RMI -- formerly known as the Rocky Mountain Institute -- is citing funding uncertainty and "volatility."

Friends of the Earth Netherlands will ask a court to block the company from developing 700 oil and gas fields that it fears Shell will tap.

The status of the Florida Hurricane Catastrophe Fund is important to Floridians regardless of where they live.

Markets have failed to provide adequate pricing signals for the real risks that lie ahead, said Cambridge scholars.

The shift adds to pressure on Brussels to soften the goal of reducing heat-trapping emissions by 90 percent.

It's fresh pain in a country where the World Food Programme says 31 million people already face food insecurity.

Nature Climate Change, Published online: 14 May 2025; doi:10.1038/s41558-025-02332-y

Using seven generations of selected zebrafish (Danio rerio), the authors consider the trade-offs and mechanisms behind evolution of warming tolerance. They show unexpected improvements in cooling tolerance in warming-adapted fish, and highlight mechanistic insights behind warming tolerance.

Imagine a radiologist examining a chest X-ray from a new patient. She notices the patient has swelling in the tissue but does not have an enlarged heart. Looking to speed up diagnosis, she might use a vision-language machine-learning model to search for reports from similar patients.

But if the model mistakenly identifies reports with both conditions, the most likely diagnosis could be quite different: If a patient has tissue swelling and an enlarged heart, the condition is very likely to be cardiac related, but with no enlarged heart there could be several underlying causes.

In a new study, MIT researchers have found that vision-language models are extremely likely to make such a mistake in real-world situations because they don’t understand negation — words like “no” and “doesn’t” that specify what is false or absent. 

“Those negation words can have a very significant impact, and if we are just using these models blindly, we may run into catastrophic consequences,” says Kumail Alhamoud, an MIT graduate student and lead author of this study.

The researchers tested the ability of vision-language models to identify negation in image captions. The models often performed as well as a random guess. Building on those findings, the team created a dataset of images with corresponding captions that include negation words describing missing objects.

They show that retraining a vision-language model with this dataset leads to performance improvements when a model is asked to retrieve images that do not contain certain objects. It also boosts accuracy on multiple choice question answering with negated captions.

But the researchers caution that more work is needed to address the root causes of this problem. They hope their research alerts potential users to a previously unnoticed shortcoming that could have serious implications in high-stakes settings where these models are currently being used, from determining which patients receive certain treatments to identifying product defects in manufacturing plants.

“This is a technical paper, but there are bigger issues to consider. If something as fundamental as negation is broken, we shouldn’t be using large vision/language models in many of the ways we are using them now — without intensive evaluation,” says senior author Marzyeh Ghassemi, an associate professor in the Department of Electrical Engineering and Computer Science (EECS) and a member of the Institute of Medical Engineering Sciences and the Laboratory for Information and Decision Systems.

Ghassemi and Alhamoud are joined on the paper by Shaden Alshammari, an MIT graduate student; Yonglong Tian of OpenAI; Guohao Li, a former postdoc at Oxford University; Philip H.S. Torr, a professor at Oxford; and Yoon Kim, an assistant professor of EECS and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. The research will be presented at Conference on Computer Vision and Pattern Recognition.

Neglecting negation

Vision-language models (VLM) are trained using huge collections of images and corresponding captions, which they learn to encode as sets of numbers, called vector representations. The models use these vectors to distinguish between different images.

A VLM utilizes two separate encoders, one for text and one for images, and the encoders learn to output similar vectors for an image and its corresponding text caption.

“The captions express what is in the images — they are a positive label. And that is actually the whole problem. No one looks at an image of a dog jumping over a fence and captions it by saying ‘a dog jumping over a fence, with no helicopters,’” Ghassemi says.

Because the image-caption datasets don’t contain examples of negation, VLMs never learn to identify it.

To dig deeper into this problem, the researchers designed two benchmark tasks that test the ability of VLMs to understand negation.

For the first, they used a large language model (LLM) to re-caption images in an existing dataset by asking the LLM to think about related objects not in an image and write them into the caption. Then they tested models by prompting them with negation words to retrieve images that contain certain objects, but not others.

For the second task, they designed multiple choice questions that ask a VLM to select the most appropriate caption from a list of closely related options. These captions differ only by adding a reference to an object that doesn’t appear in the image or negating an object that does appear in the image.

