MIT Latest News

Subscribe to MIT Latest News feed
MIT News is dedicated to communicating to the media and the public the news and achievements of the students, faculty, staff and the greater MIT community.
Updated: 1 day 45 min ago

Three honored with 2023 School of Science teaching prizes

Wed, 01/10/2024 - 3:55pm

The MIT School of Science has announced the winners of its 2023 Teaching Prizes for Graduate and Undergraduate Education. The prizes are awarded to School of Science faculty members who demonstrate excellence in teaching. Winners are chosen from nominations by their students or colleagues.

Roger Levy, a professor in the Department of Brain and Cognitive Sciences, was awarded a prize for developing and teaching class 9.19 (Computational Psycholinguistics). Levy’s nominators highlighted his success in adapting courses to synchronous and asynchronous instruction during the first year of the Covid-19 pandemic and in leading an engaging and challenging course for students across disciplines.

Pulin Li, the Eugene Bell Career Development Professor of Tissue Engineering in the Department of Biology and a member of the Whitehead Institute for Biomedical Research, was awarded the prize for teaching classes 7.06 (Cell Biology) and 7.46/7.86: (Building with Cells). Nominators praised Li’s talent for teaching complex topics effectively and her exceptional accomplishments as a teaching partner.

David McGee, associate professor and associate department head for diversity, equity, and inclusion in the Department of Earth, Atmospheric and Planetary Sciences, was awarded the prize for achieving an outstanding level of community learning in class 12.000 (Solving Complex Problems), also known as “Terrascope.” Nominators noted McGee’s extraordinary investment in both the class material and his students’ learning experiences.

The School of Science welcomes nominations for the teaching prize at the end of each semester. Nominations can be submitted at the school's website.

Researchers release open-source space debris model

Wed, 01/10/2024 - 3:45pm

MIT’s Astrodynamics, Space Robotics, and Controls Laboratory (ARCLab) announced the public beta release of the MIT Orbital Capacity Assessment Tool (MOCAT) during the 2023 Organization for Economic Cooperation and Development (OECD) Space Forum Workshop on Dec. 14. MOCAT enables users to model the long-term future space environment to understand growth in space debris and assess the effectiveness of debris-prevention mechanisms.

With the escalating congestion in low Earth orbit, driven by a surge in satellite deployments, the risk of collisions and space debris proliferation is a pressing concern. Conducting thorough space environment studies is critical for developing effective strategies for fostering responsible and sustainable use of space resources. 

MOCAT stands out among orbital modeling tools for its capability to model individual objects, diverse parameters, orbital characteristics, fragmentation scenarios, and collision probabilities. With the ability to differentiate between object categories, generalize parameters, and offer multi-fidelity computations, MOCAT emerges as a versatile and powerful tool for comprehensive space environment analysis and management.

MOCAT is intended to provide an open-source tool to empower stakeholders including satellite operators, regulators, and members of the public to make data-driven decisions. The ARCLab team has been developing these models for the last several years, recognizing that the lack of open-source implementation of evolutionary modeling tools limits stakeholders’ ability to develop consensus on actions to help improve space sustainability. This beta release is intended to allow users to experiment with the tool and provide feedback to help guide further development.

Richard Linares, the principal investigator for MOCAT and an MIT associate professor of aeronautics and astronautics, expresses excitement about the tool’s potential impact: “MOCAT represents a significant leap forward in orbital capacity assessment. By making it open-source and publicly available, we hope to engage the global community in advancing our understanding of satellite orbits and contributing to the sustainable use of space.”

MOCAT consists of two main components. MOCAT-MC evaluates space environment evolution with individual trajectory simulation and Monte Carlo parameter analysis, providing both a high-level overall view for the environment and a fidelity analysis into the individual space objects evolution. MOCAT Source Sink Evolutionary Model (MOCAT-SSEM), meanwhile, uses a lower-fidelity modeling approach that can run on personal computers within seconds to minutes. MOCAT-MC and MOCAT-SSEM can be accessed separately via GitHub.

MOCAT’s initial development has been supported by the Defense Advanced Research Projects Agency (DARPA) and NASA’s Office of Technology and Strategy.

“We are thrilled to support this groundbreaking orbital debris modeling work and the new knowledge it created,” says Charity Weeden, associate administrator for the Office of Technology, Policy, and Strategy at NASA headquarters in Washington. “This open-source modeling tool is a public good that will advance space sustainability, improve evidence-based policy analysis, and help all users of space make better decisions.”

Food for thought

Wed, 01/10/2024 - 3:00pm

MIT graduate student Juana De La O describes herself as a food-motivated organism, so it’s no surprise that she reaches for food and baking analogies when she’s discussing her thesis work in the lab of undergraduate officer and professor of biology Adam Martin

Consider the formative stages of a croissant, she offers, occasionally providing homemade croissants to accompany the presentation: When one is forming the puff pastry, the dough is folded over the butter again and again. Tissues in a developing mouse embryo must similarly fold and bend, creating layers and structures that become the spine, head, and organs — but these tissues have no hands to induce those formative movements. 

De La O is studying neural tube closure, the formation of the structure that becomes the spinal cord and the brain. Disorders like anencephaly and craniorachischisis occur when the head region fails to close in a developing fetus. It’s a heartbreaking defect, De La O says, because it’s 100 percent lethal — but the fetus fully develops otherwise. 

“Your entire central nervous system hinges on this one event happening successfully,” she says. “On the fundamental level, we have a very limited understanding of the mechanisms required for neural closure to happen at all, much less an understanding of what goes wrong that leads to those defects.” 

Hypothetically speaking

De La O hails from Chicago, where she received an undergraduate degree from the University of Chicago and worked in the lab of Ilaria Rebay. De La O’s sister was the first person in her family to go to and graduate from college — De La O, in turn, is the first person in her family to pursue a PhD. 

From her first time visiting campus, De La O could see MIT would provide a thrilling environment in which to study.

“MIT was one of the few places where the students weren’t constantly complaining about how hard their life was,” she says. “At lunch with prospective students, they’d be talking to each other and then just organically slip into conversations about science.”

The department emails acceptance letters and sends a physical copy via snail mail. De La O’s letter included a handwritten note from department head Amy Keating, then a graduate officer, who had interviewed De La O during her campus visit. 

“That’s what really sold it for me,” she recalls. “I went to my PI [principal investigator]’s office and said, ‘I have new data’” and I showed her the letter, and there was lots of unintelligible crying.” 

To prepare her for graduate school, her parents, both immigrants from Mexico, spent the summer teaching De La O to make all her favorite dishes because “comfort food feels like home.”   

When she reached MIT, however, the Covid-19 pandemic ground the world to a halt and severely limited what students could experience during rotations. Far from home and living alone, De La O taught herself to bake, creating the confections she craved but couldn’t leave her apartment to purchase. De La O didn’t get to work as extensively as she would have liked during her rotation in the Martin lab. 

Martin had recently returned from a sabbatical that was spent learning a new research model; historically a fly lab, Martin was planning to delve into mouse research. 

“My final presentation was, ‘Here’s a hypothetical project I would hypothetically do if I were hypothetically going to work with mice in a fly lab,’” De La O says. 

Martin recalls being impressed. De La O is skilled at talking about science in an earnest and engaging way, and she dug deep into the literature and identified points Martin hadn’t considered. 

“This is a level of independence that I look for in a student because it is important to the science to have someone who is contributing their ideas and independent reading and research to a project,” Martin says. 

After agreeing to join the lab — news she shared with Martin via a meme — she got to work. 

Charting mouse development

The neural tube forms from a flat sheet whose sides rise and meet to create a hollow cylinder. De La O has observed patterns of actin and myosin changing in space and time as the embryo develops. Actin and myosin are fibrous proteins that provide structure in eukaryotic cells. They are responsible for some cell movement, like muscle contraction or cell division. Fibers of actin and myosin can also connect across cells, forming vast networks that coordinate the movements of whole tissues. By looking at the structure of these networks, researchers can make predictions about how force is affecting those tissues.

De La O has found indications of a difference in the tension across the tissue during the critical stages of neural tube closure, which contributes to the tissue’s ability to fold and form a tube. They are not the first research group to propose this, she notes, but they’re suggesting that the patterns of tension are not uniform during a single stage of development.

“My project, on a really fundamental level, is an atlas for a really early stage of mouse development for actin and myosin,” De La O says. “This dataset doesn’t exist in the field yet.” 

However, De La O has been performing analyses exclusively in fixed samples, so she may be quantifying phenomena that are not actually how tissues behave. To determine whether that’s the case, De La O plans to analyze live samples.

The idea is that if one could carefully cut tissue and observe how quickly it recoils, like slicing through a taught rubber band, those measurements could be used to approximate force across the tissue. However, the techniques required are still being developed, and the greater Boston area currently lacks the equipment and expertise needed to attempt those experiments. 

A big part of her work in the lab has been figuring out how to collect and analyze relevant data. This research has already taken her far and wide, both literally and virtually. 

“We’ve found that people have been very generous with their time and expertise,” De La O says. “One of the benefits we, as fly people, brought into this field is we don’t know anything — so we’re going to question everything.”

De La O traveled to the University of Virginia to learn live imaging techniques from associate professor of cell biology Ann Sutherland, and she’s also been in contact with Gabriel Galea at University College London, where Martin and De La O are considering a visit for further training. 

“There are a lot of reasons why these experiments could go wrong, and one of them is that I’m not trained yet,” she says. “Once you know how to do things on an optimal setup, you can figure out how to make it work on a less-optimal setup.”

Collaboration and community

De La O has now expanded her cooking repertoire far beyond her family’s recipes and shares her new creations when she visits home. At MIT, she hosts dinner parties, including one where everything from the savory appetizers to the sweet desserts contained honey, thanks to an Independent Activities Period course about the producers of the sticky substance, and she made and tried apple pie for the first time with her fellow graduate students after an afternoon of apple picking. 

De La O says she’s still learning how to say no to taking on additional work outside of her regular obligations as a PhD student; she’s found there’s a lot of pressure for underrepresented students to be at the forefront of diversity efforts, and although she finds that work extremely fulfilling, she can, and has, stretched herself too thin in the past. 

“Every time I see an application that asks ‘How will you work to increase diversity,’ my strongest instinct is just to write ‘I’m brown and around — you’re welcome,’” she jokes. “The greatest amount of diversity work I will do is to get where I’m going. Me achieving my goals increases diversity inherently, but I also want to do well because I know if I do, I will make everything better for people coming after me.”

De La O is confident her path will be in academia, and troubleshooting, building up protocols, and setting up standards for her work in the Martin Lab has been “an excellent part of my training program.” 

De La O and Martin embarked on a new project in a new model for the lab for De La O’s thesis, so much of her graduate studies will be spent laying the groundwork for future research. 

“I hope her travels open Juana’s eyes to science being a larger community and to teach her about how to lead a collaboration,” Martin says. “Overall, I think this project is excellent for a student with aspirations to be a PI. I benefited from extremely open-ended projects as a student and see, in retrospect, how they prepared me for my work today.”

Richard Wiesman, professor of the practice in mechanical engineering, dies at age 69

Wed, 01/10/2024 - 2:20pm

Richard M. Wiesman ’76, SM ’76, PhD ’83, a professor of the practice in the MIT Department of Mechanical Engineering (MechE), died on Sunday, Jan. 7. He was 69. 

A technology innovator and leader who saw many complex engineering systems reach the marketplace, Wiesman’s work spanned from laboratory development to field deployment. His broad skills in all aspects of automation and robotics — including design, control, communications, locomotion, actuation, sensing, and power — brought a unique perspective to the education of MIT students and made him a tremendous educator, mentor, and colleague.

“Dr. Wiesman’s great enthusiasm for teaching, in parallel with his distinguished industry career, was a wonderful inspiration for our students,” says John Hart, department head and professor of mechanical engineering. “We will miss him very much.”

Wiesman was a lecturer in MechE in the early 1980s and from 2005 to 2007, and was named professor of the practice in 2007. He taught and supervised research in the areas of design, product development, robotics, controls, and manufacturing, and served as co-director of MIT’s Field and Space Robotics Laboratory. In recent years he served on the teaching teams for courses 2.00B, 2.007, 2.008, 2.009, and 2.810, and worked with and inspired many generations of students, including as a 2.009 instructor last fall.  

Wiesman was born on Oct. 7, 1954, to Harold and Elaine Wiesman. He had two brothers, John and Ron, and grew up in Omaha, Nebraska, before coming to study at MIT. Wiesman earned his bachelor’s, master’s, and PhD degrees in MechE at MIT. His doctoral thesis was on high-speed linear induction machines for transportation applications, which led him to work on the U.S. Navy’s Electromagnetic Aircraft Launch System and the Advanced Arresting Gear system.

Wiesman’s work on mobile robots started with the development of a new class of explosive ordnance disposal robots, which grew into a successful business in mobile robots for hazardous ground-based activities — including special robots for internal pipe inspection, robots for warehouse and packing activities, and analysis of robot team characteristics for planetary exploration.

