Ten with MIT connections win 2024 Hertz Foundation Fellowships

The Fannie and John Hertz Foundation announced that it has awarded fellowships to 10 PhD students with ties to MIT. The prestigious award provides each recipient with five years of doctoral-level research funding (up to a total of $250,000), which allows them the flexibility and autonomy to pursue their own innovative ideas.

Fellows also receive lifelong access to Hertz Foundation programs, such as events, mentoring, and networking. They join the ranks of over 1,300 former Hertz Fellows who are leaders and scholars in a range of fields in science, engineering, and technology. Connections among fellows over the years have sparked collaborations in startups, research, and technology commercialization.

The 10 MIT recipients are among a total of 18 Hertz Foundation Fellows scholars selected this year from across the country. Five of them received their undergraduate degrees at the Institute and will pursue their PhDs at other schools. Two are current MIT graduate students, and four will begin their studies here in the fall.

“For more than 60 years, Hertz Fellows have led scientific and technical innovation in national security, applied biological sciences, materials research, artificial intelligence, space exploration, and more. Their contributions have been essential in advancing U.S. competitiveness,” says Stephen Fantone, chair of the Hertz Foundation board of directors and founder and president of Optikos Corp. “I’m excited to watch our newest Hertz Fellows as they pursue challenging research and continue the strong tradition of applying their work for the greater good.”

This year’s MIT-affiliated awardees are:

Owen Dugan ’24 graduated from MIT in just two-and-a-half years with a degree in physics, and he plans to pursue a PhD in computer science at Stanford University. His research interests lie at the intersection of AI and physics. As an undergraduate, he conducted research in a broad range of areas, including using physics concepts to enhance the speed of large language models and developing machine learning algorithms that automatically discover scientific theories. He was recognized with MIT’s Outstanding Undergraduate Research Award and is a U.S. Presidential Scholar, a Neo Scholar, and a Knight-Hennessy Scholar. Dugan holds multiple patents, co-developed an app to reduce food waste, and co-founded a startup that builds tools to verify the authenticity of digital images.

Kaylie Hausknecht will begin her physics doctorate at MIT in the fall, having completing her undergraduate degree in physics and astrophysics at Harvard University. While there, her undergraduate research focused on developing new machine learning techniques to solve problems in a range of fields, such as fluid dynamics, astrophysics, and condensed matter physics. She received the Hoopes Prize for her senior thesis, was inducted into Phi Beta Kappa as a junior, and won two major writing awards. In addition, she completed five NASA internships. As an intern, she helped identify 301 new exoplanets using archival data from the Kepler Space Telescope. Hausknecht served as the co-president of Harvard’s chapter of Science Club for Girls, which works to encourage girls from underrepresented backgrounds to pursue STEM.

Elijah Lew-Smith majored in physics at Brown University and plans to pursue a doctoral degree in physics at MIT. He is a theoretical physicist with broad intellectual interests in effective field theory (EFT), which is the study of systems with many interacting degrees of freedom. EFT reveals how to extract the relevant, long-distance behavior from complicated microscopic rules. In 2023, he received a national award to work on applying EFT systematically to non-equilibrium and active systems such as fluctuating hydrodynamics or flocking birds. In addition, Lew-Smith received a scholarship from the U.S. State Department to live for a year in Dakar, Senegal, and later studied at ’École Polytechnique in Paris, France.

Rupert Li ’24 earned his bachelor’s and master’s degrees at MIT in mathematics as well as computer science, data science, and economics, with a minor in business analytics.He was named a 2024 Marshall Scholar and will study abroad for a year at Cambridge University before matriculating at Stanford University for a mathematics doctorate. As an undergraduate, Li authored 12 math research articles, primarily in combinatorics, but also including discrete geometry, probability, and harmonic analysis. He was recognized for his work with a Barry Goldwater Scholarship and an honorable mention for the Morgan Prize, one of the highest undergraduate honors in mathematics.

Amani Maina-Kilaas is a first-year doctoral student at MIT in the Department of Brain and Cognitive Sciences, where he studies computational psycholinguistics. In particular, he is interested in using artificial intelligence as a scientific tool to study how the mind works, and using what we know about the mind to develop more cognitively realistic models. Maina-Kilaas earned his bachelor’s degree in computer science and mathematics from Harvey Mudd College. There, he conducted research regarding intention perception and theoretical machine learning, earning the Astronaut Scholarship and Computing Research Association’s Outstanding Undergraduate Researcher Award.

