Weaving memory into textiles

In 2021, a curator at the Smithsonian Institution contacted Chloé Bensahel, currently the MIT 2023-24 Ida Ely Rubin Artist in Residence, and told her about some objects that had been made for space missions. “They were weavings of conductive yarn with magnetic pieces in them,” Bensahel says. “After World War II, you had these really powerful computers but no way to store data, so scientists at MIT and Harvard came up with this magnetic core memory. It was the last moment, I think, in computing history when information was visible: You can actually see the code because of the little magnets that were turned on or turned off.”

What really captured the attention of Bensahel, who works with textiles, is that those items had been woven by hand at MIT. “They’re the result of two histories in New England that are coinciding: the declining textile industry and the increasing space research,” she says. “Legend has it that the women who were getting laid off from the textile industries got hired by MIT to make these objects. They were weaving here on campus.”

Reinventing codes

Eventually, Bensahel connected with Zach Lieberman, an adjunct associate professor who runs the Future Sketches group at the MIT Media Lab, who applied for a MIT Center for Art Science and Technology (CAST) grant to bring her to campus as a visiting artist. The pair share an interest in various forms of code and communication — Bensahel, for example, sees textiles as carrying information, not just in what they visually display, like, say, a slogan on a T-shirt, but in the very way they are made. Now, they are working together at MIT, which has been unfurling in connection with Bensahel’s residency at Villa Albertine, an arts institution launched in 2021 by the French Embassy in the United States that supports cultural exchange between the United States, France and beyond, including offering more than 50 residencies each year for artists, thinkers, and creators across all disciplines.

Bensahel is building on MIT’s groundbreaking legacy in the weaving of memory technology, which complements the research conducted by her MIT collaborators, whether they are faculty members or research assistants. “We’re primarily software-oriented here,” Lieberman says, referring to his group. “We are working in the realm of bits and with language. Chloe’s work is also really intimately concerned with language, but she’s coming at it from a perspective of materials and trying to figure out how to weave them in different ways, and connect with electronics and sensing.”

Theory and craftsmanship

Born in France, Bensahel moved to the United States when she was 7. She attended Parsons School of Design, in New York City. She specialized in integrated design with a focus on textiles, and graduated in 2013. The coursework was essentially theoretical and philosophical, though, and afterward Bensahel moved to France to hone her craftsmanship. “I wanted to learn with my hands, not just my mind,” she says — no doubt making her a perfect fit for MIT, whose motto, “mens et manus,” translates as “mind and hand.”

This interest in the interaction of the physical with the ineffable continues to guide her art, which essentially renders communication tactile. “Chloe’s work is so much about listening to materials and finding ways to hear how they talk, hear the sounds that they make,” Lieberman says. This approach is in evidence at a forthcoming exhibition “Tisser L’Hybride: Chloe Bensahel” at the Palais de Tokyo in Paris, which features three interactive tapestries. According to Bensahel, the artwork in the exhibit and what she is doing at MIT are “not going to be directly connected,” but she also points out that “they benefit from one another, for sure.”

Indeed, keeping an open mind to different fields and different ways of thinking has been enriching Bensahel’s time on campus. In addition to such public-facing activities as a presentation and demonstration at the MIT Museum’s After Dark series, in March, she has been actively collaborating with various entities, faculty, and students. For instance, she has been leveraging prototyping equipment and exploring potential industrial applications of her work with the public-private partnership Advanced Functional Fabrics of America, of which MIT is a member. “I love that something that could  be in a museum could also be in a hospital,” Bensahel says. AFFOA staff members Jesse Jur, director of technical program development, and Frannie Logan, textile technologist, have been providing technical support as well.

Thriving on collaboration

Interlocutors on campus include Azra Akšamija, the director of the MIT Future Heritage Lab, and Vera van de Seyp, a research assistant in the Future Sketches group, whose interests and experiences complement Bensahel’s. “A lot of my work is text-based and I’m not a typography or graphic designer at all, so it’s really nice to work with Vera, because what we're essentially doing is thinking about form and function at the same time,” Bensahel says. “I’m working on how I can make a textile that can be magnetized, in the way that magnetic core memory was magnetic. I would like for it to tense up or move in different ways, so that essentially you have a textile that can assemble in different ways.”

Most of all, perhaps, it’s the constant intellectual activity at MIT that has spurred and inspired Bensahel, who relishes the opportunity to integrate perspectives that are new to her. “I’ve had a lot of really eye-opening conversations on what magnetism means,” she says. “I just had lunch with a researcher and she was like, ‘Bacteria sometimes have magnetic fields to know how to grow.’ This place, it's really about the people,” Bensahel continues. “It’s a very dense group of brilliant people so no matter who you're running into, they’re going to have this very powerful depth of knowledge in one specific field. Being here also shifted my perspective: I didn’t really consider myself a researcher, or a scientist for that matter, and I feel more comfortable in that space now. Every day, I find new applications or new directions.”

de MIT News https://ift.tt/nHocGZf

Three from MIT named 2024-25 Goldwater Scholars

MIT students Ben Lou, Srinath Mahankali, and Kenta Suzuki have been selected to receive Barry Goldwater Scholarships for the 2024-25 academic year. They are among just 438 recipients from across the country selected based on academic merit from an estimated pool of more than 5,000 college sophomores and juniors, approximately 1,350 of whom were nominated by their academic institution to compete for the scholarship.