Wiesman worked at Foster Miller/QinetiQ for over 40 years, starting as an engineer and ending his tenure as the executive vice president and chief technology officer. Most recently, he served as a senior fellow for General Atomics and as a member of Arsenal Capital’s Industrial Growth Advisory Board. In 2021, he shared reflections on his career in mechanical engineering with MechE students, telling them in his summation, “I believe you’ve selected an absolutely wonderful career.”

The Institute is also where Wiesman met his wife of 44 years, Suzanne. Together, they took great pleasure in traveling, hiking, snowshoeing, being with friends, and most of all, raising their three children.

Wiesman is survived by his wife; his son Josh and wife Kristina; his son David and wife Haley; his son Ben and wife Emily; and his grandchildren, Elena, John, William, and Julian, in whom he delighted as “Papa.”

In lieu of flowers, the family asks to please consider a donation to the American Heart Association

Noninvasive technique reveals how cells’ gene expression changes over time

Wed, 01/10/2024 - 5:00am

Sequencing all of the RNA in a cell can reveal a great deal of information about that cell’s function and what it is doing at a given point in time. However, the sequencing process destroys the cell, making it difficult to study ongoing changes in gene expression.

An alternative approach developed at MIT could enable researchers to track such changes over extended periods of time. The new method, which is based on a noninvasive imaging technique known as Raman spectroscopy, doesn’t harm cells and can be performed repeatedly.

Using this technique, the researchers showed that they could monitor embryonic stem cells as they differentiated into several other cell types over several days. This technique could enable studies of long-term cellular processes such as cancer progression or embryonic development, and one day might be used for diagnostics for cancer and other diseases.

“With Raman imaging you can measure many more time points, which may be important for studying cancer biology, developmental biology, and a number of degenerative diseases,” says Peter So, a professor of biological and mechanical engineering at MIT, director of MIT’s Laser Biomedical Research Center, and one of the authors of the paper.

Koseki Kobayashi-Kirschvink, a postdoc at MIT and the Broad Institute of Harvard and MIT, is the lead author of the study, which appears today in Nature Biotechnology. The paper’s senior authors are Tommaso Biancalani, a former Broad Institute scientist; Jian Shu, an assistant professor at Harvard Medical School and an associate member of the Broad Institute; and Aviv Regev, executive vice president at Genentech Research and Early Development, who is on leave from faculty positions at the Broad Institute and MIT’s Department of Biology.

Imaging gene expression

Raman spectroscopy is a noninvasive technique that reveals the chemical composition of tissues or cells by shining near-infrared or visible light on them. MIT’s Laser Biomedical Research Center has been working on biomedical Raman spectroscopy since 1985, and recently, So and others in the center have developed Raman spectroscopy-based techniques that could be used to diagnose breast cancer or measure blood glucose.

However, Raman spectroscopy on its own is not sensitive enough to detect signals as small as changes in the levels of individual RNA molecules. To measure RNA levels, scientists typically use a technique called single-cell RNA sequencing, which can reveal the genes that are active within different types of cells in a tissue sample.

In this project, the MIT team sought to combine the advantages of single-cell RNA sequencing and Raman spectroscopy by training a computational model to translate Raman signals into RNA expression states.

“RNA sequencing gives you extremely detailed information, but it’s destructive. Raman is noninvasive, but it doesn’t tell you anything about RNA. So, the idea of this project was to use machine learning to combine the strength of both modalities, thereby allowing you to understand the dynamics of gene expression profiles at the single cell level over time,” Kobayashi-Kirschvink says.

To generate data to train their model, the researchers treated mouse fibroblast cells, a type of skin cell, with factors that reprogram the cells to become pluripotent stem cells. During this process, cells can also transition into several other cell types, including neural and epithelial cells.

Using Raman spectroscopy, the researchers imaged the cells at 36 time points over 18 days as they differentiated. After each image was taken, the researchers analyzed each cell using single molecule fluorescence in situ hybridization (smFISH), which can be used to visualize specific RNA molecules within a cell. In this case, they looked for RNA molecules encoding nine different genes whose expression patterns vary between cell types.

This smFISH data can then act as a link between Raman imaging data and single-cell RNA sequencing data. To make that link, the researchers first trained a deep-learning model to predict the expression of those nine genes based on the Raman images obtained from those cells.

Then, they used a computational program called Tangram, previously developed at the Broad Institute, to link the smFISH gene expression patterns with entire genome profiles that they had obtained by performing single-cell RNA sequencing on the sample cells.

The researchers then combined those two computational models into one that they call Raman2RNA, which can predict individual cells’ entire genomic profiles based on Raman images of the cells.

Tracking cell differentiation

The researchers tested their Raman2RNA algorithm by tracking mouse embryonic stem cells as they differentiated into different cell types. They took Raman images of the cells four times a day for three days, and used their computational model to predict the corresponding RNA expression profiles of each cell, which they confirmed by comparing it to RNA sequencing measurements.

Using this approach, the researchers were able to observe the transitions that occurred in individual cells as they differentiated from embryonic stem cells into more mature cell types. They also showed that they could track the genomic changes that occur as mouse fibroblasts are reprogrammed into induced pluripotent stem cells, over a two-week period.

“It’s a demonstration that optical imaging gives additional information that allows you to directly track the lineage of the cells and the evolution of their transcription,” So says.

The researchers now plan to use this technique to study other types of cell populations that change over time, such as aging cells and cancerous cells. They are now working with cells grown in a lab dish, but in the future, they hope this approach could be developed as a potential diagnostic for use in patients.

“One of the biggest advantages of Raman is that it’s a label-free method. It’s a long way off, but there is potential for the human translation, which could not be done using the existing invasive techniques for measuring genomic profiles,” says Jeon Woong Kang, an MIT research scientist who is also an author of the study.

The research was funded by the Japan Society for the Promotion of Science Postdoctoral Fellowship for Overseas Researchers, the Naito Foundation Overseas Postdoctoral Fellowship, the MathWorks Fellowship, the Helen Hay Whitney Foundation, the U.S. National Institutes of Health, the U.S. National Institute of Biomedical Imaging and Bioengineering, HubMap, the Howard Hughes Medical Institute, and the Klarman Cell Observatory.

The future of motorcycles could be hydrogen

Wed, 01/10/2024 - 12:00am

MIT’s Electric Vehicle Team, which has a long record of building and racing innovative electric vehicles, including cars and motorcycles, in international professional-level competitions, is trying something very different this year: The team is building a hydrogen-powered electric motorcycle, using a fuel cell system, as a testbed for new hydrogen-based transportation.

The motorcycle successfully underwent its first full test-track demonstration in October. It is designed as an open-source platform that should make it possible to swap out and test a variety of different components, and for others to try their own versions based on plans the team is making freely available online.

Aditya Mehrotra, who is spearheading the project, is a graduate student working with mechanical engineering professor Alex Slocum, the Walter M. May  and A. Hazel May Chair in Emerging Technologies. Mehrotra was studying energy systems and happened to also really like motorcycles, he says, “so we came up with the idea of a hydrogen-powered bike. We did an evaluation study, and we thought that this could actually work. We [decided to] try to build it.”

Team members say that while battery-powered cars are a boon for the environment, they still face limitations in range and have issues associated with the mining of lithium and resulting emissions. So, the team was interested in exploring hydrogen-powered vehicles as a clean alternative, allowing for vehicles that could be quickly refilled just like gasoline-powered vehicles.

Unlike past projects by the team, which has been part of MIT since 2005, this vehicle will not be entering races or competitions but will be presented at a variety of conferences. The team, consisting of about a dozen students, has been working on building the prototype since January 2023. In October they presented the bike at the Hydrogen Americas Summit, and in May they will travel to the Netherlands to present it at the World Hydrogen Summit. In addition to the two hydrogen summits, the team plans to show its bike at the Consumer Electronics Show in Las Vegas this month.

“We’re hoping to use this project as a chance to start conversations around ‘small hydrogen’ systems that could increase demand, which could lead to the development of more infrastructure," Mehrotra says. "We hope the project can help find new and creative applications for hydrogen.” In addition to these demonstrations and the online information the team will provide, he adds, they are also working toward publishing papers in academic journals describing their project and lessons learned from it, in hopes of making “an impact on the energy industry.”

The motorcycle took shape over the course of the year piece by piece. “We got a couple of industry sponsors to donate components like the fuel cell and a lot of the major components of the system,” he says. They also received support from the MIT Energy Initiative, the departments of Mechanical Engineering and Electrical Engineering and Computer Science, and the MIT Edgerton Center.

Initial tests were conducted on a dynamometer, a kind of instrumented treadmill Mehrotra describes as “basically a mock road.” The vehicle used battery power during its development, until the fuel cell, provided by South Korean company Doosan, could be delivered and installed. The space the group has used to design and build the prototype, the home of the Electric Vehicle Team, is in MIT’s Building N51 and is well set up to do detailed testing of each of the bike’s components as it is developed and integrated.

Elizabeth Brennan, a senior in mechanical engineering, says she joined the team in January 2023 because she wanted to gain more electrical engineering experience, “and I really fell in love with it.” She says group members “really care and are very excited to be here and work on this bike and believe in the project.”

Brennan, who is the team’s safety lead, has been learning about the safe handling methods required for the bike’s hydrogen fuel, including the special tanks and connectors needed. The team initially used a commercially available electric motor for the prototype but is now working on an improved version, designed from scratch, she says, “which gives us a lot more flexibility.”

As part of the project, team members are developing a kind of textbook describing what they did and how they carried out each step in the process of designing and fabricating this hydrogen electric fuel-cell bike. No such motorcycle yet exists as a commercial product, though a few prototypes have been built.

That kind of guidebook to the process “just doesn’t exist,” Brennan says. She adds that “a lot of the technology development for hydrogen is either done in simulation or is still in the prototype stages, because developing it is expensive, and it’s difficult to test these kinds of systems.” One of the team’s goals for the project is to make everything available as an open-source design, and “we want to provide this bike as a platform for researchers and for education, where researchers can test ideas in both space- and funding-constrained environments.”

Unlike a design built as a commercial product, Mehrotra says, “our vehicle is fully designed for research, so you can swap components in and out, and get real hardware data on how good your designs are.” That can help people work on implementing their new design ideas and help push the industry forward, he says.

The few prototypes developed previously by some companies were inefficient and expensive, he says. “So far as we know, we are the first fully open-source, rigorously documented, tested and released-as-a-platform, [fuel cell] motorcycle in the world. No one else has made a motorcycle and tested it to the level that we have, and documented to the point that someone might actually be able to take this and scale it in the future, or use it in research.”

He adds that “at the moment, this vehicle is affordable for research, but it’s not affordable yet for commercial production because the fuel cell is a very big, expensive component.” Doosan Fuel Cell, which provided the fuel cell for the prototype bike, produces relatively small and lightweight fuel cells mostly for use in drones. The company also produces hydrogen storage and delivery systems.

The project will continue to evolve, says team member Annika Marschner, a sophomore in mechanical engineering. “It’s sort of an ongoing thing, and as we develop it and make changes, make it a stronger, better bike, it will just continue to grow over the years, hopefully,” she says.

While the Electric Vehicle Team has until now focused on battery-powered vehicles, Marschner says, “Right now we’re looking at hydrogen because it seems like something that’s been less explored than other technologies for making sustainable transportation. So, it seemed like an exciting thing for us to offer our time and effort to.”

Making it all work has been a long process. The team is using a frame from a 1999 motorcycle, with many custom-made parts added to support the electric motor, the hydrogen tank, the fuel cell, and the drive train. “Making everything fit in the frame of the bike is definitely something we’ve had to think about a lot because there’s such limited space there. So, it required trying to figure out how to mount things in clever ways so that there are not conflicts,” she says.

Marschner says, “A lot of people don’t really imagine hydrogen energy being something that’s out there being used on the roads, but the technology does exist.” She points out that Toyota and Hyundai have hydrogen-fueled vehicles on the market, and that some hydrogen fuel stations exist, mostly in California, Japan, and some European countries. But getting access to hydrogen, “for your average consumer on the East Coast, is a huge, huge challenge. Infrastructure is definitely the biggest challenge right now to hydrogen vehicles,” she says.

She sees a bright future for hydrogen as a clean fuel to replace fossil fuels over time. “I think it has a huge amount of potential,” she says. “I think one of the biggest challenges with moving hydrogen energy forward is getting these demonstration projects actually developed and showing that these things can work and that they can work well. So, we’re really excited to bring it along further.”

3 Questions: A new home for music at MIT

Tue, 01/09/2024 - 4:00pm

More than 1,500 students enroll in music classes each year at MIT. More than 500 student musicians participate in one of 30 on-campus ensembles. In spring 2025, to better provide for its thriving musical program, MIT will inaugurate its new music building, a 35,000-square-foot three-volume facility adjacent to Kresge Auditorium. The new building will feature high-quality rehearsal and performance spaces, a professional recording studio, classrooms, and laboratories for the music technology program.

