this project somehow conveys that we are all wondering about this marvelous universe together in all of our languages, I would consider that a victory.”

The project’s link to the Golden Record — an artifact launched nearly 50 years ago and now traveling beyond the solar system — strikes another chord with Carter.

“It’s unimaginably far away, and so the notion that we can connect to something in time and space, to something that’s out there, I think it is just a wonderful connection.” 

In 1949, the U.S. Air Force called upon MIT with an urgent need. Soviet aircraft carrying atomic bombs were capable of reaching the U.S. homeland, and the nation was defenseless. A dedicated center — MIT Lincoln Laboratory — was established. The brightest minds from MIT came together in service to the nation, making scientific and engineering leaps to prototype the first real-time air defense system. The commercial sector and the U.S. Department of Defense (DoD) then produced and deployed the system, called SAGE, continent-wide.

The SAGE story still describes MIT Lincoln Laboratory’s approach to national security innovation today. The laboratory works with DoD agencies to identify challenging national security gaps, determines if technology can contribute to a solution, and then executes an R&D program to advance critical technologies. The principal products of these programs are advanced technology prototypes, which are often rapidly fabricated and demonstrated through test and evaluation.

Throughout this process, the laboratory closely coordinates with the DoD and other federal agency sponsors, and then transfers the technology in many forms to industry for manufacturing at scale to meet national needs. For nearly 75 years, these technologies have saved lives, responded to emergencies, fueled the nation’s economy, and impacted the daily life of Americans and our allies. 

"Lincoln Laboratory accelerates the pace of national security technology development, in partnership with the government, private industry, and the broader national security ecosystem," says Melissa Choi, director of MIT Lincoln Laboratory. "We integrate high-performance teams with advanced facilities and the best technology available to bring novel prototypes to life, providing lasting benefits to the United States."

The Air Force and MIT recently renewed their contract for the continued operation of Lincoln Laboratory. The contract was awarded by the Air Force Lifecycle Management Center Strategic Services Division on Hanscom Air Force Base for a term of five years, with an option for an additional five years. Since Lincoln Laboratory’s founding, MIT has operated the laboratory in the national interest for no fee and strictly on a cost-reimbursement basis. The contract award is indicative of the DoD’s continuing recognition of the long-term value of, and necessity for, cutting-edge R&D in service of national security.

Critical contributions to national security

MIT Lincoln Laboratory is the DoD’s largest federally funded research and development center R&D laboratory. Sponsored by the under secretary of defense for research and engineering, it contributes to a broad range of national security missions and domains.

Among the most critical domains are air and missile defense. Laboratory researchers pioneer advanced radar systems and algorithms crucial for detecting, tracking, and targeting ballistic missiles and aircraft, and serve as scientific advisors to the Reagan Test Site. They also conduct comprehensive studies on missile defense needs, such as the recent National Defense Authorization Act–directed study on the defense of Guam, and provide actionable insights to Congress.  

MIT Lincoln Laboratory is also at the forefront of space systems and technologies, enabling the military to monitor space activities and communicate at very high bandwidths. Laboratory engineers developed the innovatively curved detector within the Space Surveillance Telescope that allows the U.S. Space Force to track tiny space objects. It also operates the world's highest-resolution long-range radar for imaging satellites. Recently, the laboratory worked closely with NASA to demonstrate laser communications systems in space, setting a record for the fastest satellite downlink and farthest lasercom link ever achieved. These breakthroughs are heralding a new era in satellite communications for defense and civil missions.

Perhaps most importantly, MIT Lincoln Laboratory is asked to rapidly prototype solutions to urgent and emerging threats. These solutions are both transferred to industry for production and fielded directly to war-fighters, saving lives. To combat improvised explosive devices in Iraq and Afghanistan, the laboratory quickly and iteratively developed several novel systems to detect and defeat explosive devices and insurgent networks. When insurgents were attacking forward-operating bases at night, the laboratory developed an advanced infrared camera system to prevent the attacks. Like other multi-use technologies developed at the laboratory, that system led to a successful commercial startup, which was recently acquired by Anduril.

Responding to domestic crises is also a key part of the laboratory’s mission. After the attacks of 9/11/2001, the laboratory quickly integrated a system to defend the airspace around critical locations in the capital region. More recently, the laboratory’s application of AI to video forensics and physical screening has resulted in commercialized systems deployed in airports and mass transit settings. Over the last decade, the laboratory has adapted its technology for many other homeland security needs, including responses to natural disasters. As one example, researchers repurposed a world-class lidar system first used by the military for terrain mapping to quickly quantify damage after hurricanes.

For all of these efforts, the laboratory exercises responsible stewardship of taxpayer funds, identifying multiple uses for the technologies it develops and introducing disruptive approaches to reduce costs for the government. Sometimes, the system architecture or design results in cost savings, as is the case with the U.S. Air Force's SensorSat; the laboratory’s unique sensor design enabled a satellite 10 times smaller and cheaper than those typically used for space surveillance. Another approach is by creating novel systems from low-cost components. For instance, laboratory researchers discovered a way to make phased-array radars using cell phone electronics instead of traditional expensive components, greatly reducing the cost of deploying the radars for weather and aircraft surveillance.

The laboratory also pursues emerging technology to bring about transformative solutions. In the 1960s, such vision brought semiconductor lasers into the world, and in the 1990s shrunk transistors more than industry imagined possible. Today, laboratory staff are pursuing other new realms: making imagers reconfigurable at the pixel level, designing quantum sensors to transform navigation technology, and developing superconducting electronics to improve computing efficiency.

A long, beneficial relationship between MIT and the DoD

"Lincoln Laboratory has created a deep understanding and knowledge base in core national security missions and associated technologies. We look forward to continuing to work closely with government sponsors, industry, and academia through our trusted, collaborative relationships to address current and future national security challenges and ensure technological superiority," says Scott Anderson, assistant director for operations at MIT Lincoln Laboratory.

"MIT has always been proud to support the nation through its operation of Lincoln Laboratory. The long-standing relationship between MIT and the Department of Defense through this storied laboratory has been a difference-maker for the safety, economy, and industrial power of the United States, and we look forward to seeing the innovations ahead of us," notes Ian Waitz, MIT vice president for research.

Under the terms of the renewed contract, MIT will ensure that Lincoln Laboratory remains ready to meet R&D challenges that are critical to national security.

When visual information enters the brain, it travels through two pathways that process different aspects of the input. For decades, scientists have hypothesized that one of these pathways, the ventral visual stream, is responsible for recognizing objects, and that it might have been optimized by evolution to do just that.

Consistent with this, in the past decade, MIT scientists have found that when computational models of the anatomy of the ventral stream are optimized to solve the task of object recognition, they are remarkably good predictors of the neural activities in the ventral stream.

However, in a new study, MIT researchers have shown that when they train these types of models on spatial tasks instead, the resulting models are also quite good predictors of the ventral stream’s neural activities. This suggests that the ventral stream may not be exclusively optimized for object recognition.

“This leaves wide open the question about what the ventral stream is being optimized for. I think the dominant perspective a lot of people in our field believe is that the ventral stream is optimized for object recognition, but this study provides a new perspective that the ventral stream could be optimized for spatial tasks as well,” says MIT graduate student Yudi Xie.

Xie is the lead author of the study, which will be presented at the International Conference on Learning Representations. Other authors of the paper include Weichen Huang, a visiting student through MIT’s Research Summer Institute program; Esther Alter, a software engineer at the MIT Quest for Intelligence; Jeremy Schwartz, a sponsored research technical staff member; Joshua Tenenbaum, a professor of brain and cognitive sciences; and James DiCarlo, the Peter de Florez Professor of Brain and Cognitive Sciences, director of the Quest for Intelligence, and a member of the McGovern Institute for Brain Research at MIT.

Beyond object recognition

When we look at an object, our visual system can not only identify the object, but also determine other features such as its location, its distance from us, and its orientation in space. Since the early 1980s, neuroscientists have hypothesized that the primate visual system is divided into two pathways: the ventral stream, which performs object-recognition tasks, and the dorsal stream, which processes features related to spatial location.

Over the past decade, researchers have worked to model the ventral stream using a type of deep-learning model known as a convolutional neural network (CNN). Researchers can train these models to perform object-recognition tasks by feeding them datasets containing thousands of images along with category labels describing the images.

The state-of-the-art versions of these CNNs have high success rates at categorizing images. Additionally, researchers have found that the internal activations of the models are very similar to the activities of neurons that process visual information in the ventral stream. Furthermore, the more similar these models are to the ventral stream, the better they perform at object-recognition tasks. This has led many researchers to hypothesize that the dominant function of the ventral stream is recognizing objects.

However, experimental studies, especially a study from the DiCarlo lab in 2016, have found that the ventral stream appears to encode spatial features as well. These features include the object’s size, its orientation (how much it is rotated), and its location within the field of view. Based on these studies, the MIT team aimed to investigate whether the ventral stream might serve additional functions beyond object recognition.

“Our central question in this project was, is it possible that we can think about the ventral stream as being optimized for doing these spatial tasks instead of just categorization tasks?” Xie says.

To test this hypothesis, the researchers set out to train a CNN to identify one or more spatial features of an object, including rotation, location, and distance. To train the models, they created a new dataset of synthetic images. These images show objects such as tea kettles or calculators superimposed on different backgrounds, in locations and orientations that are labeled to help the model learn them.

The researchers found that CNNs that were trained on just one of these spatial tasks showed a high level of “neuro-alignment” with the ventral stream — very similar to the levels seen in CNN models trained on object recognition.

The researchers measure neuro-alignment using a technique that DiCarlo’s lab has developed, which involves asking the models, once trained, to predict the neural activity that a particular image would generate in the brain. The researchers found that the better the models performed on the spatial task they had been trained on, the more neuro-alignment they showed.

“I think we cannot assume that the ventral stream is just doing object categorization, because many of these other functions, such as spatial tasks, also can lead to this strong correlation between models’ neuro-alignment and their performance,” Xie says. “Our conclusion is that you can optimize either through categorization or doing these spatial tasks, and they both give you a ventral-stream-like model, based on our current metrics to evaluate neuro-alignment.”