The models often failed at both tasks, with image retrieval performance dropping by nearly 25 percent with negated captions. When it came to answering multiple choice questions, the best models only achieved about 39 percent accuracy, with several models performing at or even below random chance.

One reason for this failure is a shortcut the researchers call affirmation bias — VLMs ignore negation words and focus on objects in the images instead.

“This does not just happen for words like ‘no’ and ‘not.’ Regardless of how you express negation or exclusion, the models will simply ignore it,” Alhamoud says.

This was consistent across every VLM they tested.

“A solvable problem”

Since VLMs aren’t typically trained on image captions with negation, the researchers developed datasets with negation words as a first step toward solving the problem.

Using a dataset with 10 million image-text caption pairs, they prompted an LLM to propose related captions that specify what is excluded from the images, yielding new captions with negation words.

They had to be especially careful that these synthetic captions still read naturally, or it could cause a VLM to fail in the real world when faced with more complex captions written by humans.

They found that finetuning VLMs with their dataset led to performance gains across the board. It improved models’ image retrieval abilities by about 10 percent, while also boosting performance in the multiple-choice question answering task by about 30 percent.

“But our solution is not perfect. We are just recaptioning datasets, a form of data augmentation. We haven’t even touched how these models work, but we hope this is a signal that this is a solvable problem and others can take our solution and improve it,” Alhamoud says.

At the same time, he hopes their work encourages more users to think about the problem they want to use a VLM to solve and design some examples to test it before deployment.

In the future, the researchers could expand upon this work by teaching VLMs to process text and images separately, which may improve their ability to understand negation. In addition, they could develop additional datasets that include image-caption pairs for specific applications, such as health care.

The MIT Press has announced that beginning in 2026, Duke University Press will join its Direct to Open (D2O) program. This collaboration marks the first such partnership with another university press for the D2O program, and reaffirms their shared commitment to open access publishing that is ethical, equitable, and sustainable.

Launched in 2021, D2O is the MIT Press’ framework for open access monographs that shifts publishing from a solely market-based purchase model, where individuals and libraries buy single e-books, to a collaborative, library-supported open access model. 

Duke University Press brings their distinguished catalog in the humanities and social sciences to Direct to Open, providing open access to 20 frontlist titles annually alongside the MIT Press’ 80 scholarly books each year. Their participation in the D2O program — which will also include free term access to a paywalled collection of 250 key backlist titles — enhances the range of openly available academic content for D2O’s library partners.

“By expanding the Direct to Open model to include one of the most innovative university presses publishing today, we’re taking a significant step toward building a more open and accessible future for academic publishing,” says Amy Brand, director and publisher of the MIT Press. “We couldn’t be more thrilled to be building this partnership with Duke University Press. This collaboration will benefit the entire scholarly community, ensuring that more books are made openly available to readers worldwide.”

“We are honored to participate in MIT Press’ dynamic and successful D2O program,” says Dean Smith, director of Duke University Press. “It greatly expands our open-access footprint and serves our mission of making bold and transformational scholarship accessible to the world.”

With Duke University Press’ involvement in 2026, D2O will feature multiple package options, combining content from both the MIT Press and Duke University Press. Participating institutions will have the opportunity to support each press individually, providing flexibility for libraries while fostering collective impact.

For details on how your institution might participate in or support Direct to Open, please visit the D2O website or contact the MIT Press library relations team.  

Starting in July, MIT’s Shaping the Future of Work Initiative in the Department of Economics will usher in a significant new era of research, policy, and education of the next generation of scholars, made possible by a gift from the James M. and Cathleen D. Stone Foundation. In recognition of the gift and the expansion of priorities it supports, on July 1 the initiative will become part of the new James M. and Cathleen D. Stone Center on Inequality and Shaping the Future of Work. This center will be officially launched at a public event in fall 2025.

The Stone Center will be led by Daron Acemoglu, Institute Professor, and co-directors David Autor, the Daniel (1972) and Gail Rubinfeld Professor in Economics, and Simon Johnson, the Ronald A. Kurtz (1954) Professor of Entrepreneurship. It will join a global network of 11 other wealth inequality centers funded by the Stone Foundation as part of an effort to advance research on the causes and consequences of the growing accumulation at the top of the wealth distribution.