Since 1989, the Barry Goldwater Scholarship and Excellence in Education Foundation has awarded nearly 11,000 Goldwater scholarships to support undergraduates who intend to pursue research careers in the natural sciences, mathematics, and engineering and have the potential to become leaders in their respective fields. Past scholars have gone on to win an impressive array of prestigious postgraduate fellowships. Almost all, including the three MIT recipients, intend to obtain doctorates in their area of research.

Ben Lou

Ben Lou is a third-year student originally from San Diego, California, majoring in physics and math with a minor in philosophy.

“My research interests are scattered across different disciplines,” says Lou. “I want to draw from a wide range of topics in math and physics, finding novel connections between them, to push forward the frontier of knowledge.”

Since January 2022, he has worked with Nergis Mavalvala, dean of the School of Science, and Hudson Loughlin, a graduate student in the LIGO group, which studies the detection of gravitational waves. Lou is working with them to advance the field of quantum measurement and better understand quantum gravity.

“Ben has enormous intellectual horsepower and works with remarkable independence,” writes Mavalvala in her recommendation letter. “I have no doubt he has an outstanding career in physics ahead of him.”

Lou, for his part, is grateful to Mavalvala and Loughlin, as well as all of his scientific mentors that have supported him along his research path. That includes MIT professors Alan Guth and Barton Zwiebach, who introduced him to quantum physics, as well as his first-year advisor, Richard Price; current advisor, Janet Conrad; Elijah Bodish and Roman Bezrukavnikov in the Department of Mathematics; and David W. Brown of the San Diego Math Circle.

In terms of his future career goals, Lou wants to be a professor of theoretical physics and study, as he says, the “fundamental aspects of reality” while also inspiring students to love math and physics.

In addition to his research, Lou is currently the vice president of the Assistive Technology Club at MIT and actively engaged in raising money for Spinal Muscular Atrophy research. In the future, he’d like to continue his philanthropy work and use his personal experience to advise an assistive technology company.

Srinath Mahankali

Srinath Mahankali is a third-year student from New York City majoring in computer science.

Since June 2022, Mahankali has been an undergraduate researcher in the MIT Computer Science and Artificial Intelligence Laboratory. Working with Pulkit Agrawal, assistant professor of electrical engineering and computer science and head of the Improbable AI Lab, Mahankali’s research is on training robots. Currently, his focus is on training quadruped robots to move in an energy-efficient manner and training agents to interact in environments with minimal feedback. But in the future, he’d like to develop robots that can complete athletic tasks like gymnastics.

“The experience of discussing research with Srinath is similar to discussions with the best PhD students in my group,” writes Agrawal in his recommendation letter. “He is fearless, willing to take risks, persistent, creative, and gets things done.”

Before coming to MIT, Mahanakli was a 2021 Regeneron STS scholar, which is one of the oldest and most prestigious awards for math and science students. In 2020, he was also a participant in the MIT PRIMES program, studying objective functions in optimization problems with Yunan Yang, an assistant professor of math at Cornell University.

“I’m deeply grateful to all my research advisors for their invaluable mentorship and guidance,” says Mahanakli, extending his thanks to PhD students Zhang-Wei Hong and Gabe Margolis, as well as assistant professor of math at Brandeis, Promit Ghosal, and all of the organizers of the PRIMES program. “I’m also very grateful to all the members of the Improbable AI Lab for their support, encouragement, and willingness to help and discuss any questions I have,”

In the future, Mahankali wants to obtain a PhD and one day lead his own lab in robotics and artificial intelligence.

Kenta Suzuki

Kenta Suzuki is a third-year student majoring in mathematics from Bloomfield Hills, Michigan, and Tokyo, Japan.

Currently, Suzuki works with professor of mathematics Roman Bezrukavnikov on research at the intersection of number and representation theory, using geometric methods to represent p-adic groups. Suzuki has also previously worked with math professors Wei Zhang and Zhiwei Yun, crediting the latter with inspiring him to pursue research in representation theory.

In his recommendation letter, Yun writes, “Kenta is the best undergraduate student that I have worked with in terms of the combination of raw talent, mathematical maturity, and research abilities.”

Before coming to MIT, Suzuki was a Yau Science Award USA finalist in 2020, receiving a gold in math, and he received honorable mention from the Davidson Institute Fellows program in 2021. He also participated in the MIT PRIMES program in 2020. Suzuki credits his PRIMES mentor, Michael Zieve at the University of Michigan, with giving him his first taste of mathematical research. In addition, he extended his thanks to all of his math mentors, including the organizers of MIT Summer Program in Undergraduate Research.

After MIT, Suzuki intends to obtain a PhD in pure math, continuing his research in representation theory and number theory and, one day, teaching at a research-oriented institution.

The Barry Goldwater Scholarship and Excellence in Education Program was established by U.S. Congress in 1986 to honor Senator Barry Goldwater, a soldier and national leader who served the country for 56 years. Awardees receive scholarships of up to $7,500 a year to cover costs related to tuition, room and board, fees, and books.

de MIT News https://ift.tt/UFT9cty

Physicists arrange atoms in extremely close proximity

Proximity is key for many quantum phenomena, as interactions between atoms are stronger when the particles are close. In many quantum simulators, scientists arrange atoms as close together as possible to explore exotic states of matter and build new quantum materials.