Comparing models

The researchers then investigated why these two approaches — training for object recognition and training for spatial features — led to similar degrees of neuro-alignment. To do that, they performed an analysis known as centered kernel alignment (CKA), which allows them to measure the degree of similarity between representations in different CNNs. This analysis showed that in the early to middle layers of the models, the representations that the models learn are nearly indistinguishable.

“In these early layers, essentially you cannot tell these models apart by just looking at their representations,” Xie says. “It seems like they learn some very similar or unified representation in the early to middle layers, and in the later stages they diverge to support different tasks.”

The researchers hypothesize that even when models are trained to analyze just one feature, they also take into account “non-target” features — those that they are not trained on. When objects have greater variability in non-target features, the models tend to learn representations more similar to those learned by models trained on other tasks. This suggests that the models are using all of the information available to them, which may result in different models coming up with similar representations, the researchers say.

“More non-target variability actually helps the model learn a better representation, instead of learning a representation that’s ignorant of them,” Xie says. “It’s possible that the models, although they’re trained on one target, are simultaneously learning other things due to the variability of these non-target features.”

In future work, the researchers hope to develop new ways to compare different models, in hopes of learning more about how each one develops internal representations of objects based on differences in training tasks and training data.

“There could be still slight differences between these models, even though our current way of measuring how similar these models are to the brain tells us they’re on a very similar level. That suggests maybe there’s still some work to be done to improve upon how we can compare the model to the brain, so that we can better understand what exactly the ventral stream is optimized for,” Xie says.

The research was funded by the Semiconductor Research Corporation and the U.S. Defense Advanced Research Projects Agency.

As we mature from childhood, our vocabulary — as well as the ways we use it — grows, and our experiences become richer, allowing us to think, reason, and interact with others with specificity and intention. Accordingly, our word choices evolve to align with our personal values, ethics, cultural norms, and views. Over time, most of us develop an internal “guide” that enables us to learn context behind conversation; it also frequently directs us away from sharing information and sentiments that are, or could be, harmful or inappropriate. As it turns out, large language models (LLMs) — which are trained on extensive, public datasets and therefore often have biases and toxic language baked in — can gain a similar capacity to moderate their own language.

A new method from MIT, the MIT-IBM Watson AI Lab, and IBM Research, called self-disciplined autoregressive sampling (SASA), allows LLMs to detoxify their own outputs, without sacrificing fluency. 

Unlike other detoxifying methods, this decoding algorithm learns a boundary between toxic/nontoxic subspaces within the LLM’s own internal representation, without altering the parameters of the model, the need for retraining, or an external reward model. Then, during inference, the algorithm assesses the toxicity value of the partially generated phrase: tokens (words) already generated and accepted, along with each potential new token that could reasonably be chosen for proximity to the classifier boundary. Next, it selects a word option that places the phrase in the nontoxic space, ultimately offering a fast and efficient way to generate less-toxic language.

“We wanted to find out a way with any existing language model [that], during the generation process, the decoding can be subject to some human values; the example here we are taking is toxicity,” says the study’s lead author Ching-Yun “Irene” Ko PhD ’24, a former graduate intern with the MIT-IBM Watson AI Lab and a current research scientist at IBM’s Thomas J. Watson Research Center in New York.

Ko’s co-authors include Luca Daniel, professor in the MIT Department of Electrical Engineering and Computer Science (EECS), a member of the MIT-IBM Watson AI Lab, and Ko’s graduate advisor; and several members of the MIT-IBM Watson AI Lab and/or IBM Research — Pin-Yu Chen, Payel Das, Youssef Mroueh, Soham Dan, Georgios Kollias, Subhajit Chaudhury, and Tejaswini Pedapati. The work will be presented at the International Conference on Learning Representations.

Finding the “guardrails”

The training resources behind LLMs almost always include content collected from public spaces like the internet and other readily available datasets. As such, curse words and bullying/unpalatable language is a component, although some of it is in the context of literary works. It then follows that LLMs can innately produce — or be tricked into generating — dangerous and/or biased content, which often contains disagreeable words or hateful language, even from innocuous prompts. Further, it’s been found that they can learn and amplify language that’s not preferred or even detrimental for many applications and downstream tasks — leading to the need for mitigation or correction strategies.

There are many ways to achieve robust language generation that’s fair and value-aligned. Some methods use LLM retraining with a sanitized dataset, which is costly, takes time, and may alter the LLM’s performance; others employ decoding external reward models, like sampling or beam search, which take longer to run and require more memory. In the case of SASA, Ko, Daniel, and the IBM Research team developed a method that leverages the autoregressive nature of LLMs, and using a decoding-based strategy during the LLM’s inference, gradually steers the generation — one token at a time — away from unsavory or undesired outputs and toward better language.

The research group achieved this by building a linear classifier that operates on the learned subspace from the LLM’s embedding. When LLMs are trained, words with similar meanings are placed closely together in vector space and further away from dissimilar words; the researchers hypothesized that an LLM’s embedding would therefore also capture contextual information, which could be used for detoxification. The researchers used datasets that contained sets of a prompt (first half of a sentence or thought), a response (the completion of that sentence), and human-attributed annotation, like toxic or nontoxic, preferred or not preferred, with continuous labels from 0-1, denoting increasing toxicity. A Bayes-optimal classifier was then applied to learn and figuratively draw a line between the binary subspaces within the sentence embeddings, represented by positive values (nontoxic space) and negative numbers (toxic space). 

The SASA system then works by re-weighting the sampling probabilities of newest potential token based on the value of it and the generated phrase’s distance to the classifier, with the goal of remaining close to the original sampling distribution.

To illustrate, if a user is generating a potential token #12 in a sentence, the LLM will look over its full vocabulary for a reasonable word, based on the 11 words that came before it, and using top-k, top-p, it will filter and produce roughly 10 tokens to select from. SASA then evaluates each of those tokens in the partially completed sentence for its proximity to the classifier (i.e., the value of tokens 1-11, plus each potential token 12). Tokens that produce sentences in the positive space are encouraged, while those in the negative space are penalized. Additionally, the further away from the classifier, the stronger the impact.

“The goal is to change the autoregressive sampling process by re-weighting the probability of good tokens. If the next token is likely to be toxic given the context, then we are going to reduce the sampling probability for those prone to be toxic tokens,” says Ko. The researchers chose to do it this way “because the things we say, whether it’s benign or not, is subject to the context.”

Tamping down toxicity for value matching

The researchers evaluated their method against several baseline interventions with three LLMs of increasing size; all were transformers and autoregressive-based: GPT2-Large, Llama2-7b, and Llama 3.1-8b-Instruct, with 762 million, 7 billion, and 8 billion parameters respectively. For each prompt, the LLM was tasked with completing the sentence/phrase 25 times, and PerspectiveAPI scored them from 0 to 1, with anything over 0.5 being toxic. The team looked at two metrics: the average maximum toxicity score over the 25 generations for all the prompts, and the toxic rate, which was the probability of producing at least one toxic phrase over 25 generations. Reduced fluency (and therefore increased perplexity) were also analyzed. SASA was tested to complete RealToxicityPrompts (RPT), BOLD, and AttaQ datasets, which contained naturally occurring, English sentence prompts.

The researchers ramped up the complexity of their trials for detoxification by SASA, beginning with nontoxic prompts from the RPT dataset, looking for harmful sentence completions. Then, they escalated it to more challenging prompts from RPT that were more likely to produce concerning results, and as well applied SASA to the instruction-tuned model to assess if their technique could further reduce unwanted ouputs. They also used the BOLD and AttaQ benchmarks to examine the general applicability of SASA in detoxification. With the BOLD dataset, the researchers further looked for gender bias in language generations and tried to achieve a balanced toxic rate between the genders. Lastly, the team looked at runtime, memory usage, and how SASA could be combined with word filtering to achieve healthy and/or helpful language generation.

“If we think about how human beings think and react in the world, we do see bad things, so it’s not about allowing the language model to see only the good things. It’s about understanding the full spectrum — both good and bad,” says Ko, “and choosing to uphold our values when we speak and act.”

Overall, SASA achieved significant toxic language generation reductions, performing on par with RAD, a state-of-the-art external reward model technique. However, it was universally observed that stronger detoxification accompanied a decrease in fluency. Before intervention, the LLMs produced more toxic responses for female labeled prompts than male; however, SASA was able to also significantly cut down harmful responses, making them more equalized. Similarly, word filtering on top of SASA did markedly lower toxicity levels, but it also hindered the ability of the LLM to respond coherently.

A great aspect of this work is that it’s a well-defined, constrained optimization problem, says Ko, meaning that balance between open language generation that sounds natural and the need to reduce unwanted language can be achieved and tuned.

Further, Ko says, SASA could work well for multiple attributes in the future: “For human beings, we have multiple human values. We don’t want to say toxic things, but we also want to be truthful, helpful, and loyal … If you were to fine-tune a model for all of these values, it would require more computational resources and, of course, additional training.” On account of the lightweight manner of SASA, it could easily be applied in these circumstances: “If you want to work with multiple values, it’s simply checking the generation’s position in multiple subspaces. It only adds marginal overhead in terms of the compute and parameters,” says Ko, leading to more positive, fair, and principle-aligned language.

This work was supported, in part, by the MIT-IBM Watson AI Lab and the National Science Foundation.

Reindustrializing America will require action from not only businesses but also a new wave of people that have the skills, experience, and drive to make things. While many efforts in this area have focused on top-down education and manufacturing initiatives, an organic, grassroots movement has been inspiring a new generation of makers across America for the last 20 years.

The first fab lab was started in 2002 by MIT’s Center for Bits and Atoms (CBA). To teach students to use the digital fabrication research facility, CBA’s leaders began teaching a rapid-prototyping class called MAS.863 (How To Make (almost) Anything). In response to overwhelming demand, CBA collaborated with civil rights activist and MIT adjunct professor Mel King to create a community-scale version of the lab, integrating tools for 3D printing and scanning, laser cutting, precision and large-format machining, molding and casting, and surface-mount electronics, as well as design software.