Zoë Marschner ’23 is a doctoral student at Carnegie Mellon University working on geometry processing, a subfield of computer graphics focused on how to represent and work with geometric data digitally; in her research, she aims to make these representations capable of enabling fundamentally better algorithms for solving geometric problems across science and engineering. As an undergraduate at MIT, she earned a bachelor’s degree in computer science and math and pursued research in geometry processing, including repairing hexahedral meshes and detecting intersections between high-order surfaces. She also interned at Walt Disney Animation Studios, where she worked on collision detection algorithms for simulation. Marschner is a recipient of the National Science Foundation’s Graduate Research Fellowship and the Goldwater Scholarship.

Zijian (William) Niu will start a doctoral program in computational and systems biology at MIT in the fall. He has a particular interest in developing new methods for imaging proteins and other biomolecules in their native cellular environments and using those data to build computational models for predicting their dynamics and molecular interactions. Niu received his bachelor’s degree in biochemistry, biophysics, and physics from the University of Pennsylvania. His undergraduate research involved developing novel computational methods for biological image analysis. He was awarded the Barry M. Goldwater Scholarship for creating a deep-learning algorithm for accurately detecting tiny diffraction-limited spots in fluorescence microscopy images that outperformed existing methods in quantifying spatial transcriptomics data.

James Roney received his bachelor’s and master’s degrees from Harvard University in computer science and statistics, respectively. He is currently working as a machine learning research engineer at D.E. Shaw Research. His past research has focused on interpreting the internal workings of AlphaFold and modeling cancer evolution. Roney plans to pursue a PhD in computational biology at MIT, with a specific interest in developing computational models of protein structure, function, and evolution and using those models to engineer novel proteins for applications in biotechnology.

Anna Sappington ’19 is a student in the Harvard University-MIT MD-PhD Program, currently in the first year of her doctoral program at MIT in electrical engineering and computer science. She is interested in building methods to predict evolutionary events, especially connections among machine learning, biology, and chemistry to develop reinforcement learning models inspired by evolutionary biology. Sappington graduated from MIT with a bachelor’s degree in computer science and molecular biology. As an undergraduate, she was awarded a 2018 Barry M. Goldwater Scholarship and selected as a Burchard Scholar and an Amgen Scholar. After graduating, she earned a master’s degree in genomic medicine from the University of Cambridge, where she studied as a Marshall Scholar, as well as a master’s degree in machine learning from University College London.

Jason Yang ’22 received his bachelor’s degree in biology with a minor in computer science from MIT and is currently a doctoral student in genetics at Stanford University. He is interested in understanding the biological processes that underlie human health and disease. At MIT, and subsequently at Massachusetts General Hospital, Yang worked on the mechanisms involved in neurodegeneration in repeat expansion diseases, uncovering a novel molecular consequence of repeat protein aggregation.

Advocating for science funding on Capitol Hill

This spring, 26 MIT students and postdocs traveled to Washington to meet with congressional staffers to advocate for increased science funding for fiscal year 2025. These conversations were impactful given the recent announcement of budget cuts for several federal science agencies for FY24. 

The participants met with 85 congressional offices representing 30 states over two days April 8-9. Overall, the group advocated for $89.46 billion in science funding across 11 federal scientific agencies. 

Every spring, the MIT Science Policy Initiative (SPI) organizes the Congressional Visit Days (CVD). The trip exposes participants to the process of U.S. federal policymaking and the many avenues researchers can use to advocate for scientific research. The participants also meet with Washington-based alumni and members of the MIT Washington Office and learn about policy careers.

This year, CVD was jointly co-organized by Marie Floryan and Andrew Fishberg, two PhD students in the departments of Mechanical Engineering and Aeronautics and Astronautics, respectively. Before the trip, the participants attended two training sessions organized by SPI, the MIT Washington Office, and the MIT Policy Lab. The participants learned how funding is appropriated at the federal level, the role of elected congressional officials and their staffers in the legislative process, and how academic researchers can get involved in advocating for policies for science.