Keeril Makan is the Michael (1949) and Sonja Koerner Music Composition Professor, section head of the MIT Music and Theater Arts Section (MTA), and was recently named associate dean of the School of Humanities, Arts, and Social Sciences. A celebrated composer, Makan has been instrumental in the conception and realization of the MIT Music Building, which will also be known as Building W18. He speaks here about the ways that music helps MIT broaden and fulfill its mission, and the opportunities that the new facilities will provide.

Q: After many years of planning, the MIT Music Building is taking shape. How will this new facility change the MIT experience?

A: There is a tremendous demand on campus for the opportunity to make music and to listen to live music. Some of our students arrive at MIT already planning to study and perform music. Others pick up the passion on campus. We have such a flourishing music community here, with so many different types of ensembles we want to support. In addition to the Western Classical tradition, like our orchestra or wind ensemble, where we’ve always been strong, there is also a strong interest in jazz on campus. In fact, we’ve just hired our first jazz professor, Miguel Zenón. More and more students want to explore and experience music from other cultures. We have our Balinese Gamelan, as well as Rambax, a Senegalese drumming group that has the second-largest enrollment for an ensemble, after our orchestra. Our building is designed to allow all of these different musical traditions to exist simultaneously, all equally respected and supported.

With such a strong interest in music among our students and MIT community, the Institute is providing the proper facilities where students and faculty can pursue and develop that interest. And a big part of that is proper acoustics. At MIT we have laboratory spaces that provide stringent environmental conditions for temperature, humidity, vibration, and particulate control. Otherwise, the samples can be contaminated, and the results altered. It’s the same thing in music — we need acoustically controlled rehearsal spaces where the students hear and perform music without contamination from other sound sources. Our performance hall is designed for the audience to hear the music exactly the way the performers hear it. They will experience the music together, in a space that fosters intimacy between the performers and their audience.

Q: Will the new music building attract a different type of student to MIT?

A: I’m not sure whether the new facility will attract a different type of student as much as keep MIT competitive in attracting the type of student who will thrive here. Undergraduates and graduate students have come to expect state-of-the-art facilities across the board for their work in STEM, but also for the parts of their lives that support or complement that work. Music is a big part of that support at MIT. In order for us to stay competitive, to continue to attract the students we believe will help us further our mission, we needed to raise the bar in terms of the level of support we offer students in music. But it’s not just about being competitive in attracting gifted students. Part of our work here is taking on and providing solutions to some of the world’s most pressing and complex challenges. Solving those problems, of course, requires technical expertise. But it also requires wisdom, emotion, and compassion. Empathizing with other members of our community can lead to solutions that will make all of our lives better. And while it’s important that this new building keeps us competitive as an institution, it’s even more important for it to keep us competitive in creating the types of people best suited to take on the world’s great problems. 

Q: How can music, and other arts, complement and support a student’s work in science and technology?

A: Making music is a physical activity. There is something about the small motions of the fingers, the voice resonating, that affects the body, that connects the body with what you are experiencing or feeling. It pulls you completely into the now. Having this building, right in the middle of our campus, makes it clear that this centering is important to MIT and its mission.

For the students rehearsing and performing in the building, or the students who compose music for our new facility, or for the students who will develop the hardware and software that engineers will use to produce music, the problem-solving inherent in those activities is very similar to what they do in STEM. Both are creative processes, where you learn to evaluate, manage, and integrate multiple parameters. Creating music or music technology requires you to rotate a series of different problems in your mind, and to devise a way for them to fit together. It fosters an internal desire for discovery, and for creativity. All of these are skills that, when mastered, easily translate into other activities, including scientific research, math, or engineering. MIT understands that music, and all the arts, are essential in helping our students take on the many challenges facing our world, like the climate crisis, or the impact of AI. Not just in creating an awareness of our humanity, but in training the minds and hearts of the people who will solve those issues. We now have the building that will support that crucial education.

A new way to swiftly eliminate micropollutants from water

Tue, 01/09/2024 - 2:55pm

“Zwitterionic” might not be a word you come across every day, but for Professor Patrick Doyle of the MIT Department of Chemical Engineering, it’s a word that’s central to the technology his group is developing to remove micropollutants from water. Derived from the German word “zwitter,” meaning “hybrid,” “zwitterionic” molecules are those with an equal number of positive and negative charges.

Devashish Gokhale, a PhD student in Doyle’s lab, uses the example of a magnet to describe zwitterionic materials. “On a magnet, you have a north pole and a south pole that stick to each other, and on a zwitterionic molecule, you have a positive charge and a negative charge which stick to each other in a similar way.” Because many inorganic micropollutants and some organic micropollutants are themselves charged, Doyle and his team have been investigating how to deploy zwitterionic molecules to capture micropollutants in water. 

In a new paper in Nature Water, Doyle, Gokhale, and undergraduate student Andre Hamelberg explain how they use zwitterionic hydrogels to sustainably capture both organic and inorganic micropollutants from water with minimal operational complexity. In the past, zwitterionic molecules have been used as coatings on membranes for water treatment because of their non-fouling properties. But in the Doyle group’s system, zwitterionic molecules are used to form the scaffold material, or backbone within the hydrogel — a porous three-dimensional network of polymer chains that contains a significant amount of water. “Zwitterionic molecules have very strong attraction to water compared to other materials which are used to make hydrogels or polymers,” says Gokhale. What’s more, the positive and negative charges on zwitterionic molecules cause the hydrogels to have lower compressibility than what has been commonly observed in hydrogels. This makes for significantly more swollen, robust, and porous hydrogels, which is important for the scale up of the hydrogel-based system for water treatment.

The early stages of this research were supported by a seed grant from MIT’s Abdul Latif Jameel Water and Food Systems Lab (J-WAFS). Doyle’s group is now pursuing commercialization of the platform for both at-home use and industrial scale applications, with support from a J-WAFS Solutions grant.

Seeking a sustainable solution

Micropollutants are chemically diverse materials that can be harmful to human health and the environment, even though they are typically found at low concentrations (micrograms to milligrams per liter) relative to conventional contaminants. Micropollutants can be organic or inorganic and can be naturally-occurring or synthetic. Organic micropollutants are mostly carbon-based molecules and include pesticides and per- and polyfluoroalkyl substances (PFAS), known as “forever chemicals.” Inorganic micropollutants, such as heavy metals like lead and arsenic, tend to be smaller than organic micropollutants. Unfortunately, both organic and inorganic micropollutants are pervasive in the environment.

Many micropollutants come from industrial processes, but the effects of human-induced climate change are also contributing to the environmental spread of micropollutants. Gokhale explains that, in California, for example, fires burn plastic electrical cables and leech micropollutants into natural ecosystems. Doyle adds that “outside of climate change, things like pandemics can spike the number of organic micropollutants in the environment due to high concentrations of pharmaceuticals in wastewater.”

It's no surprise then, that over the past few years micropollutants have become more and more of a concern. These chemicals have garnered attention in the media and led to “significant change in the environmental engineering and regulatory landscape” says Gokhale. In March 2023, the U.S. Environmental Protection Agency (EPA) proposed a strict, federal standard that would regulate six different PFAS chemicals in drinking water. Just last October, the EPA proposed banning the micropollutant trichloroethylene, a cancer-causing chemical that can be found in brake cleaners and other consumer products. And as recently as November, the EPA proposed that water utilities nationwide be required to replace all of their lead pipes to protect the public from lead exposure. Internationally, Gokhale notes the Oslo Paris Convention, whose mission is to protect the marine environment of the northeast Atlantic Ocean, including phasing out the discharge of offshore chemicals from the oil and gas industries. 

With each new, necessary regulation to protect the safety of our water resources, the need for effective water treatment processes grows. Compounding this challenge is the need to make water treatment processes that are sustainable and energy-efficient. 

The benchmark method to treat micropollutants in water is activated carbon. However, making filters with activated carbon is energy-intensive, requiring very high temperatures in large, centralized facilities. Gokhale says approximately “four kilograms of coal are needed to make one kilogram of activated carbon, so you lose a significant amount of carbon dioxide to the environment.” According to the World Economic Forum, global water and wastewater treatment accounts for 5 percent of annual emissions. In the U.S. alone, the EPA reports that drinking water and wastewater systems account for over 45 million tons of greenhouse gas emissions annually.

“We need to develop methods which have smaller climate footprints than methods which are being used industrially today,” says Gokhale.

Supporting a "high-risk" project

In September 2019, Doyle and his lab embarked on an initial project to develop a microparticle-based platform to remove a broad range of micropollutants from water. Doyle’s group had been using hydrogels in pharmaceutical processing to formulate drug molecules into pill format. When he learned about the J-WAFS seed grant opportunity for early-stage research in water and food systems, Doyle realized his pharmaceutical work with hydrogels could be applied to environmental issues like water treatment. “I would never have gotten funding for this project if I went to the NSF [National Science Foundation], because they would just say, ‘you're not a water person.’ But the J-WAFS seed grant offered a way for a high-risk, high-reward kind of project,” Doyle says.

In March 2022, Doyle, Gokhale, and MIT undergraduate Ian Chen published findings from the seed grant work, describing their use of micelles within hydrogels for water treatment. Micelles are spherical structures that form when molecules called surfactants (found in things like soap), come in contact with water or other liquids. The team was able to synthesize micelle-laden hydrogel particles that soak up micropollutants from water like a sponge. Unlike activated carbon, the hydrogel particle system is made from environmentally friendly materials. Furthermore, the system’s materials are made at room temperature, making them exceedingly more sustainable than activated carbon.

Building off the success of the seed grant, Doyle and his team were awarded a J-WAFS Solutions grant in September 2022 to help move their technology from the lab to the market. With this support, the researchers have been able to build, test, and refine pilot-scale prototypes of their hydrogel platform. System iterations during the solutions grant period have included the use of the zwitterionic molecules, a novel advancement from the seed grant work.  

Rapid elimination of micropollutants is of special importance in commercial water treatment processes, where there is a limited amount of time water can spend inside the operational filtration unit. This is referred to as contact time, explains Gokhale. In municipal-scale or industrial-scale water treatment systems, contact times are usually less than 20 minutes and can be as short as five minutes. 

“But as people have been trying to target these emerging micropollutants of concern, they realized they can’t get to sufficiently low concentrations on the same time scales as conventional contaminants,” Gokhale says. “Most technologies focus only on specific molecules or specific classes of molecules. So, you have whole technologies which are focusing only on PFAS, and then you have other technologies for lead and metals. When you start thinking about removing all of these contaminants from water, you end up with designs which have a very large number of unit operations. And that's an issue because you have plants which are in the middle of large cities, and they don't necessarily have space to expand to increase their contact times to efficiently remove multiple micropollutants,” he adds.

Since zwitterionic molecules possess unique properties that confer high porosity, the researchers have been able to engineer a system for quicker uptake of micropollutants from water. Tests show that the hydrogels can eliminate six chemically diverse micropollutants at least 10 times faster than commercial activated carbon. The system is also compatible with a diverse set of materials, making it multifunctional. Micropollutants can bind to many different sites within the hydrogel platform: organic micropollutants bind to the micelles or surfactants while inorganic micropollutants bind to the zwitterionic molecules. Micelles, surfactants, zwitterionic molecules, and other chelating agents can be swapped in and out to essentially tune the system with different functionalities based on the profile of the water being treated. This kind of “plug-and-play” addition of various functional agents does not require a change in the design or synthesis of the hydrogel platform, and adding more functionalities does not take away from existing functionality. In this way, the zwitterionic-based system can rapidly remove multiple contaminants at lower concentrations in a single step, without the need for large, industrial units or capital expenditure. 

Perhaps most importantly, the particles in the Doyle group’s system can be regenerated and used over and over again. By simply soaking the particles in an ethanol bath, they can be washed of micropollutants for indefinite use without loss of efficacy. When activated carbon is used for water treatment, the activated carbon itself becomes contaminated with micropollutants and must be treated as toxic chemical waste and disposed of in special landfills. Over time, micropollutants in landfills will reenter the ecosystem, perpetuating the problem.

Arjav Shah, a PhD-MBA candidate in MIT's Department of Chemical Engineering and the MIT Sloan School of Management, respectively, recently joined the team to lead commercialization efforts. The team has found that the zwitterionic hydrogels could be used in several real-world contexts, ranging from large-scale industrial packed beds to small-scale, portable, off-grid applications — for example, in tablets that could clean water in a canteen — and they have begun piloting the technology through a number of commercialization programs at MIT and in the greater Boston area.

The combined strengths of each member of the team continue to drive the project forward in impactful ways, including undergraduate students like Andre Hamelberg, the third author on the Nature Water paper. Hamelberg is a participant in MIT’s Undergraduate Research Opportunities Program (UROP). Gokhale, who is also a J-WAFS Fellow, provides training and mentorship to Hamelberg and other UROP students in the lab.

