The Biden-era initiatives are meant to help major polluters like Indonesia increase renewable energy generation.

The agency is making the same move its director assailed before taking office, with more than 1,200 households losing eligibility for Federal Emergency Management Agency-paid hotel rooms.

A ruling Thursday that blocks a Trump administration funding freeze cites “acute harm” from the loss of Federal Emergency Management Agency payments.

The reluctance from key Democrats comes as state lawmakers seek to spend more on climate-related projects.

Florida Crystals faces allegations that it misled consumers by falsely advertising its products as friendly to the climate and environment.

The European Commission proposed changes last week that would rein in the scope of two ESG laws. But companies with business in the EU would still comply.

While consumers won’t have to pay the price directly, the costs will likely be passed on by fuel suppliers.

The facility is expected to come online in the second half of 2025 with the capability of removing 1,000 tons of carbon dioxide per year.

Climate initiatives are also dealing with the slashing of aid and development budgets by rich European nations that are now diverting money to defense.

Nature Climate Change, Published online: 07 March 2025; doi:10.1038/s41558-025-02276-3

Rooftop photovoltaic systems are often seen as a niche solution for mitigation but could offer large-scale opportunities. Using multi-source geospatial data and artificial intelligence techniques, the authors map their potential for reducing global temperatures and analyse regional differences.

Imagine that a robot is helping you clean the dishes. You ask it to grab a soapy bowl out of the sink, but its gripper slightly misses the mark.

Using a new framework developed by MIT and NVIDIA researchers, you could correct that robot’s behavior with simple interactions. The method would allow you to point to the bowl or trace a trajectory to it on a screen, or simply give the robot’s arm a nudge in the right direction.

Unlike other methods for correcting robot behavior, this technique does not require users to collect new data and retrain the machine-learning model that powers the robot’s brain. It enables a robot to use intuitive, real-time human feedback to choose a feasible action sequence that gets as close as possible to satisfying the user’s intent.

When the researchers tested their framework, its success rate was 21 percent higher than an alternative method that did not leverage human interventions.

In the long run, this framework could enable a user to more easily guide a factory-trained robot to perform a wide variety of household tasks even though the robot has never seen their home or the objects in it.

“We can’t expect laypeople to perform data collection and fine-tune a neural network model. The consumer will expect the robot to work right out of the box, and if it doesn’t, they would want an intuitive mechanism to customize it. That is the challenge we tackled in this work,” says Felix Yanwei Wang, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this method.

His co-authors include Lirui Wang PhD ’24 and Yilun Du PhD ’24; senior author Julie Shah, an MIT professor of aeronautics and astronautics and the director of the Interactive Robotics Group in the Computer Science and Artificial Intelligence Laboratory (CSAIL); as well as Balakumar Sundaralingam, Xuning Yang, Yu-Wei Chao, Claudia Perez-D’Arpino PhD ’19, and Dieter Fox of NVIDIA. The research will be presented at the International Conference on Robots and Automation.

Mitigating misalignment

Recently, researchers have begun using pre-trained generative AI models to learn a “policy,” or a set of rules, that a robot follows to complete an action. Generative models can solve multiple complex tasks.

During training, the model only sees feasible robot motions, so it learns to generate valid trajectories for the robot to follow.

While these trajectories are valid, that doesn’t mean they always align with a user’s intent in the real world. The robot might have been trained to grab boxes off a shelf without knocking them over, but it could fail to reach the box on top of someone’s bookshelf if the shelf is oriented differently than those it saw in training.

To overcome these failures, engineers typically collect data demonstrating the new task and re-train the generative model, a costly and time-consuming process that requires machine-learning expertise.

Instead, the MIT researchers wanted to allow users to steer the robot’s behavior during deployment when it makes a mistake.

But if a human interacts with the robot to correct its behavior, that could inadvertently cause the generative model to choose an invalid action. It might reach the box the user wants, but knock books off the shelf in the process.

“We want to allow the user to interact with the robot without introducing those kinds of mistakes, so we get a behavior that is much more aligned with user intent during deployment, but that is also valid and feasible,” Wang says.