That was supposed to be the end of the story; they didn’t expect a maker movement. Then another community reached out to get help building their own fab lab. Then another. Today there are hundreds of U.S. fab labs, in nearly every state, in locations ranging from community college campuses to Main Street. The fab labs offer open access to tools and software, as well as education, training, and community to people from all backgrounds.

“In the fab labs you can make almost anything,” says Professor and CBA Director Neil Gershenfeld. “That doesn’t mean everybody will make everything, but they can make things for themselves and their communities. The success of the fab labs suggests the real way to bring manufacturing back to America is not as it was. This is a different notion of agile, just-in-time manufacturing that’s personalized, distributed, and doesn’t have a sharp boundary between producer and consumer.”

Communities of makers

A fab lab opened at Florida A&M University about a year ago, but it didn’t take long for faculty and staff to notice its impact on their students. Denaria Pringley, an elementary education teacher with no experience in STEM, first came to the lab as part of a class requirement. That’s when she realized she could build her own guitar. In a pattern that has repeated itself across the country, Pringley began coming to the lab on nights and weekends, 3D-printing the body of the guitar, drilling together the neck, sanding and polishing the finish, laser engraving pick guards, and stringing everything together. Today, she works in the fab lab and knows how to run every machine in the space.

“Her entire disposition transformed through the fab lab,” says FAMU Dean of Education Sarah Price. “Every day, students make something new. There’s so much creativity going on in the lab it astounds me.”

Gershenfeld says describing how the fab labs work is a bit like describing how the internet works. At a high level, fab labs are spaces to play, create, learn, mentor, and invent. As they started replicating, Gershenfeld and his colleague Sherry Lassiter started the Fab Foundation, a nonprofit that provides operational, technical, and logistical assistance to labs. Last year, The Boston Globe called the global network of thousands of fab labs one of MIT’s most influential contributions of the last 25 years.

Some fab labs are housed in colleges. Others are funded by local governments, businesses, or through donations. Even fab labs operated in part by colleges can be open to anyone, and many of those fab labs partner with surrounding K-12 schools and continuing education programs.

Increasingly, corporate social responsibility programs are investing in fab labs, giving their communities spaces for STEM education, workforce development, and economic development. For instance, Chevron supported the startup of the fab lab at FAMU. Lassiter, the president of the Fab Foundation, notes, “Fab labs have evolved to become community anchor organizations, building strong social connections and resilience in addition to developing technical skills and providing public access to manufacturing capabilities.”

“We’re a community resource,” says Eric Saliim, who serves as a program manager at the fab lab housed in North Carolina Central University. “We have no restrictions for how you can use our fab lab. People make everything from art to car parts, products for their home, fashion accessories, you name it.”

Many fab lab instructors say the labs are a powerful way to make abstract concepts real and spark student interest in STEM subjects.

“More schools should be using fab labs to get kids interested in computer science and coding,” says Scott Simenson, former director of the fab lab at Century College in Minnesota. “This world is going to get a lot more digitally sophisticated, and we need a workforce that’s not only highly trained but also educated around subjects like computer science and artificial intelligence.”

Minnesota’s Century College opened its fab lab in 2004 amid years of declining enrollment in its engineering and design programs.

“It’s a great bridge between the theoretical and the applied,” Simenson explains. “Frankly, it helped a lot of engineering students who were disgruntled because they felt like they didn’t get to make enough things with their hands.”

The fab lab has since helped support the creation of Century College programs in digital and additive manufacturing, welding, and bioprinting.

"Working in fab labs establishes a growth mindset for our community as well as our students,” says Kelly Zelesnik, the dean of Lorain County Community College in Ohio. “Students are so under-the-gun to get it right and the grade that they lose sight of the learning. But when they’re in the fab lab, they’re iterating, because nothing ever works the first time."

In addition to offering access to equipment and education, fab labs foster education, mentorship, and innovation. Businesses often use local fab labs to make prototypes or test new products. Students have started businesses around their art and fashion creations.

Rick Pollack was a software entrepreneur and frequent visitor to the fab lab at Lorain County Community College. Pollack became fascinated with 3D printers and eventually started the additive manufacturing company MakerGear after months of tinkering with the machines in the lab in 2009. MakerGear quickly became one of the most popular producers of 3D printers in the country.

“Everyone wants to talk about innovation with STEM education and business incubation,” Gershenfeld says. “This is delivering on that by filling in the missing scaffolding: the means of production.”

Manufacturing reimagined

Many fab labs begin with tiny spaces in forgotten corners of buildings and campuses. Over time, they attract a motley crew of people that have often struggled in structured, hierarchical classroom settings. Eventually, they become hubs for people of all backgrounds driven by making.

“Fab labs provide access to tools, but what’s really driving their success is the culture of peer-to-peer, project-based learning and production,” Gershenfeld says. “Fab labs don’t separate basic and applied work, short- and long-term goals, play and problem solving. The labs are a very bottom-up distribution of the culture at MIT.”

While the local maker movement won’t replace mass manufacturing, Gershenfeld says that mass manufacturing produces goods for consumers who all want the same thing, while local production can make more interesting things that differ for individuals.

Moreover, Gershenfeld doesn’t believe you can measure the impact of fab labs by looking only at the things produced.

“A significant part of the benefit of these labs is the act of making itself,” he says. “For instance, a fab lab in Detroit led by Blair Evans worked with at-risk youth, delivering better life outcomes than conventional social services. These labs attract interest and then build skills and communities, and so along with the things that get made, the community-building, the knowledge, the connecting, is all as important as the immediate economic impact.”

MIT Professors Andrew Vanderburg and Ariel White have been honored as Committed to Caring for their attentiveness to student needs and for creating a welcoming and inclusive culture. For MIT graduate students, the Committed to Caring program recognizes those who go above and beyond.

Professor Vanderburg “is incredibly generous with his time, resources, and passion for mentoring the next generation of astronomers,” praised one of his students. 

“Professor Ariel White has made my experience at MIT immeasurably better and I hope that one day I will be in a position to pay her kindness forward,” another student credited.

Andrew Vanderburg: Investing in student growth and development

Vanderburg is the Bruno B. Rossi Career Development Assistant Professor of Physics and is affiliated with the MIT Kavli Institute for Astrophysics and Space Research. His research focuses on studying exoplanets. Vanderburg is interested in developing cutting-edge techniques and methods to discover new planets outside of our solar system, and studying these planets to learn their detailed properties.

Ever respectful of students’ boundaries between their research and personal life, Vanderburg leads by example in striking a healthy balance. A nominator commented that he has recently been working on his wildlife photography skills, and has even shared some of his photos at the group’s meetings.

Balancing personal and work life is something that almost everyone Vanderburg knows struggles with, from undergraduate students to faculty. “I encourage my group members to spend free time doing things they enjoy outside of work,” Vanderburg says, “and I try to model that balanced behavior myself.”

Vanderburg also understands and accepts that sometimes personal lives can completely overwhelm everything else and affect work and studies. He offers, “when times like these inevitably happen, I just have to acknowledge that life is unpredictable, family comes first, and that the astronomy can wait.”

In addition, Vanderburg organizes group outings, such as hiking, apple picking, and Red Sox games, and occasionally hosts group gatherings at his home. An advisee noted that “these efforts make our group feel incredibly welcoming, and fosters friendship between all our team members.”

Vanderburg has provided individualized guidance and support to over a dozen students in his first two years as faculty at MIT. His students credit him with “meeting them where they are,” and say that he candidly addresses themes like imposter syndrome and student feelings of belonging in astronomy. Vanderburg is always ready to offer his fresh perspective and unwavering support to his students.

“I try to treat everyone in my group with kindness and support,” Vanderburg says, allowing his students to trust that he has their best interest at heart. Students feel this way as well; another nominator exclaimed that Vanderburg “genuinely and truly is one of the kindest humans I know.” 

Vanderburg went above and beyond in offering his students support and insisting that his advisees will accomplish their goals. One nominator said, “his support meant the world to me at a time where I doubted my own abilities and potential.”

The Committed to Caring honor recognizes Vanderburg’s seemingly endless capacity to share his knowledge, support his students through difficult times, and invest in his mentees’ personal growth and development.

Ariel White: Student well-being and advocacy

White is an associate professor of political science who studies voting and voting rights, race, the criminal legal system, and bureaucratic behavior. Her research uses large datasets to measure individual-level experiences, and to shed light on people's everyday interactions with government. Her recent work investigates how potential voters react to experiences with punitive government policies, such as incarceration and immigration enforcement, and how people can make their way back into political life after these experiences.

She cares deeply about student well-being and departmental culture. One of her nominators shared a personal story describing that they were frequently belittled and insulted early in their graduate school journey. They had battled with whether this hurtful treatment was part of a typical grad school journey. The experience was negatively impacting their academic performance and feeling of belonging in the department. 

When she learned of it, White immediately expressed concern and reinforced that the student deserved an environment that was conducive to learning and well-being, and then quickly took steps to talk to the peer to ensure their interactions improved. 

“She wants me to feel valued, and is dedicated to both my growth as a scholar and my well-being as a person,” the nominator expressed. “This has been especially valuable as I found the adjustment to the department difficult and isolating.”

Another student commended, “I am constantly in awe of the time and effort that Ariel puts into leading by example, actively fostering an inclusive learning environment, and ensuring students feel heard and empowered.”

White is a radiant example of a professor who can have an outstanding publishing record while still treating graduate students with kindness and respect. She shows compassion and support to students, even those she does not advise. In the words of one nominator, “Ariel is the most caring person in this department.”

White has consistently expressed her desire to support her students and advocate for them. “I think one of the hardest transitions to make is the one from being a consumer of research to a producer of it.” Students work on the rather daunting prospect of developing an idea on their own for a solo project, and it can be hard to know where to start or how to keep going.

To address this, White says that she talks with advisees about what she’s seen work for her and for other students. She also encourages them to talk with their peers for advice and try out different ways of structuring their time or plan out goals.

“I try to help by explicitly highlighting these challenges and validating them: These are difficult things for nearly everyone who goes through the PhD program,” White adds.

One student reflected, “Ariel is the type of advisor that everyone should aspire to be, and that anyone would be lucky to have.” 