They typically do this by cooling the atoms to a stand-still, then using laser light to position the particles as close as 500 nanometers apart — a limit that is set by the wavelength of light. Now, MIT physicists have developed a technique that allows them to arrange atoms in much closer proximity, down to a mere 50 nanometers. For context, a red blood cell is about 1,000 nanometers wide.

The physicists demonstrated the new approach in experiments with dysprosium, which is the most magnetic atom in nature. They used the new approach to manipulate two layers of dysprosium atoms, and positioned the layers precisely 50 nanometers apart. At this extreme proximity, the magnetic interactions were 1,000 times stronger than if the layers were separated by 500 nanometers.

What’s more, the scientists were able to measure two new effects caused by the atoms’ proximity. Their enhanced magnetic forces caused “thermalization,” or the transfer of heat from one layer to another, as well as synchronized oscillations between layers. These effects petered out as the layers were spaced farther apart.

“We have gone from positioning atoms from 500 nanometers to 50 nanometers apart, and there is a lot you can do with this,” says Wolfgang Ketterle, the John D. MacArthur Professor of Physics at MIT. “At 50 nanometers, the behavior of atoms is so much different that we’re really entering a new regime here.”

Ketterle and his colleagues say the new approach can be applied to many other atoms to study quantum phenomena. For their part, the group plans to use the technique to manipulate atoms into configurations that could generate the first purely magnetic quantum gate — a key building block for a new type of quantum computer.

The team has published their results today in the journal Science. The study’s co-authors include lead author and physics graduate student Li Du, along with Pierre Barral, Michael Cantara, Julius de Hond, and Yu-Kun Lu — all members of the MIT-Harvard Center for Ultracold Atoms, the Department of Physics, and the Research Laboratory of Electronics at MIT.

Peaks and valleys

To manipulate and arrange atoms, physicists typically first cool a cloud of atoms to temperatures approaching absolute zero, then use a system of laser beams to corral the atoms into an optical trap.

Laser light is an electromagnetic wave with a specific wavelength (the distance between maxima of the electric field) and frequency. The wavelength limits the smallest pattern into which light can be shaped to typically 500 nanometers, the so-called optical resolution limit. Since atoms are attracted by laser light of certain frequencies, atoms will be positioned at the points of peak laser intensity. For this reason, existing techniques have been limited in how close they can position atomic particles, and could not be used to explore phenomena that happen at much shorter distances.

“Conventional techniques stop at 500 nanometers, limited not by the atoms but by the wavelength of light,” Ketterle explains. “We have found now a new trick with light where we can break through that limit.”

The team’s new approach, like current techniques, starts by cooling a cloud of atoms — in this case, to about 1 microkelvin, just a hair above absolute zero — at which point, the atoms come to a near-standstill. Physicists can then use lasers to move the frozen particles into desired configurations.

Then, Du and his collaborators worked with two laser beams, each with a different frequency, or color, and circular polarization, or direction of the laser’s electric field. When the two beams travel through a super-cooled cloud of atoms, the atoms can orient their spin in opposite directions, following either of the two lasers’ polarization. The result is that the beams produce two groups of the same atoms, only with opposite spins.

Each laser beam formed a standing wave, a periodic pattern of electric field intensity with a spatial period of 500 nanometers. Due to their different polarizations, each standing wave attracted and corralled one of two groups of atoms, depending on their spin. The lasers could be overlaid and tuned such that the distance between their respective peaks is as small as 50 nanometers, meaning that the atoms gravitating to each respective laser’s peaks would be separated by the same 50 nanometers.

But in order for this to happen, the lasers would have to be extremely stable and immune to all external noise, such as from shaking or even breathing on the experiment. The team realized they could stabilize both lasers by directing them through an optical fiber, which served to lock the light beams in place in relation to each other.

“The idea of sending both beams through the optical fiber meant the whole machine could shake violently, but the two laser beams stayed absolutely stable with respect to each others,” Du says.

Magnetic forces at close range

As a first test of their new technique, the team used atoms of dysprosium — a rare-earth metal that is one of the strongest magnetic elements in the periodic table, particularly at ultracold temperatures. However, at the scale of atoms, the element’s magnetic interactions are relatively weak at distances of even 500 nanometers. As with common refrigerator magnets, the magnetic attraction between atoms increases with proximity, and the scientists suspected that if their new technique could space dysprosium atoms as close as 50 nanometers apart, they might observe the emergence of otherwise weak interactions between the magnetic atoms.

“We could suddenly have magnetic interactions, which used to be almost neglible but now are really strong,” Ketterle says.

The team applied their technique to dysprosium, first super-cooling the atoms, then passing two lasers through to split the atoms into two spin groups, or layers. They then directed the lasers through an optical fiber to stabilize them, and found that indeed, the two layers of dysprosium atoms gravitated to their respective laser peaks, which in effect separated the layers of atoms by 50 nanometers — the closest distance that any ultracold atom experiment has been able to achieve.

At this extremely close proximity, the atoms’ natural magnetic interactions were significantly enhanced, and were 1,000 times stronger than if they were positioned 500 nanometers apart. The team observed that these interactions resulted in two novel quantum phenomena: collective oscillation, in which one layer’s vibrations caused the other layer to vibrate in sync; and thermalization, in which one layer transferred heat to the other, purely through magnetic fluctuations in the atoms.

“Until now, heat between atoms could only by exchanged when they were in the same physical space and could collide,” Du notes. “Now we have seen atomic layers, separated by vacuum, and they exchange heat via fluctuating magnetic fields.”