Julian Ufert, a doctoral student in chemical engineering, says, “CVD was a remarkable opportunity to share insights from my research with policymakers, learn about U.S. politics, and serve the greater scientific community. I thoroughly enjoyed the contacts I made both on Capitol Hill and with MIT students and postdocs who share an interest in science policy.”

In addition to advocating for increased science funding, the participants advocated for topics pertaining to their research projects. A wide variety of topics were discussed, including AI, cybersecurity, energy production and storage, and biotechnology. Naturally, the recent advent of groundbreaking AI technologies, like ChatGPT, brought the topic of AI to the forefront of many offices interested, with multiple offices serving on the newly formed bipartisan AI Task Force.

These discussions were useful for both parties: The participants learned about the methods and challenges associated with enacting legislation, and the staffers directly heard from academic researchers about what is needed to promote scientific progress and innovation.

“It was fascinating to experience the interest and significant involvement of Congressional offices in policy matters related to science and technology. Most staffers were well aware of the general technological advancements and eager to learn more about how our research will impact society,” says Vipindev Vasudevan, a postdoc in electrical and computer engineering.

Dina Sharon, a PhD student in chemistry, adds, “The offices where we met with Congressional staffers were valuable classrooms! Our conversations provided insights into policymakers’ goals, how science can help reach these goals, and how scientists can help cultivate connections between the research and policy spheres.”

Participants also shared how science funding has directly impacted them, discussing how federal grants have supported their graduate education and for the need for open access research.

Unique professional development course prepares students for future careers

MIT’s unique Undergraduate Practice Opportunities Program (UPOP) is a yearlong career-development course for second-year students focused on preparing them for a summer experience in industry, research, and public service, as well as for their future careers post-MIT. The program was launched in 2001 by Thomas Magnanti, then dean of the MIT School of Engineering, who recognized that MIT students receive a best-in-class technical education, but hadn’t historically been given the opportunity to develop the softer skills that will help them succeed in the workplace.

“UPOP is a great opportunity for MIT sophomores to develop important skills that will complement what they are learning in the classroom and can help them to effectively communicate and demonstrate their value in a professional setting,” says Kendel Jester, assistant director for early career engagement in MIT Career Advising and Professional Development (CAPD). “Furthermore, the UPOP curriculum allows students to connect with tangible resources, including MIT alumni and staff, to help further their career and personal development.”

UPOP uses experiential learning to bolster students’ professional development and teaches them effective communication, teamwork, and problem-solving skills in an interactive environment. The program begins with career basics, like crafting a resume and cover letter, networking, and interview preparation, and progresses to more complex career readiness skills, such as negotiating a salary, professional communication, and fostering an inclusive environment.

“The biggest benefit of joining UPOP for me was the self-confidence which I gained as a professional,” says rising senior Jehan Ahmed. Ahmed completed UPOP in 2023 and continued on to work as a course assistant for the program. “Before starting my first industry internship, UPOP prepared me for the day-to-day collaboration which I experienced. I learned how to approach my manager from the beginning and set expectations and goals with them which became really helpful, especially as someone new to the industry. I felt more prepared to jump into my project even though I was not completely technically competent in the field as a sophomore.”

UPOP focuses on sophomores because they don’t receive as much support and targeted resources as first-year students. Completing the program gives sophomores a leg-up on summer opportunities, which are typically given precedent to juniors and seniors, by helping them become competitive candidates. The summer after sophomore year is a pivotal time in a career path, and UPOP helps its students get ahead.

“The time commitment of UPOP was low, but I got amazing connections and support systems through the program,” says rising senior Jade Durham. Durham is also a UPOP alum who returned to work for the program.

The UPOP community is a big benefit of joining the program. Students get access to UPOP’s exclusive mentor and employer networks, which opens doors to connections and opportunities that would not have been available to them otherwise. UPOP mentors are industry leaders, many of whom are MIT alumni. Meanwhile, UPOP’s 100-plus employer network members are invested in hiring UPOP students, knowing they are now equipped with skills that many other interns lack. In addition to access to these networks, students receive one-on-one advising with UPOP’s dedicated staff and exclusive opportunities to work with MIT campus partners.

“UPOP helps sophomores figure out what they want to do after graduation by connecting them with professionals in a variety of careers,” says Marianne Olsen, an MIT/UPOP alumna whose company, Chartwell, is part of the employer network. “I personally benefited significantly from meeting operations consultants through UPOP who helped me realize that a job existed that let me apply my engineering degree to manufacturing while having the variety of projects of consulting. Until then, I thought I’d have to pick one or the other.”