Tropical cyclones are hurricanes that brew over the tropical ocean and can travel over land, inundating coastal regions. The most extreme cyclones can generate devastating storm tides — seawater that is heightened by the tides and swells onto land, causing catastrophic flood events in coastal regions. A new study by MIT scientists finds that, as the planet warms, the recurrence of destructive storm tides will increase tenfold for one of the hardest-hit regions of the world.

In a study appearing today in One Earth, the scientists report that, for the highly populated coastal country of Bangladesh, what was once a 100-year event could now strike every 10 years — or more often — by the end of the century. 

In a future where fossil fuels continue to burn as they do today, what was once considered a catastrophic, once-in-a-century storm tide will hit Bangladesh, on average, once per decade. And the kind of storm tides that have occurred every decade or so will likely batter the country’s coast more frequently, every few years.

Bangladesh is one of the most densely populated countries in the world, with more than 171 million people living in a region roughly the size of New York state. The country has been historically vulnerable to tropical cyclones, as it is a low-lying delta that is easily flooded by storms and experiences a seasonal monsoon. Some of the most destructive floods in the world have occurred in Bangladesh, where it’s been increasingly difficult for agricultural economies to recover.

The study also finds that Bangladesh will likely experience tropical cyclones that overlap with the months-long monsoon season. Until now, cyclones and the monsoon have occurred at separate times during the year. But as the planet warms, the scientists’ modeling shows that cyclones will push into the monsoon season, causing back-to-back flooding events across the country.

“Bangladesh is very active in preparing for climate hazards and risks, but the problem is, everything they’re doing is more or less based on what they’re seeing in the present climate,” says study co-author Sai Ravela, principal research scientist in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS). “We are now seeing an almost tenfold rise in the recurrence of destructive storm tides almost anywhere you look in Bangladesh. This cannot be ignored. So, we think this is timely, to say they have to pause and revisit how they protect against these storms.”

Ravela’s co-authors are Jiangchao Qiu, a postdoc in EAPS, and Kerry Emanuel, professor emeritus of atmospheric science at MIT.

Height of tides

In recent years, Bangladesh has invested significantly in storm preparedness, for instance in improving its early-warning system, fortifying village embankments, and increasing access to community shelters. But such preparations have generally been based on the current frequency of storms.

In this new study, the MIT team aimed to provide detailed projections of extreme storm tide hazards, which are flooding events where tidal effects amplify cyclone-induced storm surge, in Bangladesh under various climate-warming scenarios and sea-level rise projections.

“A lot of these events happen at night, so tides play a really strong role in how much additional water you might get, depending on what the tide is,” Ravela explains.

To evaluate the risk of storm tide, the team first applied a method of physics-based downscaling, which Emanuel’s group first developed over 20 years ago and has been using since to study hurricane activity in different parts of the world. The technique involves a low-resolution model of the global ocean and atmosphere that is embedded with a finer-resolution model that simulates weather patterns as detailed as a single hurricane. The researchers then scatter hurricane “seeds” in a region of interest and run the model forward to observe which seeds grow and make landfall over time.

To the downscaled model, the researchers incorporated a hydrodynamical model, which simulates the height of a storm surge, given the pattern and strength of winds at the time of a given storm. For any given simulated storm, the team also tracked the tides, as well as effects of sea level rise, and incorporated this information into a numerical model that calculated the storm tide, or the height of the water, with tidal effects as a storm makes landfall.

Extreme overlap

With this framework, the scientists simulated tens of thousands of potential tropical cyclones near Bangladesh, under several future climate scenarios, ranging from one that resembles the current day to one in which the world experiences further warming as a result of continued fossil fuel burning. For each simulation, they recorded the maximum storm tides along the coast of Bangladesh and noted the frequency of storm tides of various heights in a given climate scenario.

“We can look at the entire bucket of simulations and see, for this storm tide of say, 3 meters, we saw this many storms, and from that you can figure out the relative frequency of that kind of storm,” Qiu says. “You can then invert that number to a return period.”

A return period is the time it takes for a storm of a particular type to make landfall again. A storm that is considered a “100-year event” is typically more powerful and destructive, and in this case, creates more extreme storm tides, and therefore more catastrophic flooding, compared to a 10-year event.

From their modeling, Ravela and his colleagues found that under a scenario of increased global warming, the storms that previously were considered 100-year events, producing the highest storm tide values, can recur every decade or less by late-century. They also observed that, toward the end of this century, tropical cyclones in Bangladesh will occur across a broader seasonal window, potentially overlapping in certain years with the seasonal monsoon season.

“If the monsoon rain has come in and saturated the soil, a cyclone then comes in and it makes the problem much worse,” Ravela says. “People won’t have any reprieve between the extreme storm and the monsoon. There are so many compound and cascading effects between the two. And this only emerges because warming happens.”

Ravela and his colleagues are using their modeling to help experts in Bangladesh better evaluate and prepare for a future of increasing storm risk. And he says that the climate future for Bangladesh is in some ways not unique to this part of the world.

“This climate change story that is playing out in Bangladesh in a certain way will be playing out in a different way elsewhere,” Ravela notes. “Maybe where you are, the story is about heat stress, or amplifying droughts, or wildfires. The peril is different. But the underlying catastrophe story is not that different.”

This research is supported in part by the MIT Climate Resilience Early Warning Systems Climate Grand Challenges project, the Jameel Observatory JO-CREWSNet project; MIT Weather and Climate Extremes Climate Grand Challenges project; and Schmidt Sciences, LLC. 

Bacteria can be engineered to sense a variety of molecules, such as pollutants or soil nutrients. In most cases, however, these signals can only be detected by looking at the cells under a microscope, making them impractical for large-scale use.

Using a new method that triggers cells to produce molecules that generate unique combinations of color, MIT engineers have shown that they can read out these bacterial signals from as far as 90 meters away. Their work could lead to the development of bacterial sensors for agricultural and other applications, which could be monitored by drones or satellites.

“It’s a new way of getting information out of the cell. If you’re standing next to it, you can’t see anything by eye, but from hundreds of meters away, using specific cameras, you can get the information when it turns on,” says Christopher Voigt, head of MIT’s Department of Biological Engineering and the senior author of the new study.

In a paper appearing today in Nature Biotechnology, the researchers showed that they could engineer two different types of bacteria to produce molecules that give off distinctive wavelengths of light across the visible and infrared spectra of light, which can be imaged with hyperspectral cameras. These reporting molecules were linked to genetic circuits that detect nearby bacteria, but this approach could also be combined with any existing sensor, such as those for arsenic or other contaminants, the researchers say.

“The nice thing about this technology is that you can plug and play whichever sensor you want,” says Yonatan Chemla, an MIT postdoc who is one of the lead authors of the paper. “There is no reason that any sensor would not be compatible with this technology.”

Itai Levin PhD ’24 is also a lead author of the paper. Other authors include former undergraduate students Yueyang Fan ’23 and Anna Johnson ’22, and Connor Coley, an associate professor of chemical engineering at MIT.

Hyperspectral imaging

There are many ways to engineer bacterial cells so that they can sense a particular chemical. Most of these work by connecting detection of a molecule to an output such as green fluorescent protein (GFP). These work well for lab studies, but such sensors can’t be measured from long distances.

For long-distance sensing, the MIT team came up with the idea to engineer cells to produce hyperspectral reporter molecules, which can be detected using hyperspectral cameras. These cameras, which were first invented in the 1970s, can determine how much of each color wavelength is present in any given pixel. Instead of showing up as simply red or green, each pixel contains information on hundreds different wavelengths of light.

Currently, hyperspectral cameras are used for applications such as detecting the presence of radiation. In the areas around Chernobyl, these cameras have been used to measure slight color changes that radioactive metals produce in the chlorophyll of plant cells. Hyperspectral cameras are also used to look for signs of malnutrition or pathogen invasion in plants.

That work inspired the MIT team to explore whether they could engineer bacterial cells to produce hyperspectral reporters when they detect a target molecule.

For a hyperspectral reporter to be most useful, it should have a spectral signature with peaks in multiple wavelengths of light, making it easier to detect. The researchers performed quantum calculations to predict the hyperspectral signatures of about 20,000 naturally occurring cell molecules, allowing them to identify those with the most unique patterns of light emission. Another key feature is the number of enzymes that would need to be engineered into a cell to get it to produce the reporter — a trait that will vary for different types of cells.

“The ideal molecule is one that’s really different from everything else, making it detectable, and requires the fewest number of enzymes to produce it in the cell,” Voigt says.

In this study, the researchers identified two different molecules that were best suited for two types of bacteria. For a soil bacterium called Pseudomonas putida, they used a reporter called biliverdin — a pigment that results from the breakdown of heme. For an aquatic bacterium called Rubrivivax gelatinosus, they used a type of bacteriochlorophyll. For each bacterium, the researchers engineered the enzymes necessary to produce the reporter into the host cell, then linked them to genetically engineered sensor circuits.

“You could add one of these reporters to a bacterium or any cell that has a genetically encoded sensor in its genome. So, it might respond to metals or radiation or toxins in the soil, or nutrients in the soil, or whatever it is you want it to respond to. Then the output of that would be the production of this molecule that can then be sensed from far away,” Voigt says.

Long-distance sensing

In this study, the researchers linked the hyperspectral reporters to circuits designed for quorum sensing, which allow cells to detect other nearby bacteria. They have also shown, in work done after this paper, that these reporting molecules can be linked to sensors for chemicals including arsenic.

When testing their sensors, the researchers deployed them in boxes so they would remain contained. The boxes were placed in fields, deserts, or on the roofs of buildings, and the cells produced signals that could be detected using hyperspectral cameras mounted on drones. The cameras take about 20 to 30 seconds to scan the field of view, and computer algorithms then analyze the signals to reveal whether the hyperspectral reporters are present.

In this paper, the researchers reported imaging from a maximum distance of 90 meters, but they are now working on extending those distances.

They envision that these sensors could be deployed for agricultural purposes such as sensing nitrogen or nutrient levels in soil. For those applications, the sensors could also be designed to work in plant cells. Detecting landmines is another potential application for this type of sensing.