The team’s results introduce a new technique that can be used to position many types of atom in close proximity. They also show that atoms, placed close enough together, can exhibit interesting quantum phenomena, that could be harnessed to build new quantum materials, and potentially, magnetically-driven atomic systems for quantum computers.

“We are really bringing super-resolution methods to the field, and it will become a general tool for doing quantum simulations,” Ketterle says. “There are many variants possible, which we are working on.”

This research was funded, in part, by the National Science Foundation and the Department of Defense.

de MIT News https://ift.tt/k1a3VAc

Epigenomic analysis sheds light on risk factors for ALS

For most patients, it’s unknown exactly what causes amyotrophic lateral sclerosis (ALS), a disease characterized by degeneration of motor neurons that impairs muscle control and eventually leads to death.

Studies have identified certain genes that confer a higher risk of the disease, but scientists believe there are many more genetic risk factors that have yet to be discovered. One reason why these drivers have been hard to find is that some are found in very few patients, making it hard to pick them out without a very large sample of patients. Additionally, some of the risk may be driven by epigenomic factors, rather than mutations in protein-coding genes.

Working with the Answer ALS consortium, a team of MIT researchers has analyzed epigenetic modifications — tags that determine which genes are turned on in a cell — in motor neurons derived from induced pluripotent stem (IPS) cells from 380 ALS patients.

This analysis revealed a strong differential signal associated with a known subtype of ALS, and about 30 locations with modifications that appear to be linked to rates of disease progression in ALS patients. The findings may help scientists develop new treatments that are targeted to patients with certain genetic risk factors.

“If the root causes are different for all these different versions of the disease, the drugs will be very different and the signals in IPS cells will be very different,” says Ernest Fraenkel, the Grover M. Hermann Professor in Health Sciences and Technology in MIT’s Department of Biological Engineering and the senior author of the study. “We may get to a point in a decade or so where we don’t even think of ALS as one disease, where there are drugs that are treating specific types of ALS that only work for one group of patients and not for another.”

MIT postdoc Stanislav Tsitkov is the lead author of the paper, which appears today in Nature Communications.

Finding risk factors

ALS is a rare disease that is estimated to affect about 30,000 people in the United States. One of the challenges in studying the disease is that while genetic variants are believed to account for about 50 percent of ALS risk (with environmental factors making up the rest), most of the variants that contribute to that risk have not been identified.

Similar to Alzheimer’s disease, there may be a large number of genetic variants that can confer risk, but each individual patient may carry only a small number of those. This makes it difficult to identify the risk factors unless scientists have a very large population of patients to analyze.

“Because we expect the disease to be heterogeneous, you need to have large numbers of patients before you can pick up on signals like this. To really be able to classify the subtypes of disease, we’re going to need to look at a lot of people,” Fraenkel says.

About 10 years ago, the Answer ALS consortium began to collect large numbers of patient samples, which could allow for larger-scale studies that might reveal some of the genetic drivers of the disease. From blood samples, researchers can create induced pluripotent stem cells and then induce them to differentiate into motor neurons, the cells most affected by ALS.

“We don’t think all ALS patients are going to be the same, just like all cancers are not the same. And the goal is being able to find drivers of the disease that could be therapeutic targets,” Fraenkel says.

In this study, Fraenkel and his colleagues wanted to see if patient-derived cells could offer any information about molecular differences that are relevant to ALS. They focused on epigenomic modifications, using a method called ATAC-seq to measure chromatin density across the genome of each cell. Chromatin is a complex of DNA and proteins that determines which genes are accessible to be transcribed by the cell, depending on how densely packed the chromatin is.

In data that were collected and analyzed over several years, the researchers did not find any global signal that clearly differentiated the 380 ALS patients in their study from 80 healthy control subjects. However, they did find a strong differential signal associated with a subtype of ALS, characterized by a genetic mutation in the C9orf72 gene.

Additionally, they identified about 30 regions that were associated with slower rates of disease progression in ALS patients. Many of these regions are located near genes related to the cellular inflammatory response; interestingly, several of the identified genes have also been implicated in other neurodegenerative diseases, such as Parkinson’s disease.

“You can use a small number of these epigenomic regions and look at the intensity of the signal there, and predict how quickly someone’s disease will progress. That really validates the hypothesis that the epigenomics can be used as a filter to better understand the contribution of the person’s genome,” Fraenkel says.

“By harnessing the very large number of participant samples and extensive data collected by the Answer ALS Consortium, these studies were able to rigorously test whether the observed changes might be artifacts related to the techniques of sample collection, storage, processing, and analysis, or truly reflective of important biology,” says Lyle Ostrow, an associate professor of neurology at the Lewis Katz School of Medicine at Temple University, who was not involved in the study. “They developed standard ways to control for these variables, to make sure the results can be accurately compared. Such studies are incredibly important for accelerating ALS therapy development, as they will enable data and samples collected from different studies to be analyzed together.”

Targeted drugs

The researchers now hope to further investigate these genomic regions and see how they might drive different aspects of ALS progression in different subsets of patients. This could help scientists develop drugs that might work in different groups of patients, and help them identify which patients should be chosen for clinical trials of those drugs, based on genetic or epigenetic markers.

Last year, the U.S. Food and Drug Administration approved a drug called tofersen, which can be used in ALS patients with a mutation in a gene called SOD1. This drug is very effective for those patients, who make up about 1 percent of the total population of people with ALS. Fraenkel’s hope is that more drugs can be developed for, and tested in, people with other genetic drivers of ALS.