Their framework accomplishes this by providing the user with three intuitive ways to correct the robot’s behavior, each of which offers certain advantages.

First, the user can point to the object they want the robot to manipulate in an interface that shows its camera view. Second, they can trace a trajectory in that interface, allowing them to specify how they want the robot to reach the object. Third, they can physically move the robot’s arm in the direction they want it to follow.

“When you are mapping a 2D image of the environment to actions in a 3D space, some information is lost. Physically nudging the robot is the most direct way to specifying user intent without losing any of the information,” says Wang.

Sampling for success

To ensure these interactions don’t cause the robot to choose an invalid action, such as colliding with other objects, the researchers use a specific sampling procedure. This technique lets the model choose an action from the set of valid actions that most closely aligns with the user’s goal.

“Rather than just imposing the user’s will, we give the robot an idea of what the user intends but let the sampling procedure oscillate around its own set of learned behaviors,” Wang explains.

This sampling method enabled the researchers’ framework to outperform the other methods they compared it to during simulations and experiments with a real robot arm in a toy kitchen.

While their method might not always complete the task right away, it offers users the advantage of being able to immediately correct the robot if they see it doing something wrong, rather than waiting for it to finish and then giving it new instructions.

Moreover, after a user nudges the robot a few times until it picks up the correct bowl, it could log that corrective action and incorporate it into its behavior through future training. Then, the next day, the robot could pick up the correct bowl without needing a nudge.

“But the key to that continuous improvement is having a way for the user to interact with the robot, which is what we have shown here,” Wang says.

In the future, the researchers want to boost the speed of the sampling procedure while maintaining or improving its performance. They also want to experiment with robot policy generation in novel environments.

Researchers from the Disruptive and Sustainable Technologies for Agricultural Precision (DiSTAP) interdisciplinary research group of the Singapore-MIT Alliance for Research and Technology (SMART), MIT’s research enterprise in Singapore, in collaboration with Temasek Life Sciences Laboratory (TLL) and MIT, have developed a groundbreaking near-infrared (NIR) fluorescent nanosensor capable of simultaneously detecting and differentiating between iron forms — Fe(II) and Fe(III) — in living plants. 

Iron is crucial for plant health, supporting photosynthesis, respiration, and enzyme function. It primarily exists in two forms: Fe(II), which is readily available for plants to absorb and use, and Fe(III), which must first be converted into Fe(II) before plants can utilize it effectively. Traditional methods only measure total iron, missing the distinction between these forms — a key factor in plant nutrition. Distinguishing between Fe(II) and Fe(III) provides insights into iron uptake efficiency, helps diagnose deficiencies or toxicities, and enables precise fertilization strategies in agriculture, reducing waste and environmental impact while improving crop productivity.

The first-of-its-kind nanosensor developed by SMART researchers enables real-time, nondestructive monitoring of iron uptake, transport, and changes between its different forms — providing precise and detailed observations of iron dynamics. Its high spatial resolution allows precise localization of iron in plant tissues or subcellular compartments, enabling the measurement of even minute changes in iron levels within plants — changes that can inform how a plant handles stress and uses nutrients. 

Traditional detection methods are destructive, or limited to a single form of iron. This new technology enables the diagnosis of deficiencies and optimization of fertilization strategies. By identifying insufficient or excessive iron intake, adjustments can be made to enhance plant health, reduce waste, and support more sustainable agriculture. While the nanosensor was tested on spinach and bok choy, it is species-agnostic, allowing it to be applied across a diverse range of plant species without genetic modification. This capability enhances our understanding of iron dynamics in various ecological settings, providing comprehensive insights into plant health and nutrient management. As a result, it serves as a valuable tool for both fundamental plant research and agricultural applications, supporting precision nutrient management, reducing fertilizer waste, and improving crop health.

“Iron is essential for plant growth and development, but monitoring its levels in plants has been a challenge. This breakthrough sensor is the first of its kind to detect both Fe(II) and Fe(III) in living plants with real-time, high-resolution imaging. With this technology, we can ensure plants receive the right amount of iron, improving crop health and agricultural sustainability,” says Duc Thinh Khong, DiSTAP research scientist and co-lead author of the paper.