“We see this as an educational opportunity,” says Gokhale, noting that the UROP students learn science and chemical engineering through the research they conduct in the lab. The J-WAFS project has also been “a way of getting undergrads interested in water treatment and the more sustainable aspects of chemical engineering,” Gokhale says. He adds that it’s “one of the few projects which goes all the way from designing specific chemistries to building small filters and units and scaling them up and commercializing them. It’s a really good learning opportunity for the undergrads and we're always excited to have them work with us.”

In four years, the technology has been able to grow from an initial idea to a technology with scalable, real-world applications, making it an exemplar J-WAFS project. The fruitful collaboration between J-WAFS and the Doyle lab serves as inspiration for any MIT faculty who may want to apply their research to water or food systems projects.

“The J-WAFS project serves as a way to demystify what a chemical engineer does,” says Doyle. “I think that there's an old idea of chemical engineering as working in just oil and gas. But modern chemical engineering is focused on things which make life and the environment better.”

Shell joins MIT.nano Consortium

Tue, 01/09/2024 - 9:00am

MIT.nano has announced that Shell, a global group of energy and petrochemical companies, has joined the MIT.nano Consortium.

“With an international perspective on the world’s energy challenges, Shell is an exciting addition to the MIT.nano Consortium,” says Vladimir Bulović, the founding faculty director of MIT.nano and the Fariborz Maseeh (1990) Professor of Emerging Technologies. “The quest to build a sustainable energy future will require creative thinking backed by broad and deep expertise that our Shell colleagues bring. They will be insightful collaborators for the MIT community and for our member companies as we work together to explore innovative technology strategies.”

Founded in 1907 when Shell Transport and Trading Co. merged with Royal Dutch, Shell has more than a century’s worth of experience in the exploration, production, refining, and marketing of oil and natural gas and the manufacturing and marketing of chemicals. Operating in over 70 countries, Shell has set a target to become a net-zero emissions energy business by 2050. To achieve this, Shell is supporting developments of low-carbon energy solutions such as biofuels, hydrogen, charging for electric vehicles, and electricity generated by solar and wind power.

“In line with our Powering Progress strategy, our research efforts to become a net-zero emission energy company by 2050 will require intense collaboration with academic leaders across a wide range of disciplines,” says Rolf van Benthem, Shell’s chief scientist for materials science. “We look forward to engaging with the top-notch PIs [principal investigators] at MIT.nano who excel in fields like materials design and nanoscale characterization for use in energy applications and carbon utilization. Together we can work on truly sustainable solutions for our society.”

Shell has been engaged in research collaborations with MIT since 2002 and is a founding member of the MIT Energy Initiative (MITEI). Recent MIT projects supported by Shell include an urban building energy model with the MIT Sustainable Design Laboratory that explores energy-saving building retrofits, a study of the role and impact of hydrogen-based technology pathways with MITEI, and a materials science and engineering project to design better batteries for electric vehicles.

The MIT.nano Consortium is a platform for academia-industry collaboration centered around research and innovation emerging from nanoscale science and engineering at MIT. Through activities that include quarterly industry consortium meetings, Shell will gain insight into the work of MIT.nano’s community of users and provide advice to help guide and advance nanoscale innovations at MIT alongside the 11 other consortium companies:

  • Analog Devices;
  • Draper;
  • Edwards;
  • Fujikura;
  • IBM Research;
  • Lam Research;
  • NC;
  • NEC;
  • Raith;
  • UpNano; and
  • Viavi Solutions.

MIT.nano continues to welcome new companies as sustaining members. For more details, visit the MIT.nano Consortium page.

Multiple AI models help robots execute complex plans more transparently

Mon, 01/08/2024 - 3:15pm

Your daily to-do list is likely pretty straightforward: wash the dishes, buy groceries, and other minutiae. It’s unlikely you wrote out “pick up the first dirty dish,” or “wash that plate with a sponge,” because each of these miniature steps within the chore feels intuitive. While we can routinely complete each step without much thought, a robot requires a complex plan that involves more detailed outlines.

MIT’s Improbable AI Lab, a group within the Computer Science and Artificial Intelligence Laboratory (CSAIL), has offered these machines a helping hand with a new multimodal framework: Compositional Foundation Models for Hierarchical Planning (HiP), which develops detailed, feasible plans with the expertise of three different foundation models. Like OpenAI’s GPT-4, the foundation model that ChatGPT and Bing Chat were built upon, these foundation models are trained on massive quantities of data for applications like generating images, translating text, and robotics.

Unlike RT2 and other multimodal models that are trained on paired vision, language, and action data, HiP uses three different foundation models each trained on different data modalities. Each foundation model captures a different part of the decision-making process and then works together when it’s time to make decisions. HiP removes the need for access to paired vision, language, and action data, which is difficult to obtain. HiP also makes the reasoning process more transparent.

What’s considered a daily chore for a human can be a robot’s “long-horizon goal” — an overarching objective that involves completing many smaller steps first — requiring sufficient data to plan, understand, and execute objectives. While computer vision researchers have attempted to build monolithic foundation models for this problem, pairing language, visual, and action data is expensive. Instead, HiP represents a different, multimodal recipe: a trio that cheaply incorporates linguistic, physical, and environmental intelligence into a robot.

“Foundation models do not have to be monolithic,” says NVIDIA AI researcher Jim Fan, who was not involved in the paper. “This work decomposes the complex task of embodied agent planning into three constituent models: a language reasoner, a visual world model, and an action planner. It makes a difficult decision-making problem more tractable and transparent.”

The team believes that their system could help these machines accomplish household chores, such as putting away a book or placing a bowl in the dishwasher. Additionally, HiP could assist with multistep construction and manufacturing tasks, like stacking and placing different materials in specific sequences.

Evaluating HiP

The CSAIL team tested HiP’s acuity on three manipulation tasks, outperforming comparable frameworks. The system reasoned by developing intelligent plans that adapt to new information.

First, the researchers requested that it stack different-colored blocks on each other and then place others nearby. The catch: Some of the correct colors weren’t present, so the robot had to place white blocks in a color bowl to paint them. HiP often adjusted to these changes accurately, especially compared to state-of-the-art task planning systems like Transformer BC and Action Diffuser, by adjusting its plans to stack and place each square as needed.

Another test: arranging objects such as candy and a hammer in a brown box while ignoring other items. Some of the objects it needed to move were dirty, so HiP adjusted its plans to place them in a cleaning box, and then into the brown container. In a third demonstration, the bot was able to ignore unnecessary objects to complete kitchen sub-goals such as opening a microwave, clearing a kettle out of the way, and turning on a light. Some of the prompted steps had already been completed, so the robot adapted by skipping those directions.

A three-pronged hierarchy

HiP’s three-pronged planning process operates as a hierarchy, with the ability to pre-train each of its components on different sets of data, including information outside of robotics. At the bottom of that order is a large language model (LLM), which starts to ideate by capturing all the symbolic information needed and developing an abstract task plan. Applying the common sense knowledge it finds on the internet, the model breaks its objective into sub-goals. For example, “making a cup of tea” turns into “filling a pot with water,” “boiling the pot,” and the subsequent actions required.

“All we want to do is take existing pre-trained models and have them successfully interface with each other,” says Anurag Ajay, a PhD student in the MIT Department of Electrical Engineering and Computer Science (EECS) and a CSAIL affiliate. “Instead of pushing for one model to do everything, we combine multiple ones that leverage different modalities of internet data. When used in tandem, they help with robotic decision-making and can potentially aid with tasks in homes, factories, and construction sites.”

These models also need some form of “eyes” to understand the environment they’re operating in and correctly execute each sub-goal. The team used a large video diffusion model to augment the initial planning completed by the LLM, which collects geometric and physical information about the world from footage on the internet. In turn, the video model generates an observation trajectory plan, refining the LLM’s outline to incorporate new physical knowledge.

This process, known as iterative refinement, allows HiP to reason about its ideas, taking in feedback at each stage to generate a more practical outline. The flow of feedback is similar to writing an article, where an author may send their draft to an editor, and with those revisions incorporated in, the publisher reviews for any last changes and finalizes.

In this case, the top of the hierarchy is an egocentric action model, or a sequence of first-person images that infer which actions should take place based on its surroundings. During this stage, the observation plan from the video model is mapped over the space visible to the robot, helping the machine decide how to execute each task within the long-horizon goal. If a robot uses HiP to make tea, this means it will have mapped out exactly where the pot, sink, and other key visual elements are, and begin completing each sub-goal.

Still, the multimodal work is limited by the lack of high-quality video foundation models. Once available, they could interface with HiP’s small-scale video models to further enhance visual sequence prediction and robot action generation. A higher-quality version would also reduce the current data requirements of the video models.

That being said, the CSAIL team’s approach only used a tiny bit of data overall. Moreover, HiP was cheap to train and demonstrated the potential of using readily available foundation models to complete long-horizon tasks. “What Anurag has demonstrated is proof-of-concept of how we can take models trained on separate tasks and data modalities and combine them into models for robotic planning. In the future, HiP could be augmented with pre-trained models that can process touch and sound to make better plans,” says senior author Pulkit Agrawal, MIT assistant professor in EECS and director of the Improbable AI Lab. The group is also considering applying HiP to solving real-world long-horizon tasks in robotics.

Ajay and Agrawal are lead authors on a paper describing the work. They are joined by MIT professors and CSAIL principal investigators Tommi Jaakkola, Joshua Tenenbaum, and Leslie Pack Kaelbling; CSAIL research affiliate and MIT-IBM AI Lab research manager Akash Srivastava; graduate students Seungwook Han and Yilun Du ’19; former postdoc Abhishek Gupta, who is now assistant professor at University of Washington; and former graduate student Shuang Li PhD ’23.

The team’s work was supported, in part, by the National Science Foundation, the U.S. Defense Advanced Research Projects Agency, the U.S. Army Research Office, the U.S. Office of Naval Research Multidisciplinary University Research Initiatives, and the MIT-IBM Watson AI Lab. Their findings were presented at the 2023 Conference on Neural Information Processing Systems (NeurIPS).

Co-creating climate futures with real-time data and spatial storytelling

Mon, 01/08/2024 - 2:25pm

Virtual story worlds and game engines aren’t just for video games anymore. They are now tools for scientists and storytellers to digitally twin existing physical spaces and then turn them into vessels to dream up speculative climate stories and build collective designs of the future. That’s the theory and practice behind the MIT WORLDING initiative.

Twice this year, WORLDING matched world-class climate story teams working in XR (extended reality) with relevant labs and researchers across MIT. One global group returned for a virtual gathering online in partnership with Unity for Humanity, while another met for one weekend in person, hosted at the MIT Media Lab.

“We are witnessing the birth of an emergent field that fuses climate science, urban planning, real-time 3D engines, nonfiction storytelling, and speculative fiction, and it is all fueled by the urgency of the climate crises,” says Katerina Cizek, lead designer of the WORLDING initiative at the Co-Creation Studio of MIT Open Documentary Lab. “Interdisciplinary teams are forming and blossoming around the planet to collectively imagine and tell stories of healthy, livable worlds in virtual 3D spaces and then finding direct ways to translate that back to earth, literally.”

At this year’s virtual version of WORLDING, five multidisciplinary teams were selected from an open call. In a week-long series of research and development gatherings, the teams met with MIT scientists, staff, fellows, students, and graduates, as well as other leading figures in the field. Guests ranged from curators at film festivals such as Sundance and Venice, climate policy specialists, and award-winning media creators to software engineers and renowned Earth and atmosphere scientists. The teams heard from MIT scholars in diverse domains, including geomorphology, urban planning as acts of democracy, and climate researchers at MIT Media Lab.

Mapping climate data

“We are measuring the Earth's environment in increasingly data-driven ways. Hundreds of terabytes of data are taken every day about our planet in order to study the Earth as a holistic system, so we can address key questions about global climate change,” explains Rachel Connolly, an MIT Media Lab research scientist focused in the “Future Worlds” research theme, in a talk to the group. “Why is this important for your work and storytelling in general? Having the capacity to understand and leverage this data is critical for those who wish to design for and successfully operate in the dynamic Earth environment.”

Making sense of billions of data points was a key theme during this year’s sessions. In another talk, Taylor Perron, an MIT professor of Earth, atmospheric and planetary sciences, shared how his team uses computational modeling combined with many other scientific processes to better understand how geology, climate, and life intertwine to shape the surfaces of Earth and other planets. His work resonated with one WORLDING team in particular, one aiming to digitally reconstruct the pre-Hispanic Lake Texcoco — where current day Mexico City is now situated — as a way to contrast and examine the region’s current water crisis.

Democratizing the future

While WORLDING approaches rely on rigorous science and the interrogation of large datasets, they are also founded on democratizing community-led approaches.

MIT Department of Urban Studies and Planning graduate Lafayette Cruise MCP '19 met with the teams to discuss how he moved his own practice as a trained urban planner to include a futurist component involving participatory methods. “I felt we were asking the same limited questions in regards to the future we were wanting to produce. We're very limited, very constrained, as to whose values and comforts are being centered. There are so many possibilities for how the future could be.”