“If you had a drug like tofersen that works for 1 percent of patients and you just gave it to a typical phase two clinical trial, you probably wouldn’t have anybody with that mutation in the trial, and it would’ve failed. And so that drug, which is a lifesaver for people, would never have gotten through,” Fraenkel says.

The MIT team is now using an approach called quantitative trait locus (QTL) analysis to try to identify subgroups of ALS patients whose disease is driven by specific genomic variants.

“We can integrate the genomics, the transcriptomics, and the epigenomics, as a way to find subgroups of ALS patients who have distinct phenotypic signatures from other ALS patients and healthy controls,” Tsitkov says. “We have already found a few potential hits in that direction.”

The research was funded by the Answer ALS program, which is supported by the Robert Packard Center for ALS Research at Johns Hopkins University, Travelers Insurance, ALS Finding a Cure Foundation, Stay Strong Vs. ALS, Answer ALS Foundation, Microsoft, Caterpillar Foundation, American Airlines, Team Gleason, the U.S. National Institutes of Health, Fishman Family Foundation, Aviators Against ALS, AbbVie Foundation, Chan Zuckerberg Initiative, ALS Association, National Football League, F. Prime, M. Armstrong, Bruce Edwards Foundation, the Judith and Jean Pape Adams Charitable Foundation, Muscular Dystrophy Association, Les Turner ALS Foundation, PGA Tour, Gates Ventures, and Bari Lipp Foundation. This work was also supported, in part, by grants from the National Institutes of Health and the MIT-GSK Gertrude B. Elion Research Fellowship Program for Drug Discovery and Disease.

de MIT News https://ift.tt/UgHlRLc

Fostering research, careers, and community in materials science

Gabrielle Wood, a junior at Howard University majoring in chemical engineering, is on a mission to improve the sustainability and life cycles of natural resources and materials. Her work in the Materials Initiative for Comprehensive Research Opportunity (MICRO) program has given her hands-on experience with many different aspects of research, including MATLAB programming, experimental design, data analysis, figure-making, and scientific writing.

Wood is also one of 10 undergraduates from 10 universities around the United States to participate in the first MICRO Summit earlier this year. The internship program, developed by the MIT Department of Materials Science and Engineering (DMSE), first launched in fall 2021. Now in its third year, the program continues to grow, providing even more opportunities for non-MIT undergraduate students — including the MICRO Summit and the program’s expansion to include Northwestern University.

“I think one of the most valuable aspects of the MICRO program is the ability to do research long term with an experienced professor in materials science and engineering,” says Wood. “My school has limited opportunities for undergraduate research in sustainable polymers, so the MICRO program allowed me to gain valuable experience in this field, which I would not otherwise have.”

Like Wood, Griheydi Garcia, a senior chemistry major at Manhattan College, values the exposure to materials science, especially since she is not able to learn as much about it at her home institution.

“I learned a lot about crystallography and defects in materials through the MICRO curriculum, especially through videos,” says Garcia. “The research itself is very valuable, as well, because we get to apply what we’ve learned through the videos in the research we do remotely.”

Expanding research opportunities

From the beginning, the MICRO program was designed as a fully remote, rigorous education and mentoring program targeted toward students from underserved backgrounds interested in pursuing graduate school in materials science or related fields. Interns are matched with faculty to work on their specific research interests.

Jessica Sandland ’99, PhD ’05, principal lecturer in DMSE and co-founder of MICRO, says that research projects for the interns are designed to be work that they can do remotely, such as developing a machine-learning algorithm or a data analysis approach.

“It’s important to note that it’s not just about what the program and faculty are bringing to the student interns,” says Sandland, a member of the MIT Digital Learning Lab, a joint program between MIT Open Learning and the Institute’s academic departments. “The students are doing real research and work, and creating things of real value. It’s very much an exchange.”

Cécile Chazot PhD ’22, now an assistant professor of materials science and engineering at Northwestern University, had helped to establish MICRO at MIT from the very beginning. Once at Northwestern, she quickly realized that expanding MICRO to Northwestern would offer even more research opportunities to interns than by relying on MIT alone — leveraging the university’s strong materials science and engineering department, as well as offering resources for biomaterials research through Northwestern’s medical school. The program received funding from 3M and officially launched at Northwestern in fall 2023. Approximately half of the MICRO interns are now in the program with MIT and half are with Northwestern. Wood and Garcia both participate in the program via Northwestern.

“By expanding to another school, we’ve been able to have interns work with a much broader range of research projects,” says Chazot. “It has become easier for us to place students with faculty and research that match their interests.”

Building community

The MICRO program received a Higher Education Innovation grant from the Abdul Latif Jameel World Education Lab, part of MIT Open Learning, to develop an in-person summit. In January 2024, interns visited MIT for three days of presentations, workshops, and campus tours — including a tour of the MIT.nano building — as well as various community-building activities.

“A big part of MICRO is the community,” says Chazot. “A highlight of the summit was just seeing the students come together.”

The summit also included panel discussions that allowed interns to gain insights and advice from graduate students and professionals. The graduate panel discussion included MIT graduate students Sam Figueroa (mechanical engineering), Isabella Caruso (DMSE), and Eliana Feygin (DMSE). The career panel was led by Chazot and included Jatin Patil PhD ’23, head of product at SiTration; Maureen Reitman ’90, ScD ’93, group vice president and principal engineer at Exponent; Lucas Caretta PhD ’19, assistant professor of engineering at Brown University; Raquel D’Oyen ’90, who holds a PhD from Northwestern University and is a senior engineer at Raytheon; and Ashley Kaiser MS ’19, PhD ’21, senior process engineer at 6K.