Before being deployed, the sensors would need to undergo regulatory approval by the U.S. Environmental Protection Agency, as well as the U.S. Department of Agriculture if used for agriculture. Voigt and Chemla have been working with both agencies, the scientific community, and other stakeholders to determine what kinds of questions need to be answered before these technologies could be approved.

“We’ve been very busy in the past three years working to understand what are the regulatory landscapes and what are the safety concerns, what are the risks, what are the benefits of this kind of technology?” Chemla says.

The research was funded by the U.S. Department of Defense; the Army Research Office, a directorate of the U.S. Army Combat Capabilities Development Command Army Research Laboratory (the funding supported engineering of environmental strains and optimization of genetically-encoded sensors and hyperspectral reporter biosynthetic pathways); and the Ministry of Defense of Israel.

Data privacy comes with a cost. There are security techniques that protect sensitive user data, like customer addresses, from attackers who may attempt to extract them from AI models — but they often make those models less accurate.

MIT researchers recently developed a framework, based on a new privacy metric called PAC Privacy, that could maintain the performance of an AI model while ensuring sensitive data, such as medical images or financial records, remain safe from attackers. Now, they’ve taken this work a step further by making their technique more computationally efficient, improving the tradeoff between accuracy and privacy, and creating a formal template that can be used to privatize virtually any algorithm without needing access to that algorithm’s inner workings.

The team utilized their new version of PAC Privacy to privatize several classic algorithms for data analysis and machine-learning tasks.

They also demonstrated that more “stable” algorithms are easier to privatize with their method. A stable algorithm’s predictions remain consistent even when its training data are slightly modified. Greater stability helps an algorithm make more accurate predictions on previously unseen data.

The researchers say the increased efficiency of the new PAC Privacy framework, and the four-step template one can follow to implement it, would make the technique easier to deploy in real-world situations.

“We tend to consider robustness and privacy as unrelated to, or perhaps even in conflict with, constructing a high-performance algorithm. First, we make a working algorithm, then we make it robust, and then private. We’ve shown that is not always the right framing. If you make your algorithm perform better in a variety of settings, you can essentially get privacy for free,” says Mayuri Sridhar, an MIT graduate student and lead author of a paper on this privacy framework.

She is joined in the paper by Hanshen Xiao PhD ’24, who will start as an assistant professor at Purdue University in the fall; and senior author Srini Devadas, the Edwin Sibley Webster Professor of Electrical Engineering at MIT. The research will be presented at the IEEE Symposium on Security and Privacy.

Estimating noise

To protect sensitive data that were used to train an AI model, engineers often add noise, or generic randomness, to the model so it becomes harder for an adversary to guess the original training data. This noise reduces a model’s accuracy, so the less noise one can add, the better.

PAC Privacy automatically estimates the smallest amount of noise one needs to add to an algorithm to achieve a desired level of privacy.

The original PAC Privacy algorithm runs a user’s AI model many times on different samples of a dataset. It measures the variance as well as correlations among these many outputs and uses this information to estimate how much noise needs to be added to protect the data.

This new variant of PAC Privacy works the same way but does not need to represent the entire matrix of data correlations across the outputs; it just needs the output variances.

“Because the thing you are estimating is much, much smaller than the entire covariance matrix, you can do it much, much faster,” Sridhar explains. This means that one can scale up to much larger datasets.

Adding noise can hurt the utility of the results, and it is important to minimize utility loss. Due to computational cost, the original PAC Privacy algorithm was limited to adding isotropic noise, which is added uniformly in all directions. Because the new variant estimates anisotropic noise, which is tailored to specific characteristics of the training data, a user could add less overall noise to achieve the same level of privacy, boosting the accuracy of the privatized algorithm.

Privacy and stability

As she studied PAC Privacy, Sridhar hypothesized that more stable algorithms would be easier to privatize with this technique. She used the more efficient variant of PAC Privacy to test this theory on several classical algorithms.

Algorithms that are more stable have less variance in their outputs when their training data change slightly. PAC Privacy breaks a dataset into chunks, runs the algorithm on each chunk of data, and measures the variance among outputs. The greater the variance, the more noise must be added to privatize the algorithm.

Employing stability techniques to decrease the variance in an algorithm’s outputs would also reduce the amount of noise that needs to be added to privatize it, she explains.

“In the best cases, we can get these win-win scenarios,” she says.

The team showed that these privacy guarantees remained strong despite the algorithm they tested, and that the new variant of PAC Privacy required an order of magnitude fewer trials to estimate the noise. They also tested the method in attack simulations, demonstrating that its privacy guarantees could withstand state-of-the-art attacks.

“We want to explore how algorithms could be co-designed with PAC Privacy, so the algorithm is more stable, secure, and robust from the beginning,” Devadas says. The researchers also want to test their method with more complex algorithms and further explore the privacy-utility tradeoff.

“The question now is: When do these win-win situations happen, and how can we make them happen more often?” Sridhar says.

“I think the key advantage PAC Privacy has in this setting over other privacy definitions is that it is a black box — you don’t need to manually analyze each individual query to privatize the results. It can be done completely automatically. We are actively building a PAC-enabled database by extending existing SQL engines to support practical, automated, and efficient private data analytics,” says Xiangyao Yu, an assistant professor in the computer sciences department at the University of Wisconsin at Madison, who was not involved with this study.

This research is supported, in part, by Cisco Systems, Capital One, the U.S. Department of Defense, and a MathWorks Fellowship.

“No cash prizes. But our friends in Kiev are calling in, and they’ll probably say thanks,”​ was the the tagline that drew students and tech professionals to join MIT-Ukraine’s first-ever hackathon this past January.

The hackathon was co-sponsored by MIT-Ukraine and Mission Innovation X and was shaped by the efforts of MIT alumni from across the world. It was led by Hosea Siu ’14, SM ’15, PhD ’18, a seasoned hackathon organizer and AI researcher, in collaboration with Phil Tinn MCP ’16, a research engineer now based at SINTEF [Foundation for Industrial and Technical Research] in Norway. The program was designed to prioritize tangible impact: 

“In a typical hackathon, you might get a weekend of sleepless nights and some flashy but mostly useless prototypes. Here, we stretched it out over four weeks, and we’re expecting real, meaningful outcomes,”​ says Siu, the hackathon director.

One week of training, three weeks of project development

In the first week, participants attended lectures with leading experts on key challenges Ukraine currently faces, from a talk on mine contamination with Andrew Heafitz PhD ’05 to a briefing on disinformation with Nina Lutz SM ’21. Then, participants formed teams to develop projects addressing these challenges, with mentorship from top MIT specialists including Phil Tinn (AI & defense), Svetlana Boriskina (energy resilience), and Gene Keselman (defense innovation and dual-use technology).

“I really liked the solid structure they gave us — walking us through exactly what’s happening in Ukraine, and potential solutions,”​ says Timur Gray, a first-year in engineering at Olin College.

The five final projects spanned demining, drone technology, AI and disinformation, education for Ukraine, and energy resilience. 

Supporting demining efforts

With current levels of technology, it is estimated that it will take 757 years to fully de-mine Ukraine. Students Timur Gray and Misha Donchenko, who is a sophomore mathematics major at MIT, came together to research the latest developments in demining technology and strategize how students could most effectively support innovations.

The team has made connections with the Ukrainian Association of Humanitarian Demining and the HALO Trust to explore opportunities for MIT students to directly support demining efforts in Ukraine. They also explored project ideas to work on tools for civilians to report on mine locations, and the team created a demo web page рішучість757, which includes an interactive database mapping mine locations.

“Being able to apply my skills to something that has a real-world impact — that’s been the best part of this hackathon,” says Donchenko.

Innovating drone production

Drone technology has been one of Ukraine’s most critical advantages on the battlefield — but government bureaucracy threatens to slow innovation, according to Oleh Deineka, who made this challenge the focus of his hackathon project. 

Joining remotely from Ukraine, where he studies post-war recovery at the Kyiv School of Economics, Deineka brought invaluable firsthand insight from living and working on the ground, enriching the experience for all participants. Prior to the hackathon, he had already begun developing UxS.AGENCY, a secure digital platform to connect drone developers with independent funders, with the aim of ensuring that the speed of innovations in drone technology is not curbed. 

He notes that Ukrainian arms manufacturers have the capacity to produce three times more weapons and military equipment than the Ukrainian government can afford to purchase. Promoting private sector development of drone production could help solve this. The platform Deineka is working on also aims to reduce the risk of corruption, allowing developers to work directly with funders, bypassing any bureaucratic interference.

Deineka is also working with MIT’s Keselman, who gave a talk during the hackathon on dual-use technology — the idea that military innovations should also have civilian applications. Deineka emphasized that developing such dual-use technology in Ukraine could help not only to win the war, but also to create sustainable civilian applications, ensuring that Ukraine’s 10,000 trained drone operators have jobs after it ends. He pointed to future applications such as drone-based urban infrastructure monitoring, precision agriculture, and even personal security — like a small drone following a child with asthma, allowing parents to monitor their well-being in real time​.

“This hackathon has connected me with MIT’s top minds in innovation and security. Being invited to collaborate with Gene Keselman and others has been an incredible opportunity," says Deineka.

Disinformation dynamics on Wikipedia

Wikipedia has long been a battleground for Russian disinformation, from the profiling of artists like Kazimir Malevich to the framing of historical events. The hackathon’s disinformation team worked together on a machine learning-based tool to detect biased edits. 

They found that Wikipedia’s moderation system is susceptible to reinforcing systemic bias, particularly when it comes to history. Their project laid the groundwork for a potential student-led initiative to track disinformation, propose corrections, and develop tools to improve fact-checking on Wikipedia.

Education for Ukraine’s future

Russia’s war against Ukraine is having a detrimental impact on education, with constant air raid sirens disrupting classes, and over 2,000 Ukrainian schools damaged or destroyed. The STEM education team focused on what they could do to support Ukrainian students. They developed a plan for adapting MIT’s Beaver Works Summer Institute in STEM for students still living in Ukraine, or potentially for Ukrainians currently displaced to neighboring countries. 

“I didn’t realize how many schools had been destroyed and how deeply that could impact kids’ futures. You hear about the war, but the hackathon made it real in a way I hadn’t thought about before,” says Catherine Tang, a senior in electrical engineering and computer science.