Scaling to reach billions

This work scales from the very local to massive global populations. Climate policymakers are concerned with reaching billions of people in the line of fire. “We have a goal to reach 1 billion people with climate resilience solutions,” says Nidhi Upadhyaya, deputy director at Atlantic Council's Adrienne Arsht-Rockefeller Foundation Resilience Center. To get that reach, Upadhyaya is turning to games. “There are 3.3 billion-plus people playing video games across the world. Half of these players are women. This industry is worth $300 billion. Africa is currently among the fastest-growing gaming markets in the world, and 55 percent of the global players are in the Asia Pacific region.” She reminded the group that this conversation is about policy and how formats of mass communication can be used for policymaking, bringing about change, changing behavior, and creating empathy within audiences.

Socially engaged game development is also connected to education at Unity Technologies, a game engine company. “We brought together our education and social impact work because we really see it as a critical flywheel for our business,” said Jessica Lindl, vice president and global head of social impact/education at Unity Technologies, in the opening talk of WORLDING. “We upscale about 900,000 students, in university and high school programs around the world, and about 800,000 adults who are actively learning and reskilling and upskilling in Unity. Ultimately resulting in our mission of the ‘world is a better place with more creators in it,’ millions of creators who reach billions of consumers — telling the world stories, and fostering a more inclusive, sustainable, and equitable world.”

Access to these technologies is key, especially the hardware. “Accessibility has been missing in XR,” explains Reginé Gilbert, who studies and teaches accessibility and disability in user experience design at New York University. “XR is being used in artificial intelligence, assistive technology, business, retail, communications, education, empathy, entertainment, recreation, events, gaming, health, rehabilitation meetings, navigation, therapy, training, video programming, virtual assistance wayfinding, and so many other uses. This is a fun fact for folks: 97.8 percent of the world hasn't tried VR [virtual reality] yet, actually.”

Meanwhile, new hardware is on its way. The WORLDING group got early insights into the highly anticipated Apple Vision Pro headset, which promises to integrate many forms of XR and personal computing in one device. “They're really pushing this kind of pass-through or mixed reality,” said Dan Miller, a Unity engineer on the poly spatial team, collaborating with Apple, who described the experience of the device as “You are viewing the real world. You're pulling up windows, you're interacting with content. It’s a kind of spatial computing device where you have multiple apps open, whether it's your email client next to your messaging client with a 3D game in the middle. You’re interacting with all these things in the same space and at different times.”

“WORLDING combines our passion for social-impact storytelling and incredible innovative storytelling,” said Paisley Smith of the Unity for Humanity Program at Unity Technologies. She added, “This is an opportunity for creators to incubate their game-changing projects and connect with experts across climate, story, and technology.”

Meeting at MIT

In a new in-person iteration of WORLDING this year, organizers collaborated closely with Connolly at the MIT Media Lab to co-design an in-person weekend conference Oct. 25 - Nov. 7 with 45 scholars and professionals who visualize climate data at NASA, the National Oceanic and Atmospheric Administration, planetariums, and museums across the United States.

A participant said of the event, “An incredible workshop that had had a profound effect on my understanding of climate data storytelling and how to combine different components together for a more [holistic] solution.”

“With this gathering under our new Future Worlds banner,” says Dava Newman, director of the MIT Media Lab and Apollo Program Professor of Astronautics chair, “the Media Lab seeks to affect human behavior and help societies everywhere to improve life here on Earth and in worlds beyond, so that all — the sentient, natural, and cosmic — worlds may flourish.” 

“WORLDING’s virtual-only component has been our biggest strength because it has enabled a true, international cohort to gather, build, and create together. But this year, an in-person version showed broader opportunities that spatial interactivity generates — informal Q&As, physical worksheets, and larger-scale ideation, all leading to deeper trust-building,” says WORLDING producer Srushti Kamat SM ’23.

The future and potential of WORLDING lies in the ongoing dialogue between the virtual and physical, both in the work itself and in the format of the workshops.

Technique could efficiently solve partial differential equations for numerous applications

Mon, 01/08/2024 - 1:30pm

In fields such as physics and engineering, partial differential equations (PDEs) are used to model complex physical processes to generate insight into how some of the most complicated physical and natural systems in the world function.

To solve these difficult equations, researchers use high-fidelity numerical solvers, which can be very time-consuming and computationally expensive to run. The current simplified alternative, data-driven surrogate models, compute the goal property of a solution to PDEs rather than the whole solution. Those are trained on a set of data that has been generated by the high-fidelity solver, to predict the output of the PDEs for new inputs. This is data-intensive and expensive because complex physical systems require a large number of simulations to generate enough data. 

In a new paper, “Physics-enhanced deep surrogates for partial differential equations,” published in December in Nature Machine Intelligence, a new method is proposed for developing data-driven surrogate models for complex physical systems in such fields as mechanics, optics, thermal transport, fluid dynamics, physical chemistry, and climate models.

The paper was authored by MIT’s professor of applied mathematics Steven G. Johnson along with Payel Das and Youssef Mroueh of the MIT-IBM Watson AI Lab and IBM Research; Chris Rackauckas of Julia Lab; and Raphaël Pestourie, a former MIT postdoc who is now at Georgia Tech. The authors call their method "physics-enhanced deep surrogate" (PEDS), which combines a low-fidelity, explainable physics simulator with a neural network generator. The neural network generator is trained end-to-end to match the output of the high-fidelity numerical solver.

“My aspiration is to replace the inefficient process of trial and error with systematic, computer-aided simulation and optimization,” says Pestourie. “Recent breakthroughs in AI like the large language model of ChatGPT rely on hundreds of billions of parameters and require vast amounts of resources to train and evaluate. In contrast, PEDS is affordable to all because it is incredibly efficient in computing resources and has a very low barrier in terms of infrastructure needed to use it.”

In the article, they show that PEDS surrogates can be up to three times more accurate than an ensemble of feedforward neural networks with limited data (approximately 1,000 training points), and reduce the training data needed by at least a factor of 100 to achieve a target error of 5 percent. Developed using the MIT-designed Julia programming language, this scientific machine-learning method is thus efficient in both computing and data.

The authors also report that PEDS provides a general, data-driven strategy to bridge the gap between a vast array of simplified physical models with corresponding brute-force numerical solvers modeling complex systems. This technique offers accuracy, speed, data efficiency, and physical insights into the process.

Says Pestourie, “Since the 2000s, as computing capabilities improved, the trend of scientific models has been to increase the number of parameters to fit the data better, sometimes at the cost of a lower predictive accuracy. PEDS does the opposite by choosing its parameters smartly. It leverages the technology of automatic differentiation to train a neural network that makes a model with few parameters accurate.”

“The main challenge that prevents surrogate models from being used more widely in engineering is the curse of dimensionality — the fact that the needed data to train a model increases exponentially with the number of model variables,” says Pestourie. “PEDS reduces this curse by incorporating information from the data and from the field knowledge in the form of a low-fidelity model solver.”

The researchers say that PEDS has the potential to revive a whole body of the pre-2000 literature dedicated to minimal models — intuitive models that PEDS could make more accurate while also being predictive for surrogate model applications.

"The application of the PEDS framework is beyond what we showed in this study,” says Das. “Complex physical systems governed by PDEs are ubiquitous, from climate modeling to seismic modeling and beyond. Our physics-inspired fast and explainable surrogate models will be of great use in those applications, and play a complementary role to other emerging techniques, like foundation models."

The research was supported by the MIT-IBM Watson AI Lab and the U.S. Army Research Office through the Institute for Soldier Nanotechnologies. 

Stripes in a flowing liquid crystal suggest a route to “chiral” fluids

Mon, 01/08/2024 - 5:00am

Hold your hands out in front of you, and no matter how you rotate them, it’s impossible to superimpose one over the other. Our hands are a perfect example of chirality — a geometric configuration by which an object cannot be superimposed onto its mirror image.

Chirality is everywhere in nature, from our hands to the arrangement of our internal organs to the spiral structure of DNA. Chiral molecules and materials have been the key to many drug therapies, optical devices, and functional metamaterials. Scientists have until now assumed that chirality begets chirality — that is, chiral structures emerge from chiral forces and building blocks. But that assumption may need some retuning.

MIT engineers recently discovered that chirality can also emerge in an entirely nonchiral material, and through nonchiral means. In a study appearing today in Nature Communications, the team reports observing chirality in a liquid crystal — a material that flows like a liquid and has non ordered, crystal-like microstructure like a solid. They found that when the fluid flows slowly, its normally nonchiral microstructures spontaneously assemble into large, twisted, chiral structures. The effect is as if a conveyor belt of crayons, all symmetrically aligned, were to suddenly rearrange into large, spiral patterns once the belt reaches a certain speed.

The geometric transformation is unexpected, given that the liquid crystal is naturally nonchiral, or “achiral.” The team’s study thus opens a new path to generating chiral structures. The researchers envision that the structures, once formed, could serve as spiral scaffolds in which to assemble intricate molecular structures. The chiral liquid crystals could also be used as optical sensors, as their structural transformation would change the way they interact with light.

“This is exciting, because this gives us an easy way to structure these kinds of fluids,” says study co-author Irmgard Bischofberger, associate professor of mechanical engineering at MIT. “And from a fundamental level, this is a new way in which chirality can emerge.”

The study’s co-authors include lead author Qing Zhang PhD ’22, Weiqiang Wang and Rui Zhang of Hong Kong University of Science and Technology, and Shuang Zhou of the University of Massachusetts at Amherst.

Striking stripes

A liquid crystal is a phase of matter that embodies properties of both a liquid and a solid. Such in-between materials flow like liquid, and are molecularly structured like solids. Liquid crystals are used as the main element in pixels that make up LCD displays, as the symmetric alignment of their molecules can be uniformly switched with voltage to collectively create high-resolution images.

Bischofberger’s group at MIT studies how fluids and soft materials spontaneously form patterns in nature and in the lab. The team seeks to understand the mechanics underlying fluid transformations, which could be used to create new, reconfigurable materials.

In their new study, the researchers focused on a special type of nematic liquid crystal — a water-based fluid that contains microscopic, rod-like molecular structures. The rods normally align in the same direction throughout the fluid. Zhang was initially curious how the fluid would behave under various flow conditions.

“I tried this experiment for the first time at home, in 2020,” Zhang recalls. “I had samples of the fluid, and a small microscope, and one day I just set it to a low flow. When I came back, I saw this really striking pattern.”

She and her colleagues repeated her initial experiments in the lab. They fabricated a microfluidic channel out of two glass slides, separated by a very thin space, and connected to a main reservoir. The team slowly pumped samples of the liquid crystal through the reservoir and into the space between the plates, then took microscopy images of fluid as it flowed through.

Like Zhang’s initial experiments, the team observed an unexpected transformation: The normally uniform fluid began to form tiger-like stripes as it slowly moved through the channel.

“It was surprising that it formed any structure, but even more surprising once we actually knew what type of structure it formed,” Bischofberger says. “That’s where chirality comes in.”

Twist and flow

The team discovered that the fluid’s stripes were unexpectedly chiral, by using various optical and modeling techniques to effectively retrace the fluid’s flow. They observed that, when unmoving, the fluid’s microscopic rods are normally aligned in near-perfect formation. When the fluid is pumped through the channel quickly, the rods are in complete disarray. But at a slower, in-between flow, the structures start to wiggle, then progressively twist like tiny propellers, each one turning slightly more than the next.

If the fluid continues its slow flow, the twisting crystals assemble into large spiral structures that appear as stripes under the microscope.

“There’s this magic region, where if you just gently make them flow, they form these large spiral structures,” Zhang says.

The researchers modeled the fluid’s dynamics and found that the large spiral patterns emerged when the fluid arrived at a balance between two forces: viscosity and elasticity. Viscosity describes how easily a material flows, while elasticity is essentially how likely a material is to deform (for instance, how easily the fluid’s rods wiggle and twist).

“When these two forces are about the same, that’s when we see these spiral structures,” Bischofberger explains. “It’s kind of amazing that individual structures, on the order of nanometers, can assemble into much larger, millimeter-scale structures that are very ordered, just by pushing them a little bit out of equilibrium.”

The team realized that the twisted assemblages have a chiral geometry: If a mirror image was made of one spiral, it would not be possible to superimpose it over the original, no matter how the spirals were rearranged. The fact that the chiral spirals emerged from a nonchiral material, and through nonchiral means, is a first and points to a relatively simple way to engineer structured fluids.

“The results are indeed surprising and intriguing,” says Giuliano Zanchetta, associate professor at the University of Milan, who was not involved with the study. “It would be interesting to explore the boundaries of this phenomenon. I would see the reported chiral patterns as a promising way to periodically modulate optical properties at the microscale.”

“We now have some knobs to tune this structure,” Bischofberger says. “This might give us a new optical sensor that interacts with light in certain ways. It could also be used as scaffolds to grow and transport molecules for drug delivery. We’re excited to explore this whole new phase space.”