Students also had an opportunity to share their work with each other through research presentations. Their presentations covered a wide range of topics, including: developing a computer program to calculate solubility parameters for polymers used in textile manufacturing; performing a life-cycle analysis of a photonic chip and evaluating its environmental impact in comparison to a standard silicon microchip; and applying machine learning algorithms to scanning transmission electron microscopy images of CrSBr, a two-dimensional magnetic material. 

“The summit was wonderful and the best academic experience I have had as a first-year college student,” says MICRO intern Gabriella La Cour, who is pursuing a major in chemistry and dual degree biomedical engineering at Spelman College and participates in MICRO through MIT. “I got to meet so many students who were all in grades above me … and I learned a little about how to navigate college as an upperclassman.” 

“I actually have an extremely close friendship with one of the students, and we keep in touch regularly,” adds La Cour. “Professor Chazot gave valuable advice about applications and recommendation letters that will be useful when I apply to REUs [Research Experiences for Undergraduates] and graduate schools.”

Looking to the future, MICRO organizers hope to continue to grow the program’s reach.

“We would love to see other schools taking on this model,” says Sandland. “There are a lot of opportunities out there. The more departments, research groups, and mentors that get involved with this program, the more impact it can have.”

de MIT News https://ift.tt/wmkb0ce

Natural language boosts LLM performance in coding, planning, and robotics

Large language models (LLMs) are becoming increasingly useful for programming and robotics tasks, but for more complicated reasoning problems, the gap between these systems and humans looms large. Without the ability to learn new concepts like humans do, these systems fail to form good abstractions — essentially, high-level representations of complex concepts that skip less-important details — and thus sputter when asked to do more sophisticated tasks.

Luckily, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) researchers have found a treasure trove of abstractions within natural language. In three papers to be presented at the International Conference on Learning Representations this month, the group shows how our everyday words are a rich source of context for language models, helping them build better overarching representations for code synthesis, AI planning, and robotic navigation and manipulation.

The three separate frameworks build libraries of abstractions for their given task: LILO (library induction from language observations) can synthesize, compress, and document code; Ada (action domain acquisition) explores sequential decision-making for artificial intelligence agents; and LGA (language-guided abstraction) helps robots better understand their environments to develop more feasible plans. Each system is a neurosymbolic method, a type of AI that blends human-like neural networks and program-like logical components.

LILO: A neurosymbolic framework that codes

Large language models can be used to quickly write solutions to small-scale coding tasks, but cannot yet architect entire software libraries like the ones written by human software engineers. To take their software development capabilities further, AI models need to refactor (cut down and combine) code into libraries of succinct, readable, and reusable programs.

Refactoring tools like the previously developed MIT-led Stitch algorithm can automatically identify abstractions, so, in a nod to the Disney movie “Lilo & Stitch,” CSAIL researchers combined these algorithmic refactoring approaches with LLMs. Their neurosymbolic method LILO uses a standard LLM to write code, then pairs it with Stitch to find abstractions that are comprehensively documented in a library.

LILO’s unique emphasis on natural language allows the system to do tasks that require human-like commonsense knowledge, such as identifying and removing all vowels from a string of code and drawing a snowflake. In both cases, the CSAIL system outperformed standalone LLMs, as well as a previous library learning algorithm from MIT called DreamCoder, indicating its ability to build a deeper understanding of the words within prompts. These encouraging results point to how LILO could assist with things like writing programs to manipulate documents like Excel spreadsheets, helping AI answer questions about visuals, and drawing 2D graphics.

“Language models prefer to work with functions that are named in natural language,” says Gabe Grand SM '23, an MIT PhD student in electrical engineering and computer science, CSAIL affiliate, and lead author on the research. “Our work creates more straightforward abstractions for language models and assigns natural language names and documentation to each one, leading to more interpretable code for programmers and improved system performance.”

When prompted on a programming task, LILO first uses an LLM to quickly propose solutions based on data it was trained on, and then the system slowly searches more exhaustively for outside solutions. Next, Stitch efficiently identifies common structures within the code and pulls out useful abstractions. These are then automatically named and documented by LILO, resulting in simplified programs that can be used by the system to solve more complex tasks.

The MIT framework writes programs in domain-specific programming languages, like Logo, a language developed at MIT in the 1970s to teach children about programming. Scaling up automated refactoring algorithms to handle more general programming languages like Python will be a focus for future research. Still, their work represents a step forward for how language models can facilitate increasingly elaborate coding activities.

Ada: Natural language guides AI task planning

Just like in programming, AI models that automate multi-step tasks in households and command-based video games lack abstractions. Imagine you’re cooking breakfast and ask your roommate to bring a hot egg to the table — they’ll intuitively abstract their background knowledge about cooking in your kitchen into a sequence of actions. In contrast, an LLM trained on similar information will still struggle to reason about what they need to build a flexible plan.

Named after the famed mathematician Ada Lovelace, who many consider the world’s first programmer, the CSAIL-led “Ada” framework makes headway on this issue by developing libraries of useful plans for virtual kitchen chores and gaming. The method trains on potential tasks and their natural language descriptions, then a language model proposes action abstractions from this dataset. A human operator scores and filters the best plans into a library, so that the best possible actions can be implemented into hierarchical plans for different tasks.