Vlad Duda, founder of Nomad AI, also contributed to the education track of the hackathon with a focus on language accessibility and learning support. One of the prototypes he presented, MOVA, is a Chrome extension that uses AI to translate online resources into Ukrainian — an especially valuable tool for high school students in Ukraine, who often lack the English proficiency needed to engage with complex academic content. Duda also developed OpenBookLM, an AI-powered tool that helps students turn notes into audio and personalized study guides, similar in concept to Google’s NotebookLM but designed to be open-source and adaptable to different languages and educational contexts.

Energy resilience 

The energy resilience team worked on exploring cheaper, more reliable heating and cooling technologies so Ukrainian homes can be less dependent on traditional energy grids that are susceptible to Russian attacks.

The team tested polymer filaments that generate heat when stretched and cool when released, which could potentially offer low-cost, durable home heating solutions in Ukraine. Their work focused on finding the most effective braid structure to enhance durability and efficiency.

From hackathon to reality

Unlike most hackathons, where projects end when the event does, MIT-Ukraine’s goal is to ensure these ideas don’t stop here. All the projects developed during the hackathon will be considered as potential avenues for MIT’s Undergraduate Research Opportunities Program (UROP) and MISTI Ukraine summer internship programs. Last year, 15 students worked on UROP and MISTI projects for Ukraine, contributing in areas such as STEM education and reconstruction in Ukraine. With the many ideas generated during the hackathon, MIT-Ukraine is committed to expanding opportunities for student-led projects and collaborations in the coming year.

"The MIT-Ukraine program is about learning by doing, and making an impact beyond MIT’s campus. This hackathon proved that students, researchers, and professionals can work together to develop solutions that matter — and Ukraine’s urgent challenges demand nothing less," says Elizabeth Wood, Ford International Professor of History at MIT and the faculty director of the MIT-Ukraine Program at the Center for International Studies. 

For the past decade, the Abdul Latif Jameel Water and Food Systems Lab (J-WAFS) has been instrumental in promoting student engagement across the Institute to help solve the world’s most pressing water and food system challenges. As part of J-WAFS’ central mission of securing the world’s water and food supply, J-WAFS aims to cultivate the next generation of leaders in the water and food sectors by encouraging MIT student involvement through a variety of programs and mechanisms that provide research funding, mentorship, and other types of support.

J-WAFS offers a range of opportunities for both undergraduate and graduate students to engage in the advancement of water and food systems research. These include graduate student fellowships, travel grants for participation in conferences, funding for research projects in India, video competitions highlighting students’ water and food research, and support for student-led organizations and initiatives focused on critical areas in water and food.

As J-WAFS enters its second decade, it continues to expose students across the Institute to experiential hands-on water and food research, career and other networking opportunities, and a platform to develop their innovative and collaborative solutions.

Graduate student fellowships

In 2017, J-WAFS inaugurated two graduate student fellowships: the Rasikbhai L. Meswani Fellowship for Water Solutions and the J-WAFS Graduate Student Fellowship Program. The Rasikbhai L. Meswani Fellowship for Water Solutions is a doctoral fellowship for students pursuing research related to water for human need at MIT. The fellowship is made possible by Elina and Nikhil Meswani and family. Each year, up to two outstanding students are selected to receive fellowship support for one academic semester. Through it, J-WAFS seeks to support distinguished MIT students who are pursuing solutions to the pressing global water supply challenges of our time. The J-WAFS Fellowship for Water and Food Solutions is funded by the J-WAFS Research Affiliate Program, which offers companies the opportunity to collaborate with MIT on water and food research. A portion of each research affiliate’s fees supports this fellowship.

Aditya Avinash Ghodgaonkar, a PhD student in the Department of Mechanical Engineering (MechE), reflects on how receiving a J-WAFS graduate student fellowship positively impacted his research on the design of low-cost emitters for affordable, resilient drip irrigation for farmers: “My J-WAFS fellowship gave me the flexibility and financial support needed to explore new directions in the area of clog-resistant drip irrigation that had a higher risk element that might not have been feasible to manage on an industrially sponsored project,” Ghodgaonkar explains. Emitters, which control the volume and flow rate of water used during irrigation, often clog due to small particles like sand. Ghodgaonkar worked with Professor Amos Winter, and with farmers in resource-constrained communities in countries like Jordan and Morocco, to develop an emitter that is mechanically more resistant to clogging. Ghodgaonkar reports that their energy-efficient, compact, clog-resistant drip emitters are being commercialized by Toro and may be available for retail in the next few years. The opportunities and funding support Ghodgaonkar has received from J-WAFS contributed greatly to his entrepreneurial success and the advancement of the water and agricultural sectors.

Linzixuan (Rhoda) Zhang, a PhD student advised by Professor Robert Langer and Principal Research Scientist Ana Jaklenec of the Department of Chemical Engineering, was a 2022 J-WAFS Graduate Student Fellow. With the fellowship, Zhang was able to focus on her innovative research on a novel micronutrient delivery platform that fortifies food with essential vitamins and nutrients. “We intake micronutrients from basically all the healthy food that we eat; however, around the world there are about 2 billion people currently suffering from micronutrient deficiency because they do not have access to very healthy, very fresh food,” Zhang says. Her research involves the development of biodegradable polymers that can deliver these micronutrients in harsh environments in underserved regions of the world. “Vitamin A is not very stable, for example; we have vitamin A in different vegetables but when we cook them, the vitamin can easily degrade,” Zhang explains. However, when vitamin A is encapsulated in the microparticle platform, simulation of boiling and of the stomach environment shows that vitamin A was stabilized. “The meaningful factors behind this experiment are real,” says Zhang. The J-WAFS Fellowship helped position Zhang to win the 2024 Collegiate Inventors Competition for this work.

J-WAFS grant for water and food projects in India

J-WAFS India Grants are intended to further the work being pursued by MIT individuals as a part of their research, innovation, entrepreneurship, coursework, or related activities. Faculty, research staff, and undergraduate and graduate students are eligible to apply. The program aims to support projects that will benefit low-income communities in India, and facilitates travel and other expenses related to directly engaging with those communities.

Gokul Sampath, a PhD student in the Department of Urban Studies and Planning, and Jonathan Bessette, a PhD student in MechE, initially met through J-WAFS-sponsored conference travel, and discovered their mutual interest in the problem of arsenic in water in India. Together, they developed a cross-disciplinary proposal that received a J-WAFS India Grant. Their project is studying how women in rural India make decisions about where they fetch water for their families, and how these decisions impact exposure to groundwater contaminants like naturally-occurring arsenic. Specifically, they are developing low-cost remote sensors to better understand water-fetching practices. The grant is enabling Sampath and Bessette to equip Indian households with sensor-enabled water collection devices (“smart buckets”) that will provide them data about fetching practices in arsenic-affected villages. By demonstrating the efficacy of a sensor-based approach, the team hopes to address a major data gap in international development. “It is due to programs like the Jameel Water and Food Systems Lab that I was able to obtain the support for interdisciplinary work on connecting water security, public health, and regional planning in India,” says Sampath.

J-WAFS travel grants for water conferences

In addition to funding graduate student research, J-WAFS also provides grants for graduate students to attend water conferences worldwide. Typically, students will only receive travel funding to attend conferences where they are presenting their research. However, the J-WAFS travel grants support learning, networking, and career exploration opportunities for exceptional MIT graduate students who are interested in a career in the water sector, whether in academia, nonprofits, government, or industry.

Catherine Lu ’23, MNG ’24 was awarded a 2023 Travel Grant to attend the UNC Water and Health Conference in North Carolina. The conference serves as a curated space for policymakers, practitioners, and researchers to convene and assess data, scrutinize scientific findings, and enhance new and existing strategies for expanding access to and provision of services for water, sanitation, and hygiene (WASH). Lu, who studied civil and environmental engineering, worked with Professor Dara Entekhabi on modeling and predicting droughts in Africa using satellite Soil Moisture Active Passive (SMAP) data. As she evaluated her research trajectory and career options in the water sector, Lu found the conference to be informative and enlightening. “I was able to expand my knowledge on all the sectors and issues that are related to water and the implications they have on my research topic.” Furthermore, she notes: “I was really impressed by the diverse range of people that were able to attend the conference. The global perspective offered at the conference provided a valuable context for understanding the challenges and successes of different regions around the world — from WASH education in schools in Zimbabwe and India to rural water access disparities in the United States … Being able to engage with such passionate and dedicated people has motivated me to continue progress in this sector.” Following graduation, Lu secured a position as a water resources engineer at CDM Smith, an engineering and construction firm.

Daniela Morales, a master’s student in city planning in the Department of Urban Studies and Planning, was a 2024 J-WAFS Travel Grant recipient who attended World Water Week in Stockholm, Sweden. The annual global conference is organized by the Stockholm International Water Institute and convenes leading experts, decision-makers, and professionals in the water sector to actively engage in discussions and developments addressing critical water-related challenges. Morales’ research interests involve drinking water quality and access in rural and peri-urban areas affected by climate change impacts, the effects of municipal water shutoffs on marginalized communities, and the relationship between regional water management and public health outcomes. When reflecting on her experience at the conference, Morales writes: “Being part of this event has given me so much motivation to continue my professional and academic journey in water management as it relates to public health and city planning … There was so much energy that was collectively generated in the conference, and so many new ideas that I was able to process around my own career interests and my role as a future planner in water management, that the last day of the conference felt less like an ending and more of the beginning of a new chapter. I am excited to take all the information I learned to work towards my own research, and continue to build relationships with all the new contacts I made.” Morales also notes that without the support of the J-WAFS grant, “I would not have had the opportunity to make it to Stockholm and participate in such a unique week of water wisdom.”

Seed grants and Solutions grants

J-WAFS offers seed grants for early-stage research and Solutions Grants for later-stage research that is ready to move from the lab to the commercial world. Proposals for both types of grants must be submitted and led by an MIT principal investigator, but graduate students, and sometimes undergraduates, are often supported by these grants.