“In enabling non-destructive real-time tracking of iron speciation in plants, this sensor opens new avenues for understanding plant iron metabolism and the implications of different iron variations for plants. Such knowledge will help guide the development of tailored management approaches to improve crop yield and more cost-effective soil fertilization strategies,” says Grace Tan, TLL research scientist and co-lead author of the paper.

The research, recently published in Nano Letters and titled, “Nanosensor for Fe(II) and Fe(III) Allowing Spatiotemporal Sensing in Planta,” builds upon SMART DiSTAP’s established expertise in plant nanobionics, leveraging the Corona Phase Molecular Recognition (CoPhMoRe) platform pioneered by the Strano Lab at SMART DiSTAP and MIT. The new nanosensor features single-walled carbon nanotubes (SWNTs) wrapped in a negatively charged fluorescent polymer, forming a helical corona phase structure that interacts differently with Fe(II) and Fe(III). Upon introduction into plant tissues and interaction with iron, the sensor emits distinct NIR fluorescence signals based on the iron type, enabling real-time tracking of iron movement and chemical changes.

The CoPhMoRe technique was used to develop highly selective fluorescent responses, allowing precise detection of iron oxidation states. The NIR fluorescence of SWNTs offers superior sensitivity, selectivity, and tissue transparency while minimizing interference, making it more effective than conventional fluorescent sensors. This capability allows researchers to track iron movement and chemical changes in real time using NIR imaging. 

“This sensor provides a powerful tool to study plant metabolism, nutrient transport, and stress responses. It supports optimized fertilizer use, reduces costs and environmental impact, and contributes to more nutritious crops, better food security, and sustainable farming practices,” says Professor Daisuke Urano, TLL senior principal investigator, DiSTAP principal investigator, National University of Singapore adjunct assistant professor, and co-corresponding author of the paper.

“This set of sensors gives us access to an important type of signalling in plants, and a critical nutrient necessary for plants to make chlorophyll. This new tool will not just help farmers to detect nutrient deficiency, but also give access to certain messages within the plant. It expands our ability to understand the plant response to its growth environment,” says Professor Michael Strano, DiSTAP co-lead principal investigator, Carbon P. Dubbs Professor of Chemical Engineering at MIT, and co-corresponding author of the paper.

Beyond agriculture, this nanosensor holds promise for environmental monitoring, food safety, and health sciences, particularly in studying iron metabolism, iron deficiency, and iron-related diseases in humans and animals. Future research will focus on leveraging this nanosensor to advance fundamental plant studies on iron homeostasis, nutrient signaling, and redox dynamics. Efforts are also underway to integrate the nanosensor into automated nutrient management systems for hydroponic and soil-based farming and expand its functionality to detect other essential micronutrients. These advancements aim to enhance sustainability, precision, and efficiency in agriculture.

The research is carried out by SMART, and supported by the National Research Foundation under its Campus for Research Excellence And Technological Enterprise program.

For over 30 years, science photographer Felice Frankel has helped MIT professors, researchers, and students communicate their work visually. Throughout that time, she has seen the development of various tools to support the creation of compelling images: some helpful, and some antithetical to the effort of producing a trustworthy and complete representation of the research. In a recent opinion piece published in Nature magazine, Frankel discusses the burgeoning use of generative artificial intelligence (GenAI) in images and the challenges and implications it has for communicating research. On a more personal note, she questions whether there will still be a place for a science photographer in the research community.

Q: You’ve mentioned that as soon as a photo is taken, the image can be considered “manipulated.” There are ways you’ve manipulated your own images to create a visual that more successfully communicates the desired message. Where is the line between acceptable and unacceptable manipulation?

A: In the broadest sense, the decisions made on how to frame and structure the content of an image, along with which tools used to create the image, are already a manipulation of reality. We need to remember the image is merely a representation of the thing, and not the thing itself. Decisions have to be made when creating the image. The critical issue is not to manipulate the data, and in the case of most images, the data is the structure. For example, for an image I made some time ago, I digitally deleted the petri dish in which a yeast colony was growing, to bring attention to the stunning morphology of the colony. The data in the image is the morphology of the colony. I did not manipulate that data. However, I always indicate in the text if I have done something to an image. I discuss the idea of image enhancement in my handbook, “The Visual Elements, Photography.”