“Traditionally, large language models have struggled with more complex tasks because of problems like reasoning about abstractions,” says Ada lead researcher Lio Wong, an MIT graduate student in brain and cognitive sciences, CSAIL affiliate, and LILO coauthor. “But we can combine the tools that software engineers and roboticists use with LLMs to solve hard problems, such as decision-making in virtual environments.”

When the researchers incorporated the widely-used large language model GPT-4 into Ada, the system completed more tasks in a kitchen simulator and Mini Minecraft than the AI decision-making baseline “Code as Policies.” Ada used the background information hidden within natural language to understand how to place chilled wine in a cabinet and craft a bed. The results indicated a staggering 59 and 89 percent task accuracy improvement, respectively.

With this success, the researchers hope to generalize their work to real-world homes, with the hopes that Ada could assist with other household tasks and aid multiple robots in a kitchen. For now, its key limitation is that it uses a generic LLM, so the CSAIL team wants to apply a more powerful, fine-tuned language model that could assist with more extensive planning. Wong and her colleagues are also considering combining Ada with a robotic manipulation framework fresh out of CSAIL: LGA (language-guided abstraction).

Language-guided abstraction: Representations for robotic tasks

Andi Peng SM ’23, an MIT graduate student in electrical engineering and computer science and CSAIL affiliate, and her coauthors designed a method to help machines interpret their surroundings more like humans, cutting out unnecessary details in a complex environment like a factory or kitchen. Just like LILO and Ada, LGA has a novel focus on how natural language leads us to those better abstractions.

In these more unstructured environments, a robot will need some common sense about what it’s tasked with, even with basic training beforehand. Ask a robot to hand you a bowl, for instance, and the machine will need a general understanding of which features are important within its surroundings. From there, it can reason about how to give you the item you want. 

In LGA’s case, humans first provide a pre-trained language model with a general task description using natural language, like “bring me my hat.” Then, the model translates this information into abstractions about the essential elements needed to perform this task. Finally, an imitation policy trained on a few demonstrations can implement these abstractions to guide a robot to grab the desired item.

Previous work required a person to take extensive notes on different manipulation tasks to pre-train a robot, which can be expensive. Remarkably, LGA guides language models to produce abstractions similar to those of a human annotator, but in less time. To illustrate this, LGA developed robotic policies to help Boston Dynamics’ Spot quadruped pick up fruits and throw drinks in a recycling bin. These experiments show how the MIT-developed method can scan the world and develop effective plans in unstructured environments, potentially guiding autonomous vehicles on the road and robots working in factories and kitchens.

“In robotics, a truth we often disregard is how much we need to refine our data to make a robot useful in the real world,” says Peng. “Beyond simply memorizing what’s in an image for training robots to perform tasks, we wanted to leverage computer vision and captioning models in conjunction with language. By producing text captions from what a robot sees, we show that language models can essentially build important world knowledge for a robot.”

The challenge for LGA is that some behaviors can’t be explained in language, making certain tasks underspecified. To expand how they represent features in an environment, Peng and her colleagues are considering incorporating multimodal visualization interfaces into their work. In the meantime, LGA provides a way for robots to gain a better feel for their surroundings when giving humans a helping hand. 

An “exciting frontier” in AI

“Library learning represents one of the most exciting frontiers in artificial intelligence, offering a path towards discovering and reasoning over compositional abstractions,” says assistant professor at the University of Wisconsin-Madison Robert Hawkins, who was not involved with the papers. Hawkins notes that previous techniques exploring this subject have been “too computationally expensive to use at scale” and have an issue with the lambdas, or keywords used to describe new functions in many languages, that they generate. “They tend to produce opaque 'lambda salads,' big piles of hard-to-interpret functions. These recent papers demonstrate a compelling way forward by placing large language models in an interactive loop with symbolic search, compression, and planning algorithms. This work enables the rapid acquisition of more interpretable and adaptive libraries for the task at hand.”

By building libraries of high-quality code abstractions using natural language, the three neurosymbolic methods make it easier for language models to tackle more elaborate problems and environments in the future. This deeper understanding of the precise keywords within a prompt presents a path forward in developing more human-like AI models.

MIT CSAIL members are senior authors for each paper: Joshua Tenenbaum, a professor of brain and cognitive sciences, for both LILO and Ada; Julie Shah, head of the Department of Aeronautics and Astronautics, for LGA; and Jacob Andreas, associate professor of electrical engineering and computer science, for all three. The additional MIT authors are all PhD students: Maddy Bowers and Theo X. Olausson for LILO, Jiayuan Mao and Pratyusha Sharma for Ada, and Belinda Z. Li for LGA. Muxin Liu of Harvey Mudd College was a coauthor on LILO; Zachary Siegel of Princeton University, Jaihai Feng of the University of California at Berkeley, and Noa Korneev of Microsoft were coauthors on Ada; and Ilia Sucholutsky, Theodore R. Sumers, and Thomas L. Griffiths of Princeton were coauthors on LGA. 