Arjav Shah, a PhD-MBA student in MIT’s Department of Chemical Engineering and the MIT Sloan School of Management, is currently pursuing the commercialization of a water treatment technology that was first supported through a 2019 J-WAFS seed grant and then a 2022 J-WAFS Solutions Grant with Professor Patrick Doyle. The technology uses hydrogels to remove a broad range of micropollutants from water. The Solutions funding enables entrepreneurial students and postdocs to lay the groundwork to commercialize a technology by assessing use scenarios and exploring business needs with actual potential customers. “With J-WAFS’ support, we were not only able to scale up the technology, but also gain a deeper understanding of market needs and develop a strong business case,” says Shah. Shah and the Solutions team have discovered that the hydrogels could be used in several real-world contexts, ranging from large-scale industrial use to small-scale, portable, off-grid applications. “We are incredibly grateful to J-WAFS for their support, particularly in fostering industry connections and facilitating introductions to investors, potential customers, and experts,” Shah adds.

Shah was also a 2023 J-WAFS Travel Grant awardee who attended Stockholm World Water Week that year. He says, “J-WAFS has played a pivotal role in both my academic journey at MIT and my entrepreneurial pursuits. J-WAFS support has helped me grow both as a scientist and an aspiring entrepreneur. The exposure and opportunities provided have allowed me to develop critical skills such as customer discovery, financial modeling, business development, fundraising, and storytelling — all essential for translating technology into real-world impact. These experiences provided invaluable insights into what it takes to bring a technology from the lab to market.”

Shah is currently leading efforts to spin out a company to commercialize the hydrogel research. Since receiving J-WAFS support, the team has made major strides toward launching a startup company, including winning the Pillar VC Moonshot Prize, Cleantech Open National Grand Prize, MassCEC Catalyst Award, and participation in the NSF I-Corps National Program.

J-WAFS student video competitions

J-WAFS has hosted two video competitions: MIT Research for a Water Secure Future and MIT Research for a Food Secure Future, in honor of World Water Day and Word Food Day, respectively. In these competitions, students are tasked with creating original videos showcasing their innovative water and food research conducted at MIT. The opportunity is open to MIT students, postdocs, and recent alumni.

Following a review by a distinguished panel of judges, Vishnu Jayaprakash SM ’19, PhD ’22 won first place in the 2022 J-WAFS World Food Day Student Video Competition for his video focused on eliminating pesticide pollution and waste. Jayaprakash delved into the science behind AgZen-Cloak, a new generation of agricultural sprays that prevents pesticides from bouncing off of plants and seeping into the ground, thus causing harmful runoff. The J-WAFS competition provided Jayaprakash with a platform to highlight the universal, low-cost, and environmentally sustainable benefits of AgZen-Cloak. Jayaprakash worked on similar technology as a funded student on a J-WAFS Solutions grant with Professor Kripa Varanasi. The Solutions grant, in fact, helped Jayaprakash and Varanasi to launch AgZen, a company that deploys AgZen-Cloak and other products and technologies to control the interactions of droplets and sprays with crop surfaces. AgZen is currently helping farmers sustainably tend to their agricultural plots while also protecting the environment.  

In 2021, Hilary Johnson SM ’18, PhD ’22, won first place in the J-WAFS World Water Day video competition. Her video highlighted her work on a novel pump that uses adaptive hydraulics for improved pump efficiency. The pump was part of a sponsored research project with Xylem Inc., a J-WAFS Research Affiliate company, and Professor Alex Slocum of MechE. At the time, Johnson was a PhD student in Slocum’s lab. She was instrumental in the development of the pump by engineering the volute to expand and contract to meet changing system flow rates. Johnson went on to later become a 2021-22 J-WAFS Fellow, and is now a full-time mechanical engineer at the Lawrence Livermore National Laboratory.

J-WAFS-supported student clubs

J-WAFS-supported student clubs provide members of the MIT student community the opportunity for networking and professional advancement through events focused on water and food systems topics.

J-WAFS is a sponsor of the MIT Water Club, a student-led group that supports and promotes the engagement of the MIT community in water-sector-related activism, dissemination of information, and research innovation. The club allows students to spearhead the organization of conferences, lectures, outreach events, research showcases, and entrepreneurship competitions including the former MIT Water Innovation Prize and MIT Water Summit. J-WAFS not only sponsors the MIT Water Club financially, but offers mentorship and guidance to the leadership team.

The MIT Food and Agriculture Club is also supported by J-WAFS. The club’s mission is to promote the engagement of the MIT community in food and agriculture-related topics. In doing so, the students lead initiatives to share the innovative technology and business solutions researchers are developing in food and agriculture systems. J-WAFS assists in the connection of passionate MIT students with those who are actively working in the food and agriculture industry beyond the Institute. From 2015 to 2022, J-WAFS also helped the club co-produce the Rabobank-MIT Food and Agribusiness Innovation Prize — a student business plan competition for food and agricultural startups.

From 2023 onward, the MIT Water Club and the MIT Food and Ag Club have been joining forces to organize a combined prize competition: The MIT Water, Food and Agriculture (WFA) Innovation Prize. The WFA Innovation Prize is a business plan competition for student-led startups focused on any region or market. The teams present business plans involving a technology, product, service, or process that is aimed at solving a problem related to water, food, or agriculture. The competition encourages all approaches to innovation, from engineering and product design to policy and data analytics. The goal of the competition is to help emerging entrepreneurs translate research and ideas into businesses, access mentors and resources, and build networks in the water, food, and agriculture industries. J-WAFS offers financial and in-kind support, working with student leaders to plan, organize, and implement the stages of the competition through to the final pitch event. This year, J-WAFS is continuing to support the WFA team, which is led by Ali Decker, an MBA student at MIT Sloan, and Sam Jakshtis, a master’s student in MIT’s science in real estate development program. The final pitch event will take place on April 30 in the MIT Media Lab.

“I’ve had the opportunity to work with Renee Robins, executive director of J-WAFS, on MIT’s Water, Food and Agriculture Innovation Prize for the past two years, and it has been both immensely valuable and a delight to have her support,” says Decker. “Renee has helped us in all areas of prize planning: brainstorming new ideas, thinking through startup finalist selection, connecting to potential sponsors and partners, and more. Above all, she supports us with passion and joy; each time we meet, I look forward to our discussion,” Decker adds.

J-WAFS events

Throughout the year, J-WAFS aims to offer events that will engage any in the MIT student community who are working in water or food systems. For example, on April 19, 2023, J-WAFS teamed up with the MIT Energy Initiative (MITEI) and the Environmental Solutions Initiative (ESI) to co-host an MIT student poster session for Earth Month. The theme of the poster session was “MIT research for a changing planet,” and it featured work from 11 MIT students with projects in water, food, energy, and the environment. The students, who represented a range of MIT departments, labs, and centers, were on hand to discuss their projects and engage with those attending the event. Attendees could vote for their favorite poster after being asked to consider which poster most clearly communicated the research problem and the potential solution. At the end of the night, votes were tallied and the winner of the “People’s Choice Award” for best poster was Elaine Liu ’24, an undergraduate in mathematics at the time of the event. Liu’s poster featured her work on managing failure cascades in systems with wind power.

J-WAFS also hosts less-structured student networking events. For instance, during MIT’s Independent Activities Period (IAP) in January 2024, J-WAFS hosted an ice cream social for student networking. The informal event was an opportunity for graduate and undergraduate students from across the Institute to meet and mingle with like-minded peers working in, or interested in, water and food systems. Students were able to explain their current and future research, interests, and projects and ask questions while exchanging ideas, engaging with one another, and potentially forming collaborations, or at the very least sharing insights.

Looking ahead to 10 more years of student impact

Over the past decade, J-WAFS has demonstrated a strong commitment to empowering students in the water and food sectors, fostering an environment where they can confidently drive meaningful change and innovation. PhD student Jonathan Bessette sums up the J-WAFS community as a “one-of-a-kind community that enables essential research in water and food that otherwise would not be pursued. It’s this type of research that is not often the focus of major funding, yet has such a strong impact in sustainable development.”

J-WAFS aims to provide students with the support and tools they need to conduct authentic and meaningful water and food-related research that will benefit communities around the world. This support, coupled with an MIT education, enables students to become leaders in sustainable water and food systems. As the second decade of J-WAFS programming begins, the J-WAFS team remains committed to fostering student collaboration across the Institute, driving innovative solutions to revitalize the world’s water and food systems while empowering the next generation of pioneers in these critical fields. 

MIT graduate students Sreekar Mantena and Arjun Ramani, and recent MIT alumni Rupert Li ’24 and Jupneet Singh ’23, have been named 2025 P.D. Soros Fellows. In addition, Soros Fellow Andre Ye will begin a PhD in computer science at MIT this fall.

Each year, the P.D. Soros Fellowship for New Americans awards 30 outstanding immigrants and children of immigrants $90,000 in graduate school financial support over a two-year period. The merit-based program selects fellows based on their achievements, potential to make meaningful contributions to their fields and communities, and dedication to the ideals of the United States represented in the Bill of Rights and the Constitution. This year’s fellows were selected from a competitive pool of more than 2,600 applicants nationwide.

Rupert Li ’24

The son of Chinese immigrants, Rupert Li was born and raised in Portland, Oregon. He graduated from MIT in 2024 with a double major in mathematics and computer science, economics, and data science, and earned an MEng in the latter subject.

Li was named a Marshall Scholar in 2023 and is currently pursuing a master’s degree in the Part III mathematics program at Cambridge University. His P.D. Soros Fellowship will support his pursuit of a PhD in mathematics at Stanford University.

Li’s first experience with mathematics research was as a high school student participant in the MIT PRIMES-USA program. He continued research in mathematics as an undergraduate at MIT, where he worked with professors Henry Cohn, Nike Sun, and Elchanan Mossel in the Department of Mathematics. Li also spent two summers at the Duluth REU (Research Experience for Undergraduates) program with Professor Joe Gallian.

Li’s research in probability, discrete geometry, and combinatorics culminated in him receiving the Barry Goldwater Scholarship, an honorable mention for the Frank and Brennie Morgan Prize for Outstanding Research in Mathematics by an Undergraduate Student, the Marshall Scholarship, and the Hertz Fellowship.