Q: What can researchers do to make sure their research is communicated correctly and ethically?

A: With the advent of AI, I see three main issues concerning visual representation: the difference between illustration and documentation, the ethics around digital manipulation, and a continuing need for researchers to be trained in visual communication. For years, I have been trying to develop a visual literacy program for the present and upcoming classes of science and engineering researchers. MIT has a communication requirement which mostly addresses writing, but what about the visual, which is no longer tangential to a journal submission? I will bet that most readers of scientific articles go right to the figures, after they read the abstract. 

We need to require students to learn how to critically look at a published graph or image and decide if there is something weird going on with it. We need to discuss the ethics of “nudging” an image to look a certain predetermined way. I describe in the article an incident when a student altered one of my images (without asking me) to match what the student wanted to visually communicate. I didn’t permit it, of course, and was disappointed that the ethics of such an alteration were not considered. We need to develop, at the very least, conversations on campus and, even better, create a visual literacy requirement along with the writing requirement.

Q: Generative AI is not going away. What do you see as the future for communicating science visually?

A: For the Nature article, I decided that a powerful way to question the use of AI in generating images was by example. I used one of the diffusion models to create an image using the following prompt:

“Create a photo of Moungi Bawendi’s nano crystals in vials against a black background, fluorescing at different wavelengths, depending on their size, when excited with UV light.”

The results of my AI experimentation were often cartoon-like images that could hardly pass as reality — let alone documentation — but there will be a time when they will be. In conversations with colleagues in research and computer-science communities, all agree that we should have clear standards on what is and is not allowed. And most importantly, a GenAI visual should never be allowed as documentation.

But AI-generated visuals will, in fact, be useful for illustration purposes. If an AI-generated visual is to be submitted to a journal (or, for that matter, be shown in a presentation), I believe the researcher MUST

• clearly label if an image was created by an AI model;
• indicate what model was used;
• include what prompt was used; and
• include the image, if there is one, that was used to help the prompt.

Senior Kevin Guo, a computer science major, and junior Erin Hovendon, studying mechanical engineering, are on widely divergent paths at MIT. But their lives do intersect in one dimension: They share an understanding that their political science and public policy minors provide crucial perspectives on their research and future careers.

For Guo, the connection between computer science and policy emerged through his work at MIT's Election Data and Science Lab. “When I started, I was just looking for a place to learn how to code and do data science,” he reflects. “But what I found was this fascinating intersection where technical skills could directly shape democratic processes.”

Hovendon is focused on sustainable methods for addressing climate change. She is currently participating in a multisemester research project at MIT's Environmental Dynamics Lab (ENDLab) developing monitoring technology for marine carbon dioxide removal (mCDR).

She believes the success of her research today and in the future depends on understanding its impact on society. Her academic track in policy provides that grounding. “When you’re developing a new technology, you need to focus as well on how it will be applied,” she says. “This means learning about the policies required to scale it up, and about the best ways to convey the value of what you’re working on to the public.”

Bridging STEM and policy

For both Hovendon and Guo, interdisciplinary study is proving to be a valuable platform for tangibly addressing real-world challenges.

Guo came to MIT from Andover, Massachusetts, the son of parents who specialize in semiconductors and computer science. While math and computer science were a natural track for him, Guo was also keenly interested in geopolitics. He enrolled in class 17.40 (American Foreign Policy). “It was my first engagement with MIT political science and I liked it a lot, because it dealt with historical episodes I wanted to learn more about, like World War II, the Korean War, and Vietnam,” says Guo.

He followed up with a class on American Military History and on the Rise of Asia, where he found himself enrolled with graduate students and active duty U.S. military officers. “I liked attending a course with people who had unusual insights,” Guo remarks. “I also liked that these humanities classes were small seminars, and focused a lot on individual students.”

From coding to elections

It was in class 17.835 (Machine Learning and Data Science in Politics) that Guo first realized he could directly connect his computer science and math expertise to the humanities. “They gave us big political science datasets to analyze, which was a pretty cool application of the skills I learned in my major,” he says.