This research was supported, in part, by the U.S. National Science Foundation.

Inhalable sensors could enable early lung cancer detection

Fri, 01/05/2024 - 2:00pm

Using a new technology developed at MIT, diagnosing lung cancer could become as easy as inhaling nanoparticle sensors and then taking a urine test that reveals whether a tumor is present.

The new diagnostic is based on nanosensors that can be delivered by an inhaler or a nebulizer. If the sensors encounter cancer-linked proteins in the lungs, they produce a signal that accumulates in the urine, where it can be detected with a simple paper test strip.

This approach could potentially replace or supplement the current gold standard for diagnosing lung cancer, low-dose computed tomography (CT). It could have an especially significant impact in low- and middle-income countries that don’t have widespread availability of CT scanners, the researchers say.

“Around the world, cancer is going to become more and more prevalent in low- and middle-income countries. The epidemiology of lung cancer globally is that it’s driven by pollution and smoking, so we know that those are settings where accessibility to this kind of technology could have a big impact,” says Sangeeta Bhatia, the John and Dorothy Wilson Professor of Health Sciences and Technology and of Electrical Engineering and Computer Science at MIT, and a member of MIT’s Koch Institute for Integrative Cancer Research and the Institute for Medical Engineering and Science.

Bhatia is the senior author of the paper, which appears today in Science Advances. Qian Zhong, an MIT research scientist, and Edward Tan, a former MIT postdoc, are the lead authors of the study.

Inhalable particles

To help diagnose lung cancer as early as possible, the U.S. Preventive Services Task Force recommends that heavy smokers over the age of 50 undergo annual CT scans. However, not everyone in this target group receives these scans, and the high false-positive rate of the scans can lead to unnecessary, invasive tests.

Bhatia has spent the last decade developing nanosensors for use in diagnosing cancer and other diseases, and in this study, she and her colleagues explored the possibility of using them as a more accessible alternative to CT screening for lung cancer.

These sensors consist of polymer nanoparticles coated with a reporter, such as a DNA barcode, that is cleaved from the particle when the sensor encounters enzymes called proteases, which are often overactive in tumors. Those reporters eventually accumulate in the urine and are excreted from the body.

Previous versions of the sensors, which targeted other cancer sites such as the liver and ovaries, were designed to be given intravenously. For lung cancer diagnosis, the researchers wanted to create a version that could be inhaled, which could make it easier to deploy in lower resource settings.

“When we developed this technology, our goal was to provide a method that can detect cancer with high specificity and sensitivity, and also lower the threshold for accessibility, so that hopefully we can improve the resource disparity and inequity in early detection of lung cancer,” Zhong says.

To achieve that, the researchers created two formulations of their particles: a solution that can be aerosolized and delivered with a nebulizer, and a dry powder that can be delivered using an inhaler.

Once the particles reach the lungs, they are absorbed into the tissue, where they encounter any proteases that may be present. Human cells can express hundreds of different proteases, and some of them are overactive in tumors, where they help cancer cells to escape their original locations by cutting through proteins of the extracellular matrix. These cancerous proteases cleave DNA barcodes from the sensors, allowing the barcodes to circulate in the bloodstream until they are excreted in the urine.

In the earlier versions of this technology, the researchers used mass spectrometry to analyze the urine sample and detect DNA barcodes. However, mass spectrometry requires equipment that might not be available in low-resource areas, so for this version, the researchers created a lateral flow assay, which allows the barcodes to be detected using a paper test strip.

The researchers designed the strip to detect up to four different DNA barcodes, each of which indicates the presence of a different protease. No pre-treatment or processing of the urine sample is required, and the results can be read about 20 minutes after the sample is obtained.

“We were really pushing this assay to be point-of-care available in a low-resource setting, so the idea was to not do any sample processing, not do any amplification, just to be able to put the sample right on the paper and read it out in 20 minutes,” Bhatia says.

Accurate diagnosis

The researchers tested their diagnostic system in mice that are genetically engineered to develop lung tumors similar to those seen in humans. The sensors were administered 7.5 weeks after the tumors started to form, a time point that would likely correlate with stage 1 or 2 cancer in humans.

In their first set of experiments in the mice, the researchers measured the levels of 20 different sensors designed to detect different proteases. Using a machine learning algorithm to analyze those results, the researchers identified a combination of just four sensors that was predicted to give accurate diagnostic results. They then tested that combination in the mouse model and found that it could accurately detect early-stage lung tumors.

For use in humans, it’s possible that more sensors might be needed to make an accurate diagnosis, but that could be achieved by using multiple paper strips, each of which detects four different DNA barcodes, the researchers say.

The researchers now plan to analyze human biopsy samples to see if the sensor panels they are using would also work to detect human cancers. In the longer term, they hope to perform clinical trials in human patients. A company called Sunbird Bio has already run phase 1 trials on a similar sensor developed by Bhatia’s lab, for use in diagnosing liver cancer and a form of hepatitis known as nonalcoholic steatohepatitis (NASH).

In parts of the world where there is limited access to CT scanning, this technology could offer a dramatic improvement in lung cancer screening, especially since the results can be obtained during a single visit.

“The idea would be you come in and then you get an answer about whether you need a follow-up test or not, and we could get patients who have early lesions into the system so that they could get curative surgery or lifesaving medicines,” Bhatia says.

The research was funded by the Johnson & Johnson Lung Cancer Initiative, the Howard Hughes Medical Institute, the Koch Institute Support (core) Grant from the National Cancer Institute, and the National Institute of Environmental Health Sciences.

Improving patient safety using principles of aerospace engineering

Thu, 01/04/2024 - 1:10pm

Approximately 13 billion laboratory tests are administered every year in the United States, but not every result is timely or accurate. Laboratory missteps prevent patients from receiving appropriate, necessary, and sometimes lifesaving care. These medical errors are the third-leading cause of death in the nation. 

To help reverse this trend, a research team from the MIT Department of Aeronautics and Astronautics (AeroAstro) Engineering Systems Lab and Synensys, a safety management contractor, examined the ecosystem of diagnostic laboratory data. Their findings, including six systemic factors contributing to patient hazards in laboratory diagnostics tests, offer a rare holistic view of this complex network — not just doctors and lab technicians, but also device manufacturers, health information technology (HIT) providers, and even government entities such as the White House. By viewing the diagnostic laboratory data ecosystem as an integrated system, an approach based on systems theory, the MIT researchers have identified specific changes that can lead to safer behaviors for health care workers and healthier outcomes for patients. 

A report of the study, which was conducted by AeroAstro Professor Nancy Leveson, who serves as head of the System Safety and Cybersecurity group, along with Research Engineer John Thomas and graduate students Polly Harrington and Rodrigo Rose, was submitted to the U.S. Food and Drug Administration this past fall. Improving the infrastructure of laboratory data has been a priority for the FDA, who contracted the study through Synensis.

Hundreds of hazards, six causes

In a yearlong study that included more than 50 interviews, the Leveson team found the diagnostic laboratory data ecosystem to be vast yet fractured. No one understood how the whole system functioned or the totality of substandard treatment patients received. Well-intentioned workers were being influenced by the system to carry out unsafe actions, MIT engineers wrote.  

Test results sent to the wrong patients, incompatible technologies that strain information sharing between the doctor and lab technician, and specimens transported to the lab without guarantees of temperature control were just some of the hundreds of hazards the MIT engineers identified. The sheer volume of potential risks, known as unsafe control actions (UCAs), should not dissuade health care stakeholders from seeking change, Harrington says. 

“While there are hundreds of UCAs, there are only six systemic factors that are causing these hazards,” she adds. “Using a system-based methodology, the medical community can address many of these issues with one swoop.” 

Four of the systemic factors — decentralization, flawed communication and coordination, insufficient focus on safety-related regulations, and ambiguous or outdated standards — reflect the need for greater oversight and accountability. The two remaining systemic factors — misperceived notions of risk and lack of systems theory integration — call for a fundamental shift in perspective and operations. For instance, the medical community, including doctors themselves, tends to blame physicians when errors occur. Understanding the real risk levels associated with laboratory data and HIT might prompt more action for change, the report’s authors wrote. 

“There’s this expectation that doctors will catch every error,” Harrington says. “It’s unreasonable and unfair to expect that, especially when they have no reason to assume the data they're getting is flawed.”

Think like an engineer

Systems theory may be a new concept to the medical community, but the aviation industry has used it for decades. 

“After World War II, there were so many commercial aviation crashes that the public was scared to fly,” says Leveson, a leading expert in system and software safety. In the early 2000s, she developed the System-Theoretic Process Analysis (STPA), a technique based on systems theory that offers insights into how complex systems can become safer. Researchers used STPA in its report to the FDA. “Industry and government worked together to put controls and error reporting in place. Today, there are nearly zero crashes in the U.S. What’s happening in health care right now is like having a Boeing 787 crash every day.” 

Other engineering principles that work well in aviation, such as control systems, could be applied to health care as well, Thomas says. For instance, closed-loop controls solicit feedback so a system can change and adapt. Having laboratories confirm that physicians received their patients’ test results or investigating all reports of diagnostic errors are examples of closed-loop controls that are not mandated in the current ecosystem, Thomas says. 

“Operating without controls is like asking a robot to navigate a city street blindfolded,” Thomas says. “There’s no opportunity for course correction. Closed-loop controls help inform future decision-making, and, at this point in time, it’s missing in the U.S. health-care system.” 

The Leveson team will continue working with Synensys on behalf of the FDA. Their next study will investigate diagnostic screenings outside the laboratory, such as at a physician’s office (point of care) or at home (over the counter). Since the start of the Covid-19 pandemic, nonclinical lab testing has surged in the country. About 600 million Covid-19 tests were sent to U.S. households between January and September 2022, according to Synensys. Yet, few systems are in place to aggregate these data or report findings to public health agencies.  

“There’s a lot of well-meaning people trying to solve this and other lab data challenges,” Rose says. “If we can convince people to think of health care as an engineered system, we can go a long way in solving some of these entrenched problems.”

The Synensys research contract is art of the Systemic Harmonization and Interoperability Enhancement for Laboratory Data (SHIELD) campaign, an agency initiative that seeks assistance and input in using systems theory to address this challenge. 

Inclusive research for social change

Thu, 01/04/2024 - 12:50pm

Pair a decades-old program dedicated to creating research opportunities for underrepresented minorities and populations with a growing initiative committed to tackling the very issues at the heart of such disparities, and you’ll get a transformative partnership that only MIT can deliver. 

Since 1986, the MIT Summer Research Program (MSRP) has led an institutional effort to prepare underrepresented students (minorities, women in STEM, or students with low socioeconomic status) for doctoral education by pairing them with MIT labs and research groups. For the past three years, the Initiative on Combatting Systemic Racism (ICSR), a cross-disciplinary research collaboration led by MIT’s Institute for Data, Systems, and Society (IDSS), has joined them in their mission, helping bring the issue full circle by providing MSRP students with the opportunity to use big data and computational tools to create impactful changes toward racial equity.

“ICSR has further enabled our direct engagement with undergrads, both within and outside of MIT,” says Fotini Christia, the Ford International Professor of the Social Sciences, associate director of IDSS, and co-organizer for the initiative. “We've found that this line of research has attracted students interested in examining these topics with the most rigorous methods.”

The initiative fits well under the IDSS banner, as IDSS research seeks solutions to complex societal issues through a multidisciplinary approach that includes statistics, computation, modeling, social science methodologies, human behavior, and an understanding of complex systems. With the support of faculty and researchers from all five schools and the MIT Schwarzman College of Computing, the objective of ICSR is to work on an array of different societal aspects of systemic racism through a set of verticals including policing, housing, health care, and social media.

Where passion meets impact

Grinnell senior Mia Hines has always dreamed of using her love for computer science to support social justice. She has experience working with unhoused people and labor unions, and advocating for Indigenous peoples’ rights. When applying to college, she focused her essay on using technology to help Syrian refugees.

“As a Black woman, it's very important to me that we focus on these areas, especially on how we can use technology to help marginalized communities,” Hines says. “And also, how do we stop technology or improve technology that is already hurting marginalized communities?”   

Through MSRP, Hines was paired with research advisor Ufuoma Ovienmhada, a fourth-year doctoral student in the Department of Aeronautics and Astronautics at MIT. A member of Professor Danielle Wood’s Space Enabled research group at MIT’s Media Lab, Ovienmhada received funding from an ICSR Seed Grant and NASA's Applied Sciences Program to support her ongoing research measuring environmental injustice and socioeconomic disparities in prison landscapes. 

“I had been doing satellite remote sensing for environmental challenges and sustainability, starting out looking at coastal ecosystems, when I learned about an issue called ‘prison ecology,’” Ovienmhada explains. “This refers to the intersection of mass incarceration and environmental justice.”

Ovienmhada’s research uses satellite remote sensing and environmental data to characterize exposures to different environmental hazards such as air pollution, extreme heat, and flooding. “This allows others to use these datasets for real-time advocacy, in addition to creating public awareness,” she says.