UPOP is a course offering three credits for the full year, but it boasts a much lighter workload and more flexibility than other classes at MIT. It consists of three or four hour-long milestone workshops during the fall and spring semesters, which cover the career readiness curriculum described above.

In addition to the milestone workshops, UPOP’s cornerstone event is Team Training Workshop (TTW), a multi-day, intensive experiential learning opportunity that places students in small teams assigned to UPOP mentors. Teams work together on a series of activities focused on building the skills they will need in their future professional endeavors, regardless of what their MIT course is. TTW’s unique programming immerses sophomores in a wide range of practice opportunities, such as project management, negotiations, and presenting professionally, while still prioritizing camaraderie and fun.

“Considering all the networking practice and professional skills that you get to learn from experienced mentors, TTW is definitely worth your time,” says Ahmed. “You get an opportunity to learn more about different fields of work from experts. You also get the chance to learn about the communication and emotional intelligence skills that are necessary to be successful at your job, which we may not get the chance to practice in our academic/technical classes.”

UPOP’s mission puts students’ career readiness needs as a No. 1 priority, and the program is constantly evolving to meet those needs. This year, UPOP started programming earlier than ever before to account for students whose chosen fields have internship application deadlines in the fall. This includes a brand-new First-Year Speed Networking event, which took place on April 23. The event gave prospective UPOP applicants a chance to hone their elevator pitch with each other and members of the MIT and UPOP community, including program alumni and employers within the network.

The UPOP application is now open. Admissions is rolling throughout the summer until it closes on Sept. 13.

“I would tell my first-year self that it was a great opportunity to build up my confidence for networking and a wonderful resource during the internship hunting season,” says Ahmed.

New technique reveals how gene transcription is coordinated in cells

The human genome contains about 23,000 genes, but only a fraction of those genes are turned on inside a cell at any given time. The complex network of regulatory elements that controls gene expression includes regions of the genome called enhancers, which are often located far from the genes that they regulate.

This distance can make it difficult to map the complex interactions between genes and enhancers. To overcome that, MIT researchers have invented a new technique that allows them to observe the timing of gene and enhancer activation in a cell. When a gene is turned on around the same time as a particular enhancer, it strongly suggests the enhancer is controlling that gene.

Learning more about which enhancers control which genes, in different types of cells, could help researchers identify potential drug targets for genetic disorders. Genomic studies have identified mutations in many non-protein-coding regions that are linked to a variety of diseases. Could these be unknown enhancers?

“When people start using genetic technology to identify regions of chromosomes that have disease information, most of those sites don’t correspond to genes. We suspect they correspond to these enhancers, which can be quite distant from a promoter, so it’s very important to be able to identify these enhancers,” says Phillip Sharp, an MIT Institute Professor Emeritus and member of MIT’s Koch Institute for Integrative Cancer Research.

Sharp is the senior author of the new study, which appears today in Nature. MIT Research Assistant D.B. Jay Mahat is the lead author of the paper.

Hunting for eRNA

Less than 2 percent of the human genome consists of protein-coding genes. The rest of the genome includes many elements that control when and how those genes are expressed. Enhancers, which are thought to turn genes on by coming into physical contact with gene promoter regions through transiently forming a complex, were discovered about 45 years ago.

More recently, in 2010, researchers discovered that these enhancers are transcribed into RNA molecules, known as enhancer RNA or eRNA. Scientists suspect that this transcription occurs when the enhancers are actively interacting with their target genes. This raised the possibility that measuring eRNA transcription levels could help researchers determine when an enhancer is active, as well as which genes it’s targeting.

“That information is extraordinarily important in understanding how development occurs, and in understanding how cancers change their regulatory programs and activate processes that lead to de-differentiation and metastatic growth,” Mahat says.

However, this kind of mapping has proven difficult to perform because eRNA is produced in very small quantities and does not last long in the cell. Additionally, eRNA lacks a modification known as a poly-A tail, which is the “hook” that most techniques use to pull RNA out of a cell.

One way to capture eRNA is to add a nucleotide to cells that halts transcription when incorporated into RNA. These nucleotides also contain a tag called biotin that can be used to fish the RNA out of a cell. However, this current technique only works on large pools of cells and doesn’t give information about individual cells.