Guo springboarded from this class to a three-year, undergraduate research project in the Election Data and Science Lab. “The hardest part is data collection, which I worked on for an election audit project that looked at whether there were significant differences between original vote counts and audit counts in all the states, at the precinct level,” says Guo. “We had to scrape data, raw PDFs, and create a unified dataset, standardized to our format, that we could publish.”

The data analysis skills he acquired in the lab have come in handy in the professional sphere in which he has begun training: investment finance.

“The workflow is very similar: clean the data to see what you want, analyze it to see if I can find an edge, and then write some code to implement it,” he says. “The biggest difference between finance and the lab research is that the development cycle is a lot faster, where you want to act on a dataset in a few days, rather than weeks or months.”

Engineering environmental solutions

Hovendon, a native of North Carolina with a deep love for the outdoors, arrived at MIT committed “to doing something related to sustainability and having a direct application in the world around me,” she says.

Initially, she headed toward environmental engineering, “but then I realized that pretty much every major can take a different approach to that topic,” she says. “So I ended up switching to mechanical engineering because I really enjoy the hands-on aspects of the field.”

In parallel to her design and manufacturing, and mechanics and materials courses, Hovendon also immersed herself in energy and environmental policy classes. One memorable anthropology class, 21A.404 (Living through Climate Change), asked students to consider whether technological or policy solutions could be fully effective on their own for combating climate change. “It was useful to apply holistic ways of exploring human relations to the environment,” says Hovendon.

Hovendon brings this well-rounded perspective to her research at ENDLab in marine carbon capture and fluid dynamics. She is helping to develop verification methods for mCDR at a pilot treatment plant in California. The facility aims to remove 100 tons of carbon dioxide directly from the ocean by enhancing natural processes. Hovendon hopes to design cost-efficient monitoring systems to demonstrate the efficacy of this new technology. If scaled up, mCDR could enable oceans to store significantly more atmospheric carbon, helping cool the planet.

But Hovendon is well aware that innovation with a major impact cannot emerge on the basis of technical efficacy alone.

“You're going to have people who think that you shouldn't be trying to replicate or interfere with a natural system, and if you're putting one of these facilities somewhere in water, then you're using public spaces and resources,” she says. “It's impossible to come up with any kind of technology, but especially any kind of climate-related technology, without first getting the public to buy into it.”

She recalls class 17.30J (Making Public Policy), which emphasized the importance of both economic and social analysis to the successful passage of highly impactful legislation, such as the Affordable Care Act.

“I think that breakthroughs in science and engineering should be evaluated not just through their technological prowess, but through the success of their implementation for general societal benefit,” she says. “Understanding the policy aspects is vital for improving accessibility for scientific advancements.”

Beyond the dome

Guo will soon set out for a career as a quantitative financial trader, and he views his political science background as essential to his success. While his expertise in data cleaning and analysis will come into play, he believes other skills will as well: “Understanding foreign policy, considering how U.S. policy impacts other places, that's actually very important in finance,” he explains. “Macroeconomic changes and politics affect trading volatility and markets in general, so it's very important to understand what's going on.”

With one year to go, Hovendon is contemplating graduate school in mechanical engineering, perhaps designing renewable energy technologies. “I just really hope that I'm working on something I'm genuinely passionate about, something that has a broader purpose,” she says. “In terms of politics and technology, I also hope that at least some government research and development will still go to climate work, because I'm sure there will be an urgent need for it.”

Interesting article—with photos!—of the US/UK “Combined Cipher Machine” from WWII.

A northern Colorado group that works to reduce wildfires in two conservative counties wonders if the Trump administration will ever pay out federal awards.

Should the state improve air quality and preserve land? Or should it pay for projects that will protect utility customers from rate hikes?

The president's withdrawal of the White House's power to write National Environmental Policy Act rules has created uncertainty for project developers.

Aaron Szabo's carefully worded message for senators was clear: Climate change is real, but his job would not be fighting it.

Jay Bhattacharya said at his confirmation hearing that President Donald Trump's executive orders don't target health initiatives for minorities.