Beyond research, Li finds fulfillment in opportunities to give back to the math community that has supported him throughout his mathematical journey. This year marks the second time he has served as a graduate student mentor for the PRIMES-USA program, which sparked his mathematical career, and his first year as an advisor for the Duluth REU program.

Sreekar Mantena

Sreekar Mantena graduated Phi Beta Kappa from Harvard College with a degree in statistics and molecular biology. He is currently an MD student in biomedical informatics in the Harvard-MIT Program in Health Sciences and Technology (HST), where he works under Professor Soumya Raychaudhuri of the Broad Institute of MIT and Harvard. He is also pursuing a PhD in bioinformatics and integrative genomics at Harvard Medical School. In the future, Mantena hopes to blend compassion with computation as a physician-scientist who harnesses the power of machine learning and statistics to advance equitable health care delivery.

The son of Indian-American immigrants, Mantena was raised in North Carolina, where he grew up as fond of cheese grits as of his mother’s chana masala. Every summer of his childhood, he lived with his grandparents in Southern India, who instilled in him the importance of investing in one’s community and a love of learning.

As an undergraduate at Harvard, Mantena was inspired by the potential of statistics and data science to address gaps in health-care delivery. He founded the Global Alliance for Medical Innovation, a nonprofit organization that has partnered with physicians in six countries to develop data-driven medical technologies for underserved communities, including devices to detect corneal disease.

Mantena also pursued research in Professor Pardis Sabeti’s lab at the Broad Institute, where he built new algorithms to design diagnostic assays that improve the detection of infectious pathogens in resource-limited settings. He has co-authored over 20 scientific publications, and his lead-author work has been published in many journals, including Nature Biotechnology, The Lancet Digital Health, and the Journal of Pediatrics.

Arjun Ramani

Arjun Ramani, from West Lafayette, Indiana, is the son of immigrants from Tamil Nadu, India. He is currently pursuing a PhD in economics at MIT, where he studies technological change and innovation. He hopes his research can inform policies and business practices that generate broadly shared economic growth.

Ramani’s dual interests in technology and the world led him to Stanford University, where he studied economics as an undergraduate and pursued a master’s in computer science, specializing in artificial intelligence. As data editor of the university’s newspaper, he started the Stanford Open Data Project to improve campus data transparency. During college, Ramani also spent time at the White House working on economic policy, in Ghana helping startups scale, and at Citadel in financial markets — all of which cultivated a broad interest in the economic world.

After graduating from Stanford, Ramani became The Economist’s global business and economics correspondent. He first covered technology and finance and later shifted to covering artificial intelligence after the technology took the world by storm in 2022.

In 2023, Ramani moved to India to cover the Indian economy in the lead-up to its election. There, he gained a much deeper appreciation for the social and institutional barriers that slowed technology adoption and catch-up growth. Ramani wrote or co-wrote six cover stories, was shortlisted for U.K. financial journalist of the year in 2024 for his AI and economics reporting, and co-authored a six-part special report on India’s economy.

Jupneet Singh ’23

Jupneet Singh, the daughter of Indian immigrants, is a Sikh-American who grew up deeply connected to her Punjabi and Sikh heritage in Somis, California. The Soros Fellowship will support her MD studies at Harvard Medical School’s HST program under the U.S. Air Force Health Professions Scholarship Program.

Singh plans to complete her medical residency as an active-duty U.S. Air Force captain, and after serving as a surgeon in the USAF she hopes to enter the United States Public Health Commissioned Corps. While Singh is the first in her family to serve in the U.S. armed services, she is proud to be carrying on a long Sikh military legacy.

Singh graduated from MIT in 2023 with a degree in chemistry and a concentration in history and won a Rhodes Scholarship to pursue two degrees at the University of Oxford: a master’s in public policy and a master’s in translational health sciences. At MIT, she served as the commander (highest-ranked cadet) of the Air Force ROTC Detachment and is now commissioned as a 2nd Lieutenant. She is the first woman Air Force ROTC Rhodes Scholar.

Singh has worked in de-addiction centers in Punjab, India. She also worked at the Ventura County Family Justice Center and Ventura County Medical Center Trauma Center, and published a first-author paper in The American Surgeon. She founded Pathways to Promise, a program to support the health of children affected by domestic violence. She has conducted research on fatty liver disease under Professor Alex Shalek at MIT and on maternal health inequalities at the National Perinatal Epidemiological Unit at Oxford.

Insect-scale robots can squeeze into places their larger counterparts can’t, like deep into a collapsed building to search for survivors after an earthquake.

However, as they move through the rubble, tiny crawling robots might encounter tall obstacles they can’t climb over or slanted surfaces they will slide down. While aerial robots could avoid these hazards, the amount of energy required for flight would severely limit how far the robot can travel into the wreckage before it needs to return to base and recharge.

To get the best of both locomotion methods, MIT researchers developed a hopping robot that can leap over tall obstacles and jump across slanted or uneven surfaces, while using far less energy than an aerial robot.

The hopping robot, which is smaller than a human thumb and weighs less than a paperclip, has a springy leg that propels it off the ground, and four flapping-wing modules that give it lift and control its orientation.

The robot can jump about 20 centimeters into the air, or four times its height, at a lateral speed of about 30 centimeters per second, and has no trouble hopping across ice, wet surfaces, and uneven soil, or even onto a hovering drone. All the while, the hopping robot consumes about 60 percent less energy than its flying cousin.

Due to its light weight and durability, and the energy efficiency of the hopping process, the robot could carry about 10 times more payload than a similar-sized aerial robot, opening the door to many new applications.

“Being able to put batteries, circuits, and sensors on board has become much more feasible with a hopping robot than a flying one. Our hope is that one day this robot could go out of the lab and be useful in real-world scenarios,” says Yi-Hsuan (Nemo) Hsiao, an MIT graduate student and co-lead author of a paper on the hopping robot.

Hsiao is joined on the paper by co-lead authors Songnan Bai, a research assistant professor at The University of Hong Kong; and Zhongtao Guan, an incoming MIT graduate student who completed this work as a visiting undergraduate; as well as Suhan Kim and Zhijian Ren of MIT; and senior authors Pakpong Chirarattananon, an associate professor of the City University of Hong Kong; and Kevin Chen, an associate professor in the MIT Department of Electrical Engineering and Computer Science and head of the Soft and Micro Robotics Laboratory within the Research Laboratory of Electronics. The research appears today in Science Advances.

Maximizing efficiency

Jumping is common among insects, from fleas that leap onto new hosts to grasshoppers that bound around a meadow. While jumping is less common among insect-scale robots, which usually fly or crawl, hopping affords many advantages for energy efficiency.

When a robot hops, it transforms potential energy, which comes from its height off the ground, into kinetic energy as it falls. This kinetic energy transforms back to potential energy when it hits the ground, then back to kinetic as it rises, and so on.

To maximize efficiency of this process, the MIT robot is fitted with an elastic leg made from a compression spring, which is akin to the spring on a click-top pen. This spring converts the robot’s downward velocity to upward velocity when it strikes the ground.

“If you have an ideal spring, your robot can just hop along without losing any energy. But since our spring is not quite ideal, we use the flapping modules to compensate for the small amount of energy it loses when it makes contact with the ground,” Hsiao explains.

As the robot bounces back up into the air, the flapping wings provide lift, while ensuring the robot remains upright and has the correct orientation for its next jump. Its four flapping-wing mechanisms are powered by soft actuators, or artificial muscles, that are durable enough to endure repeated impacts with the ground without being damaged.

“We have been using the same robot for this entire series of experiments, and we never needed to stop and fix it,” Hsiao adds.

Key to the robot’s performance is a fast control mechanism that determines how the robot should be oriented for its next jump. Sensing is performed using an external motion-tracking system, and an observer algorithm computes the necessary control information using sensor measurements.

As the robot hops, it follows a ballistic trajectory, arcing through the air. At the peak of that trajectory, it estimates its landing position. Then, based on its target landing point, the controller calculates the desired takeoff velocity for the next jump. While airborne, the robot flaps its wings to adjust its orientation so it strikes the ground with the correct angle and axis to move in the proper direction and at the right speed.

Durability and flexibility

The researchers put the hopping robot, and its control mechanism, to the test on a variety of surfaces, including grass, ice, wet glass, and uneven soil — it successfully traversed all surfaces. The robot could even hop on a surface that was dynamically tilting.

“The robot doesn’t really care about the angle of the surface it is landing on. As long as it doesn’t slip when it strikes the ground, it will be fine,” Hsiao says.

Since the controller can handle multiple terrains, the robot can easily transition from one surface to another without missing a beat.

For instance, hopping across grass requires more thrust than hopping across glass, since blades of grass cause a damping effect that reduces its jump height. The controller can pump more energy to the robot’s wings during its aerial phase to compensate.

Due to its small size and light weight, the robot has an even smaller moment of inertia, which makes it more agile than a larger robot and better able to withstand collisions.

The researchers showcased its agility by demonstrating acrobatic flips. The featherweight robot could also hop onto an airborne drone without damaging either device, which could be useful in collaborative tasks.

In addition, while the team demonstrated a hopping robot that carried twice its weight, the maximum payload may be much higher. Adding more weight doesn’t hurt the robot’s efficiency. Rather, the efficiency of the spring is the most significant factor that limits how much the robot can carry.

Moving forward, the researchers plan to leverage its ability to carry heavy loads by installing batteries, sensors, and other circuits onto the robot, in the hopes of enabling it to hop autonomously outside the lab.

“Multimodal robots (those combining multiple movement strategies) are generally challenging and particularly impressive at such a tiny scale. The versatility of this tiny multimodal robot — flipping, jumping on rough or moving terrain, and even another robot — makes it even more impressive,” says Justin Yim, assistant professor at the University of Illinois at Urbana-Champagne, who was not involved with this work. “Continuous hopping shown in this research enables agile and efficient locomotion in environments with many large obstacles.”

This research is funded, in part, by the U.S. National Science Foundation and the MIT MISTI program. Chirarattananon was supported by the Research Grants Council of the Hong Kong Special Administrative Region of China. Hsiao is supported by a MathWorks Fellowship, and Kim is supported by a Zakhartchenko Fellowship.