LILO and Ada were supported, in part, by ​​MIT Quest for Intelligence, the MIT-IBM Watson AI Lab, Intel, U.S. Air Force Office of Scientific Research, the U.S. Defense Advanced Research Projects Agency, and the U.S. Office of Naval Research, with the latter project also receiving funding from the Center for Brains, Minds and Machines. LGA received funding from the U.S. National Science Foundation, Open Philanthropy, the Natural Sciences and Engineering Research Council of Canada, and the U.S. Department of Defense.

de MIT News https://ift.tt/85Jwmzb

Nuno Loureiro named director of MIT’s Plasma Science and Fusion Center

Nuno Loureiro, professor of nuclear science and engineering and of physics, has been appointed the new director of the MIT Plasma Science and Fusion Center, effective May 1.

Loureiro is taking the helm of one of MIT’s largest labs: more than 250 full-time researchers, staff members, and students work and study in seven buildings with 250,000 square feet of lab space. A theoretical physicist and fusion scientist, Loureiro joined MIT as a faculty member in 2016, and was appointed deputy director of the Plasma Science and Fusion Center (PSFC) in 2022. Loureiro succeeds Dennis Whyte, who stepped down at the end of 2023 to return to teaching and research.

Stepping into his new role as director, Loureiro says, “The PSFC has an impressive tradition of discovery and leadership in plasma and fusion science and engineering. Becoming director of the PSFC is an incredible opportunity to shape the future of these fields. We have a world-class team, and it’s an honor to be chosen as its leader.”

Loureiro’s own research ranges widely. He is recognized for advancing the understanding of multiple aspects of plasma behavior, particularly turbulence and the physics underpinning solar flares and other astronomical phenomena. In the fusion domain, his work enables the design of fusion devices that can more efficiently control and harness the energy of fusing plasmas, bringing the dream of clean, near-limitless fusion power that much closer. 

Plasma physics is foundational to advancing fusion science, a fact Loureiro has embraced and that is relevant as he considers the direction of the PSFC’s multidisciplinary research. “But plasma physics is only one aspect of our focus. Building a scientific agenda that continues and expands on the PSFC’s history of innovation in all aspects of fusion science and engineering is vital, and a key facet of that work is facilitating our researchers’ efforts to produce the breakthroughs that are necessary for the realization of fusion energy.”

As the climate crisis accelerates, fusion power continues to grow in appeal: It produces no carbon emissions, its fuel is plentiful, and dangerous “meltdowns” are impossible. The sooner that fusion power is commercially available, the greater impact it can have on reducing greenhouse gas emissions and meeting global climate goals. While technical challenges remain, “the PSFC is well poised to meet them, and continue to show leadership. We are a mission-driven lab, and our students and staff are incredibly motivated,” Loureiro comments.

“As MIT continues to lead the way toward the delivery of clean fusion power onto the grid, I have no doubt that Nuno is the right person to step into this key position at this critical time,” says Maria T. Zuber, MIT’s presidential advisor for science and technology policy. “I look forward to the steady advance of plasma physics and fusion science at MIT under Nuno’s leadership.”

Over the last decade, there have been massive leaps forward in the field of fusion energy, driven in part by innovations like high-temperature superconducting magnets developed at the PSFC. Further progress is guaranteed: Loureiro believes that “The next few years are certain to be an exciting time for us, and for fusion as a whole. It’s the dawn of a new era with burning plasma experiments” — a reference to the collaboration between the PSFC and Commonwealth Fusion Systems, a startup company spun out of the PSFC, to build SPARC, a fusion device that is slated to turn on in 2026 and produce a burning plasma that yields more energy than it consumes. “It’s going to be a watershed moment,” says Loureiro.

He continues, “In addition, we have strong connections to inertial confinement fusion experiments, including those at Lawrence Livermore National Lab, and we’re looking forward to expanding our research into stellarators, which are another kind of magnetic fusion device.” Over recent years, the PSFC has significantly increased its collaboration with industrial partners such Eni, IBM, and others. Loureiro sees great value in this: “These collaborations are mutually beneficial: they allow us to grow our research portfolio while advancing companies’ R&D efforts. It’s very dynamic and exciting.”

Loureiro’s directorship begins as the PSFC is launching key tech development projects like LIBRA, a “blanket” of molten salt that can be wrapped around fusion vessels and perform double duty as a neutron energy absorber and a breeder for tritium (the fuel for fusion). Researchers at the PSFC have also developed a way to rapidly test the durability of materials being considered for use in a fusion power plant environment, and are now creating an experiment that will utilize a powerful particle accelerator called a gyrotron to irradiate candidate materials.

Interest in fusion is at an all-time high; the demand for researchers and engineers, particularly in the nascent commercial fusion industry, is reflected by the record number of graduate students that are studying at the PSFC — more than 90 across seven affiliated MIT departments. The PSFC’s classrooms are full, and Loureiro notes a palpable sense of excitement. “Students are our greatest strength,” says Loureiro. “They come here to do world-class research but also to grow as individuals, and I want to give them a great place to do that. Supporting those experiences, making sure they can be as successful as possible is one of my top priorities.” Loureiro plans to continue teaching and advising students after his appointment begins.

MIT President Sally Kornbluth’s recently announced Climate Project is a clarion call for Loureiro: “It’s not hyperbole to say MIT is where you go to find solutions to humanity’s biggest problems,” he says. “Fusion is a hard problem, but it can be solved with resolve and ingenuity — characteristics that define MIT. Fusion energy will change the course of human history. It’s both humbling and exciting to be leading a research center that will play a key role in enabling that change.” 

de MIT News https://ift.tt/GASv2Fc