Focused especially on extreme heat, Hines used satellite remote sensing to monitor the fluctuation of temperature to assess the risk being imposed on prisoners, including death, especially in states like Texas, where 75 percent of prisons either don't have full air conditioning or have none at all.

“Before this project I had done little to no work with geospatial data, and as a budding data scientist, getting to work with and understanding different types of data and resources is really helpful,” Hines says. “I was also funded and afforded the flexibility to take advantage of IDSS’s Data Science and Machine Learning online course. It was really great to be able to do that and learn even more.”

Filling the gap

Much like Hines, Harvey Mudd senior Megan Li was specifically interested in the IDSS-supported MSRP projects. She was drawn to the interdisciplinary approach, and she seeks in her own work to apply computational methods to societal issues and to make computer science more inclusive, considerate, and ethical. 

Working with Aurora Zhang, a grad student in IDSS’s Social and Engineering Systems PhD program, Li used county-level data on income and housing prices to quantify and visualize how affordability based on income alone varies across the United States. She then expanded the analysis to include assets and debt to determine the most common barriers to home ownership.

“I spent my day-to-day looking at census data and writing Python scripts that could work with it,” reports Li. “I also reached out to the Census Bureau directly to learn a little bit more about how they did their data collection, and discussed questions related to some of their previous studies and working papers that I had reviewed.” 

Outside of actual day-to-day research, Li says she learned a lot in conversations with fellow researchers, particularly changing her “skeptical view” of whether or not mortgage lending algorithms would help or hurt home buyers in the approval process. “I think I have a little bit more faith now, which is a good thing.”

“Harvey Mudd is undergraduate-only, and while professors do run labs here, my specific research areas are not well represented,” Li says. “This opportunity was enormous in that I got the experience I need to see if this research area is actually something that I want to do long term, and I got more mirrors into what I would be doing in grad school from talking to students and getting to know faculty.”

Closing the loop

While participating in MSRP offered crucial research experience to Hines, the ICSR projects enabled her to engage in topics she's passionate about and work that could drive tangible societal change.

“The experience felt much more concrete because we were working on these very sophisticated projects, in a supportive environment where people were very excited to work with us,” she says.

A significant benefit for Li was the chance to steer her research in alignment with her own interests. “I was actually given the opportunity to propose my own research idea, versus supporting a graduate student's work in progress,” she explains. 

For Ovienmhada, the pairing of the two initiatives solidifies the efforts of MSRP and closes a crucial loop in diversity, equity, and inclusion advocacy. 

“I've participated in a lot of different DEI-related efforts and advocacy and one thing that always comes up is the fact that it’s not just about bringing people in, it's also about creating an environment and opportunities that align with people’s values,” Ovienmhada says. “Programs like MSRP and ICSR create opportunities for people who want to do work that’s aligned with certain values by providing the needed mentoring and financial support.”

Researchers 3D print components for a portable mass spectrometer

Thu, 01/04/2024 - 12:00am

Mass spectrometers, devices that identify chemical substances, are widely used in applications like crime scene analysis, toxicology testing, and geological surveying. But these machines are bulky, expensive, and easy to damage, which limits where they can be effectively deployed.

Using additive manufacturing, MIT researchers produced a mass filter, which is the core component of a mass spectrometer, that is far lighter and cheaper than the same type of filter made with traditional techniques and materials.

Their miniaturized filter, known as a quadrupole, can be completely fabricated in a matter of hours for a few dollars. The 3D-printed device is as precise as some commercial-grade mass filters that can cost more than $100,000 and take weeks to manufacture.

Built from durable and heat-resistant glass-ceramic resin, the filter is 3D printed in one step, so no assembly is required. Assembly often introduces defects that can hamper the performance of quadrupoles.

This lightweight, cheap, yet precise quadrupole is one important step in Luis Fernando Velásquez-García’s 20-year quest to produce a 3D-printed, portable mass spectrometer.

“We are not the first ones to try to do this. But we are the first ones who succeeded at doing this. There are other miniaturized quadrupole filters, but they are not comparable with professional-grade mass filters. There are a lot of possibilities for this hardware if the size and cost could be smaller without adversely affecting the performance,” says Velásquez-García, a principal research scientist in MIT’s Microsystems Technology Laboratories (MTL) and senior author of a paper detailing the miniaturized quadrupole.

For instance, a scientist could bring a portable mass spectrometer to remote areas of the rainforest, using it to rapidly analyze potential pollutants without shipping samples back to a lab. And a lightweight device would be cheaper and easier to send into space, where it could monitor chemicals in Earth’s atmosphere or on those of distant planets.

Velásquez-García is joined on the paper by lead author Colin Eckhoff, an MIT graduate student in electrical engineering and computer science (EECS); Nicholas Lubinsky, a former MIT postdoc; and Luke Metzler and Randall Pedder of Ardara Technologies. The research is published in Advanced Science.

Size matters

At the heart of a mass spectrometer is the mass filter. This component uses electric or magnetic fields to sort charged particles based on their mass-to-charge ratio. In this way, the device can measure the chemical components in a sample to identify an unknown substance.

A quadrupole, a common type of mass filter, is composed of four metallic rods surrounding an axis. Voltages are applied to the rods, which produce an electromagnetic field. Depending on the properties of the electromagnetic field, ions with a specific mass-to-charge ratio will swirl around through the middle of the filter, while other particles escape out the sides. By varying the mix of voltages, one can target ions with different mass-to-charge ratios.

While fairly simple in design, a typical stainless-steel quadrupole might weigh several kilograms. But miniaturizing a quadrupole is no easy task. Making the filter smaller usually introduces errors during the manufacturing process. Plus, smaller filters collect fewer ions, which makes chemical analysis less sensitive.

“You can’t make quadrupoles arbitrarily smaller — there is a tradeoff,” Velásquez-García adds.

His team balanced this tradeoff by leveraging additive manufacturing to make miniaturized quadrupoles with the ideal size and shape to maximize precision and sensitivity.

They fabricate the filter from a glass-ceramic resin, which is a relatively new printable material that can withstand temperatures up to 900 degrees Celsius and performs well in a vacuum.

The device is produced using vat photopolymerization, a process where a piston pushes into a vat of liquid resin until it nearly touches an array of LEDs at the bottom. These illuminate, curing the resin that remains in the minuscule gap between the piston and the LEDs. A tiny layer of cured polymer is then stuck to the piston, which rises up and repeats the cycle, building the device one tiny layer at a time.

“This is a relatively new technology for printing ceramics that allows you to make very precise 3D objects. And one key advantage of additive manufacturing is that you can aggressively iterate the designs,” Velásquez-García says.

Since the 3D printer can form practically any shape, the researchers designed a quadrupole with hyperbolic rods. This shape is ideal for mass filtering but difficult to make with conventional methods. Many commercial filters employ rounded rods instead, which can reduce performance.

They also printed an intricate network of triangular lattices surrounding the rods, which provides durability while ensuring the rods remain positioned correctly if the device is moved or shaken.

To finish the quadrupole, the researchers used a technique called electroless plating to coat the rods with a thin metal film, which makes them electrically conductive. They cover everything but the rods with a masking chemical and then submerge the quadrupole in a chemical bath heated to a precise temperature and stirring conditions. This deposits a thin metal film on the rods uniformly without damaging the rest of the device or shorting the rods.

“In the end, we made quadrupoles that were the most compact but also the most precise that could be made, given the constraints of our 3D printer,” Velásquez-García says.

Maximizing performance

To test their 3D-printed quadrupoles, the team swapped them into a commercial system and found that they could attain higher resolutions than other types of miniature filters. Their quadrupoles, which are about 12 centimeters in length, are one-quarter the density of comparable stainless-steel filters.

In addition, further experiments suggest that their 3D-printed quadrupoles could achieve precision that is on par with that of largescale commercial filters.

“Mass spectrometry is one of the most important of all scientific tools, and Velásquez-Garcia and co-workers describe the design, construction, and performance of a quadrupole mass filter that has several advantages over earlier devices,” says Graham Cooks, the Henry Bohn Hass Distinguished Professor of Chemistry in the Aston Laboratories for Mass Spectrometry at Purdue University, who was not involved with this work. “The advantages derive from these facts: It is much smaller and lighter than most commercial counterparts and it is fabricated monolithically, using additive construction. … It is an open question as to how well the performance will compare with that of quadrupole ion traps, which depend on the same electric fields for mass measurement but which do not have the stringent geometrical requirements of quadrupole mass filters.”

“This paper represents a real advance in the manufacture of quadrupole mass filters (QMF). The authors bring together their knowledge of manufacture using advanced materials, QMF drive electronics, and mass spectrometry to produce a novel system with good performance at low cost,” adds Steve Taylor, professor of electrical engineering and electronics at the University of Liverpool, who was also not involved with this paper. “Since QMFs are at the heart of the ‘analytical engine’ in many other types of mass spectrometry systems, the paper has an important significance across the whole mass spectrometry field, which worldwide represents a multibillion-dollar industry.”

In the future, the researchers plan to boost the quadrupole’s performance by making the filters longer. A longer filter can enable more precise measurements since more ions that are supposed to be filtered out will escape as the chemical travels along its length. They also intend to explore different ceramic materials that could better transfer heat.

“Our vision is to make a mass spectrometer where all the key components can be 3D printed, contributing to a device with much less weight and cost without sacrificing performance. There is still a lot of work to do, but this is a great start,” Velásquez-Garcia adds.

This work was funded by Empiriko Corporation.

MIT community members elected to the National Academy of Inventors for 2023

Wed, 01/03/2024 - 3:30pm

The National Academy of Inventors (NAI) recently announced the election of more than 160 individuals to their 2023 class of fellows. Among them are two members of the MIT Koch Institute for Integrative Cancer Research, Professor Daniel G. Anderson and Principal Research Scientist Ana Jaklenec. In addition, 11 MIT alumni were also recognized.

The highest professional distinction accorded solely to academic inventors, election to the NAI recognizes individuals who have created or facilitated outstanding inventions that have made a tangible impact on quality of life, economic development, and the welfare of society.  

“Daniel and Ana embody some of the Koch Institute’s core values of interdisciplinary innovation and drive to translate their discoveries into real impact for patients,” says Matthew Vander Heiden, director of the Koch Institute. “Their election to the academy is very well-deserved, and we are honored to count them both among the Koch Institute’s and MIT’s research community.”

Daniel Anderson is the Joseph R. Mares (1924) Professor of Chemical Engineering, and a core member of the Institute for Medical Engineering and Science. He is a leading researcher in the fields of nanotherapeutics and biomaterials. Anderson’s work has led to advances in a range of areas, including medical devices, cell therapy, drug delivery, gene therapy, and material science, and has resulted in the publication of more than 500 papers, patents, and patent applications. He has founded several companies, including Living Proof, Olivo Labs, Crispr Therapeutics (CRSP), Sigilon Therapeutics, Verseau Therapeutics, oRNA, and VasoRx. He is a member of National Academy of Medicine, the Harvard-MIT Division of Health Science and Technology, and is an affiliate of the Broad Institute of MIT and Harvard and the Ragon Institute of MGH, MIT and Harvard.

Ana Jaklenec, a principal research scientist and principal investigator at the Koch Institute, is a leader in the fields of bioengineering and materials science, focused on controlled delivery and stability of therapeutics for global health. She is an inventor of several drug delivery technologies that have the potential to enable equitable access to medical care globally. Her lab is developing new manufacturing techniques for the design of materials at the nano- and micro-scale for self-boosting vaccines, 3D printed on-demand microneedles, heat-stable polymer-based carriers for oral delivery of micronutrients and probiotics, and long-term drug delivery systems for cancer immunotherapy. She has published over 100 manuscripts, patents, and patent applications and has founded three companies: Particles for Humanity, VitaKey, and OmniPulse Biosciences.

The 11 MIT alumni who were elected to the NAI for 2023 include:

  • Michel Barsoum PhD ’85 (Materials Science and Engineering);
  • Eric Burger ’84 (Electrical Engineering and Computer Science);
  • Kevin Kelly SM ’88, PhD ’91 (Mechanical Engineering);
  • Ali Khademhosseini PhD ’05 (Biological Engineering);
  • Joshua Makower ’85 (Mechanical Engineering);
  • Marcela Maus ’97 (Biology);
  • Milos Popovic SM ’02, PhD ’08 (Electrical Engineering and Computer Science);
  • Milica Radisic PhD ’04 (Chemical Engineering);
  • David Reinkensmeyer ’88 (Electrical Engineering);
  • Boris Rubinsky PhD ’81 (Mechanical Engineering); and
  • Paul S. Weiss ’80, SM ’80 (Chemistry).

Since its inception in 2012, the NAI Fellows program has grown to include 1,898 exceptional researchers and innovators, who hold over 63,000 U.S. patents and 13,000 licensed technologies. NAI Fellows are known for the societal and economic impact of their inventions, contributing to major advancements in science and consumer technologies. Their innovations have generated over $3 trillion in revenue and generated 1 million jobs.    