“This generous gift from the Stone Foundation advances our pioneering economics research on inequality, technology, and the future of the workforce. This work will create a pipeline of scholars in this critical area of study, and it will help to inform the public and policymakers,” says Provost Cynthia Barnhart.

Originally established as part of MIT Blueprint Labs with a foundational gift from the William and Flora Hewlett Foundation, the Shaping the Future of Work Initiative is a nonpartisan research organization that applies economics research to identify innovative ways to move the labor market onto a more equitable trajectory, with a central focus on revitalizing labor market opportunities for workers without a college education. Building on frontier micro- and macro-economics, economic sociology, political economy, and other disciplines, the initiative seeks to answer key questions about the decline in labor market opportunities for non-college workers in recent decades. These labor market changes have been a major driver of growing wealth inequality, a phenomenon that has, in turn, broadly reshaped our economy, democracy, and society.

Support from the Stone Foundation will allow the new Stone Center to build on the Shaping the Future of Work Initiative’s ongoing research agenda and extend its focus to include a growing emphasis on the interplay between technologies and inequality, as well as the technology sector’s role in defining future inequality.

Core objectives of the James M. and Cathleen D. Stone Center on Inequality and Shaping the Future of Work will include fostering connections between scholars doing pathbreaking research on automation, AI, the intersection of work and technology, and wealth inequality across disciplines, including within the Department of Economics, the MIT Sloan School of Management, and the MIT Stephen A. Schwarzman College of Computing; strengthening the pipeline of emerging scholars focused on these issues; and using research to inform and engage a wider audience including the public, undergraduate and graduate students, and policymakers.     

The Stone Foundation’s support will allow the center to strengthen and expand its commitments to produce new research, convene additional events to share research findings, promote connection and collaboration between scholars working on related topics, provide new resources for the center’s research affiliates, and expand public outreach to raise awareness of this important emerging challenge. “Cathy and I are thrilled to welcome MIT to the growing family of Stone Centers dedicated to studying the urgent challenges of accelerating wealth inequality,” James M. Stone says.

Agustín Rayo, dean of the School of Humanities, Arts, and Social Sciences, says, “I am thrilled to celebrate the creation of the James M. and Cathleen D. Stone Center in the MIT economics department. Not only will it enhance the cutting-edge work of MIT’s social scientists, but it will support cross-disciplinary interactions that will enable new insights and solutions to complex social challenges.”

Jonathan Gruber, chair of the Department of Economics, adds, “I couldn’t be more excited about the Stone Foundation’s support for the Shaping the Future of Work Initiative. The initiative’s leaders have been far ahead of the curve in anticipating the rapid changes that technological forces are bringing to the labor market, and their influential studies have helped us understand the potential effects of AI and other technologies on U.S. workers. The generosity of the Stone Foundation will allow them to continue this incredible work, while expanding their priorities to include other critical issues around inequality. This is a great moment for the paradigm-shifting research that Acemoglu, Autor, and Johnson are leading here at MIT.”

“We are grateful to the James M. and Cathleen D. Stone Foundation for their generous support enabling us to study two defining challenges of our age: inequality and the future of work,” says Acemoglu, who was awarded the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel in 2024 (with co-laureates Simon Johnson and James A. Robinson). “We hope to go beyond exploring the causes of inequality and the determinants of the availability of good jobs in the present and in the future, but also develop ideas about how society can shape both the work of the future and inequality by its choices of institutions and technological trajectories.”

“We are incredibly fortunate to be joining the family of Stone Centers around the world. Jim and Cathleen Stone are far-sighted and generous donors, and we are delighted that they are willing to back us and MIT in this way,” says Johnson. “We look forward to working with all our colleagues, at MIT and around the world, to advance understanding and practical approaches to inequality and the future of work.”

Autor adds, “This support will enable us — and many others — to focus our scholarship, teaching and public outreach towards shaping a labor market that offers opportunity, mobility, and economic security to a far broader set of people.” 

Just 10 to 15 minutes of mindfulness practice a day led to reduced stress and anxiety for autistic adults who participated in a study led by scientists at MIT’s McGovern Institute for Brain Research. Participants in the study used a free smartphone app to guide their practice, giving them the flexibility to practice when and where they chose.

Mindfulness is a state in which the mind is focused only on the present moment. It is a way of thinking that can be cultivated with practice, often through meditation or breathing exercises — and evidence is accumulating that practicing mindfulness has positive effects on mental health. The new open-access study, reported April 8 in the journal Mindfulness, adds to that evidence, demonstrating clear benefits for autistic adults.