“This year’s class of NAI Fellows showcases the caliber of researchers that are found within the innovation ecosystem. Each of these individuals are making significant contributions to both science and society through their work,” says Paul R. Sanberg, president of the NAI. “This new class, in conjunction with our existing fellows, are creating innovations that are driving crucial advancements across a variety of disciplines and are stimulating the global and national economy in immeasurable ways as they move these technologies from lab to marketplace.” 

AI agents help explain other AI systems

Wed, 01/03/2024 - 3:10pm

Explaining the behavior of trained neural networks remains a compelling puzzle, especially as these models grow in size and sophistication. Like other scientific challenges throughout history, reverse-engineering how artificial intelligence systems work requires a substantial amount of experimentation: making hypotheses, intervening on behavior, and even dissecting large networks to examine individual neurons. To date, most successful experiments have involved large amounts of human oversight. Explaining every computation inside models the size of GPT-4 and larger will almost certainly require more automation — perhaps even using AI models themselves. 

Facilitating this timely endeavor, researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a novel approach that uses AI models to conduct experiments on other systems and explain their behavior. Their method uses agents built from pretrained language models to produce intuitive explanations of computations inside trained networks.

Central to this strategy is the “automated interpretability agent” (AIA), designed to mimic a scientist’s experimental processes. Interpretability agents plan and perform tests on other computational systems, which can range in scale from individual neurons to entire models, in order to produce explanations of these systems in a variety of forms: language descriptions of what a system does and where it fails, and code that reproduces the system’s behavior. Unlike existing interpretability procedures that passively classify or summarize examples, the AIA actively participates in hypothesis formation, experimental testing, and iterative learning, thereby refining its understanding of other systems in real time. 

Complementing the AIA method is the new “function interpretation and description” (FIND) benchmark, a test bed of functions resembling computations inside trained networks, and accompanying descriptions of their behavior. One key challenge in evaluating the quality of descriptions of real-world network components is that descriptions are only as good as their explanatory power: Researchers don’t have access to ground-truth labels of units or descriptions of learned computations. FIND addresses this long-standing issue in the field by providing a reliable standard for evaluating interpretability procedures: explanations of functions (e.g., produced by an AIA) can be evaluated against function descriptions in the benchmark.  

For example, FIND contains synthetic neurons designed to mimic the behavior of real neurons inside language models, some of which are selective for individual concepts such as “ground transportation.” AIAs are given black-box access to synthetic neurons and design inputs (such as “tree,” “happiness,” and “car”) to test a neuron’s response. After noticing that a synthetic neuron produces higher response values for “car” than other inputs, an AIA might design more fine-grained tests to distinguish the neuron’s selectivity for cars from other forms of transportation, such as planes and boats. When the AIA produces a description such as “this neuron is selective for road transportation, and not air or sea travel,” this description is evaluated against the ground-truth description of the synthetic neuron (“selective for ground transportation”) in FIND. The benchmark can then be used to compare the capabilities of AIAs to other methods in the literature. 

Sarah Schwettmann PhD '21, co-lead author of a paper on the new work and a research scientist at CSAIL, emphasizes the advantages of this approach. “The AIAs’ capacity for autonomous hypothesis generation and testing may be able to surface behaviors that would otherwise be difficult for scientists to detect. It’s remarkable that language models, when equipped with tools for probing other systems, are capable of this type of experimental design,” says Schwettmann. “Clean, simple benchmarks with ground-truth answers have been a major driver of more general capabilities in language models, and we hope that FIND can play a similar role in interpretability research.”

Automating interpretability 

Large language models are still holding their status as the in-demand celebrities of the tech world. The recent advancements in LLMs have highlighted their ability to perform complex reasoning tasks across diverse domains. The team at CSAIL recognized that given these capabilities, language models may be able to serve as backbones of generalized agents for automated interpretability. “Interpretability has historically been a very multifaceted field,” says Schwettmann. “There is no one-size-fits-all approach; most procedures are very specific to individual questions we might have about a system, and to individual modalities like vision or language. Existing approaches to labeling individual neurons inside vision models have required training specialized models on human data, where these models perform only this single task. Interpretability agents built from language models could provide a general interface for explaining other systems — synthesizing results across experiments, integrating over different modalities, even discovering new experimental techniques at a very fundamental level.” 

As we enter a regime where the models doing the explaining are black boxes themselves, external evaluations of interpretability methods are becoming increasingly vital. The team’s new benchmark addresses this need with a suite of functions with known structure, that are modeled after behaviors observed in the wild. The functions inside FIND span a diversity of domains, from mathematical reasoning to symbolic operations on strings to synthetic neurons built from word-level tasks. The dataset of interactive functions is procedurally constructed; real-world complexity is introduced to simple functions by adding noise, composing functions, and simulating biases. This allows for comparison of interpretability methods in a setting that translates to real-world performance.      

In addition to the dataset of functions, the researchers introduced an innovative evaluation protocol to assess the effectiveness of AIAs and existing automated interpretability methods. This protocol involves two approaches. For tasks that require replicating the function in code, the evaluation directly compares the AI-generated estimations and the original, ground-truth functions. The evaluation becomes more intricate for tasks involving natural language descriptions of functions. In these cases, accurately gauging the quality of these descriptions requires an automated understanding of their semantic content. To tackle this challenge, the researchers developed a specialized “third-party” language model. This model is specifically trained to evaluate the accuracy and coherence of the natural language descriptions provided by the AI systems, and compares it to the ground-truth function behavior. 

FIND enables evaluation revealing that we are still far from fully automating interpretability; although AIAs outperform existing interpretability approaches, they still fail to accurately describe almost half of the functions in the benchmark. Tamar Rott Shaham, co-lead author of the study and a postdoc in CSAIL, notes that “while this generation of AIAs is effective in describing high-level functionality, they still often overlook finer-grained details, particularly in function subdomains with noise or irregular behavior. This likely stems from insufficient sampling in these areas. One issue is that the AIAs’ effectiveness may be hampered by their initial exploratory data. To counter this, we tried guiding the AIAs’ exploration by initializing their search with specific, relevant inputs, which significantly enhanced interpretation accuracy.” This approach combines new AIA methods with previous techniques using pre-computed examples for initiating the interpretation process.

The researchers are also developing a toolkit to augment the AIAs’ ability to conduct more precise experiments on neural networks, both in black-box and white-box settings. This toolkit aims to equip AIAs with better tools for selecting inputs and refining hypothesis-testing capabilities for more nuanced and accurate neural network analysis. The team is also tackling practical challenges in AI interpretability, focusing on determining the right questions to ask when analyzing models in real-world scenarios. Their goal is to develop automated interpretability procedures that could eventually help people audit systems — e.g., for autonomous driving or face recognition — to diagnose potential failure modes, hidden biases, or surprising behaviors before deployment. 

Watching the watchers

The team envisions one day developing nearly autonomous AIAs that can audit other systems, with human scientists providing oversight and guidance. Advanced AIAs could develop new kinds of experiments and questions, potentially beyond human scientists’ initial considerations. The focus is on expanding AI interpretability to include more complex behaviors, such as entire neural circuits or subnetworks, and predicting inputs that might lead to undesired behaviors. This development represents a significant step forward in AI research, aiming to make AI systems more understandable and reliable.

“A good benchmark is a power tool for tackling difficult challenges,” says Martin Wattenberg, computer science professor at Harvard University who was not involved in the study. “It's wonderful to see this sophisticated benchmark for interpretability, one of the most important challenges in machine learning today. I'm particularly impressed with the automated interpretability agent the authors created. It's a kind of interpretability jiu-jitsu, turning AI back on itself in order to help human understanding.”

Schwettmann, Rott Shaham, and their colleagues presented their work at NeurIPS 2023 in December.  Additional MIT coauthors, all affiliates of the CSAIL and the Department of Electrical Engineering and Computer Science (EECS), include graduate student Joanna Materzynska, undergraduate student Neil Chowdhury, Shuang Li PhD ’23, Assistant Professor Jacob Andreas, and Professor Antonio Torralba. Northeastern University Assistant Professor David Bau is an additional coauthor.

The work was supported, in part, by the MIT-IBM Watson AI Lab, Open Philanthropy, an Amazon Research Award, Hyundai NGV, the U.S. Army Research Laboratory, the U.S. National Science Foundation, the Zuckerman STEM Leadership Program, and a Viterbi Fellowship.

Complex, unfamiliar sentences make the brain’s language network work harder

Wed, 01/03/2024 - 5:00am

With help from an artificial language network, MIT neuroscientists have discovered what kind of sentences are most likely to fire up the brain’s key language processing centers.

The new study reveals that sentences that are more complex, either because of unusual grammar or unexpected meaning, generate stronger responses in these language processing centers. Sentences that are very straightforward barely engage these regions, and nonsensical sequences of words don’t do much for them either.

For example, the researchers found this brain network was most active when reading unusual sentences such as “Buy sell signals remains a particular,” taken from a publicly available language dataset called C4. However, it went quiet when reading something very straightforward, such as “We were sitting on the couch.”

“The input has to be language-like enough to engage the system,” says Evelina Fedorenko, Associate Professor of Neuroscience at MIT and a member of MIT’s McGovern Institute for Brain Research. “And then within that space, if things are really easy to process, then you don’t have much of a response. But if things get difficult, or surprising, if there’s an unusual construction or an unusual set of words that you’re maybe not very familiar with, then the network has to work harder.”

Fedorenko is the senior author of the study, which appears today in Nature Human Behavior. MIT graduate student Greta Tuckute is the lead author of the paper.

Processing language

In this study, the researchers focused on language-processing regions found in the left hemisphere of the brain, which includes Broca’s area as well as other parts of the left frontal and temporal lobes of the brain.

“This language network is highly selective to language, but it’s been harder to actually figure out what is going on in these language regions,” Tuckute says. “We wanted to discover what kinds of sentences, what kinds of linguistic input, drive the left hemisphere language network.”

The researchers began by compiling a set of 1,000 sentences taken from a wide variety of sources — fiction, transcriptions of spoken words, web text, and scientific articles, among many others.

Five human participants read each of the sentences while the researchers measured their language network activity using functional magnetic resonance imaging (fMRI). The researchers then fed those same 1,000 sentences into a large language model — a model similar to ChatGPT, which learns to generate and understand language from predicting the next word in huge amounts of text — and measured the activation patterns of the model in response to each sentence.

Once they had all of those data, the researchers trained a mapping model, known as an “encoding model,” which relates the activation patterns seen in the human brain with those observed in the artificial language model. Once trained, the model could predict how the human language network would respond to any new sentence based on how the artificial language network responded to these 1,000 sentences.

The researchers then used the encoding model to identify 500 new sentences that would generate maximal activity in the human brain (the “drive” sentences), as well as sentences that would elicit minimal activity in the brain’s language network (the “suppress” sentences).

In a group of three new human participants, the researchers found these new sentences did indeed drive and suppress brain activity as predicted.

“This ‘closed-loop’ modulation of brain activity during language processing is novel,” Tuckute says. “Our study shows that the model we’re using (that maps between language-model activations and brain responses) is accurate enough to do this. This is the first demonstration of this approach in brain areas implicated in higher-level cognition, such as the language network.”

Linguistic complexity

To figure out what made certain sentences drive activity more than others, the researchers analyzed the sentences based on 11 different linguistic properties, including grammaticality, plausibility, emotional valence (positive or negative), and how easy it is to visualize the sentence content.

For each of those properties, the researchers asked participants from crowd-sourcing platforms to rate the sentences. They also used a computational technique to quantify each sentence’s “surprisal,” or how uncommon it is compared to other sentences.

This analysis revealed that sentences with higher surprisal generate higher responses in the brain. This is consistent with previous studies showing people have more difficulty processing sentences with higher surprisal, the researchers say.

Another linguistic property that correlated with the language network’s responses was linguistic complexity, which is measured by how much a sentence adheres to the rules of English grammar and how plausible it is, meaning how much sense the content makes, apart from the grammar.

Sentences at either end of the spectrum — either extremely simple, or so complex that they make no sense at all — evoked very little activation in the language network. The largest responses came from sentences that make some sense but require work to figure them out, such as “Jiffy Lube of — of therapies, yes,” which came from the Corpus of Contemporary American English dataset.

“We found that the sentences that elicit the highest brain response have a weird grammatical thing and/or a weird meaning,” Fedorenko says. “There’s something slightly unusual about these sentences.”

The researchers now plan to see if they can extend these findings in speakers of languages other than English. They also hope to explore what type of stimuli may activate language processing regions in the brain’s right hemisphere.

The research was funded by an Amazon Fellowship from the Science Hub, an International Doctoral Fellowship from the American Association of University Women, the MIT-IBM Watson AI Lab, the National Institutes of Health, the McGovern Institute, the Simons Center for the Social Brain, and MIT’s Department of Brain and Cognitive Sciences.

Pages