“Everything you want from this on behalf of somebody you care about happened: reduced reports of anxiety, reduced reports of stress, reduced reports of negative emotions, and increased reports of positive emotions,” says McGovern investigator and MIT Professor John Gabrieli, who led the research with Liron Rozenkrantz, an investigator at the Azrieli Faculty of Medicine at Bar-Ilan University in Israel and a research affiliate in Gabrieli’s lab. “Every measure that we had of well-being moved in significantly in a positive direction,” adds Gabrieli, who is also the Grover Hermann Professor of Health Sciences and Technology and a professor of brain and cognitive sciences at MIT.

One of the reported benefits of practicing mindfulness is that it can reduce the symptoms of anxiety disorders. This prompted Gabrieli and his colleagues to wonder whether it might benefit adults with autism, who tend to report above average levels of anxiety and stress, which can interfere with daily living and quality of life. As many as 65 percent of autistic adults may also have an anxiety disorder.

Gabrieli adds that the opportunity for autistic adults to practice mindfulness with an app, rather than needing to meet with a teacher or class, seemed particularly promising. “The capacity to do it at your own pace in your own home, or any environment you like, might be good for anybody,” he says. “But maybe especially for people for whom social interactions can sometimes be challenging.”

The research team, including Cindy Li, the autism recruitment and outreach coordinator in Gabrieli’s lab, recruited 89 autistic adults to participate in their study. Those individuals were split into two groups: one would try the mindfulness practice for six weeks, while the others would wait and try the intervention later.

Participants were asked to practice daily using an app called Healthy Minds, which guides participants through seated or active meditations, each lasting 10 to 15 minutes. Participants reported that they found the app easy to use and had little trouble making time for the daily practice.

After six weeks, participants reported significant reductions in anxiety and perceived stress. These changes were not experienced by the wait-list group, which served as a control. However, after their own six weeks of practice, people in the wait-list group reported similar benefits. “We replicated the result almost perfectly. Every positive finding we found with the first sample we found with the second sample,” Gabrieli says.

The researchers followed up with study participants after another six weeks. Almost everyone had discontinued their mindfulness practice — but remarkably, their gains in well-being had persisted. Based on this finding, the team is eager to further explore the long-term effects of mindfulness practice in future studies. “There’s a hypothesis that a benefit of gaining mindfulness skills or habits is they stick with you over time — that they become incorporated in your daily life,” Gabrieli says. “If people are using the approach to being in the present and not dwelling on the past or worrying about the future, that’s what you want most of all. It’s a habit of thought that’s powerful and helpful.”

Even as they plan future studies, the researchers say they are already convinced that mindfulness practice can have clear benefits for autistic adults. “It’s possible mindfulness would be helpful at all kinds of ages,” Gabrieli says. But he points out the need is particularly great for autistic adults, who usually have fewer resources and support than autistic children have access to through their schools. Gabrieli is eager for more people with autism to try the Healthy Minds app. “Having scientifically proven resources for adults who are no longer in school systems might be a valuable thing,” he says.

This research was funded, in part, by The Hock E. Tan and K. Lisa Yang Center for Autism Research at MIT and the Yang Tan Collective.

The case is over:

A jury has awarded WhatsApp $167 million in punitive damages in a case the company brought against Israel-based NSO Group for exploiting a software vulnerability that hijacked the phones of thousands of users.

I’m sure it’ll be appealed. Everything always is.

The House Energy and Commerce Committee wants to use the budget process to repeal program authorizations. That may not fly in the Senate.

The proposal, which is circulating through federal agencies, could help fossil fuel projects move forward more quickly.

A majority on the state Supreme Court allowed the climate lawsuit to proceed at the state level, but a minority dissent warned of "chaos."

About 11 million U.S. residents had to relocate to another part of the country in 2024.

The council, which will meet for the first time next week, has until Nov. 16 to issue a report on the disaster agency's future.

Climate research is under pressure as U.S. funding cuts stand to reduce scientific activity in the region and Russia’s isolation hinders collaboration.

Clergy and faithful in the Amazon region see Pope Leo XIV as a pontiff who could protect the region and fight against climate change.

Shota Ikeda is calling for emissions reductions of at least 75 percent by 2035.