While brainstorming ideas for new ways to capture eRNA, Mahat and Sharp considered using click chemistry, a technique that can be used to join two molecules together if they are each tagged with “click handles” that can react together.

The researchers designed nucleotides labeled with one click handle, and once these nucleotides are incorporated into growing eRNA strands, the strands can be fished out with a tag containing the complementary handle. This allowed the researchers to capture eRNA and then purify, amplify, and sequence it. Some RNA is lost at each step, but Mahat estimates that they can successfully pull out about 10 percent of the eRNA from a given cell.

Using this technique, the researchers obtained a snapshot of the enhancers and genes that are being actively transcribed at a given time in a cell.

“You want to be able to determine, in every cell, the activation of transcription from regulatory elements and from their corresponding gene. And this has to be done in a single cell because that’s where you can detect synchrony or asynchrony between regulatory elements and genes,” Mahat says.

Timing of gene expression

Demonstrating their technique in mouse embryonic stem cells, the researchers found that they could calculate approximately when a particular region starts to be transcribed, based on the length of the RNA strand and the speed of the polymerase (the enzyme responsible for transcription) — that is, how far the polymerase transcribes per second. This allowed them to determine which genes and enhancers were being transcribed around the same time.

The researchers used this approach to determine the timing of the expression of cell cycle genes in more detail than has previously been possible. They were also able to confirm several sets of known gene-enhancer pairs and generated a list of about 50,000 possible enhancer-gene pairs that they can now try to verify.

Learning which enhancers control which genes would prove valuable in developing new treatments for diseases with a genetic basis. Last year, the U.S. Food and Drug Administration approved the first gene therapy treatment for sickle cell anemia, which works by interfering with an enhancer that results in activation of a fetal globin gene, reducing the production of sickled blood cells.

The MIT team is now applying this approach to other types of cells, with a focus on autoimmune diseases. Working with researchers at Boston Children’s Hospital, they are exploring immune cell mutations that have been linked to lupus, many of which are found in non-coding regions of the genome.

“It’s not clear which genes are affected by these mutations, so we are beginning to tease apart the genes these putative enhancers might be regulating, and in what cell types these enhancers are active,” Mahat says. “This is a tool for creating gene-to-enhancer maps, which are fundamental in understanding the biology, and also a foundation for understanding disease.”

The findings of this study also offer evidence for a theory that Sharp has recently developed, along with MIT professors Richard Young and Arup Chakraborty, that gene transcription is controlled by membraneless droplets known as condensates. These condensates are made of large clusters of enzymes and RNA, which Sharp suggests may include eRNA produced at enhancer sites.

“We picture that the communication between an enhancer and a promoter is a condensate-type, transient structure, and RNA is part of that. This is an important piece of work in building the understanding of how RNAs from enhancers could be active,” he says.

The research was funded by the National Cancer Institute, the National Institutes of Health, and the Emerald Foundation Postdoctoral Transition Award. 

Physicists create five-lane superhighway for electrons

MIT physicists and colleagues have created a five-lane superhighway for electrons that could allow ultra-efficient electronics and more. 

The work, reported in the May 10 issue of Science, is one of several important discoveries by the same team over the past year involving a material that is a unique form of graphene.

“This discovery has direct implications for low-power electronic devices because no energy is lost during the propagation of electrons, which is not the case in regular materials where the electrons are scattered,” says Long Ju, an assistant professor in the Department of Physics and corresponding author of the Science paper.

The phenomenon is akin to cars traveling down an open turnpike as opposed to those moving through neighborhoods. The neighborhood cars can be stopped or slowed by other drivers making abrupt stops or U-turns that disrupt an otherwise smooth commute.

A new material

The material behind this work, known as rhombohedral pentalayer graphene, was discovered two years ago by physicists led by Ju. “We found a goldmine, and every scoop is revealing something new,” says Ju, who is also affiliated with MIT’s Materials Research Laboratory.

In a Nature Nanotechnology paper last October, Ju and colleagues reported the discovery of three important properties arising from rhombohedral graphene. For example, they showed that it could be topological, or allow the unimpeded movement of electrons around the edge of the material but not through the middle. That resulted in a superhighway, but required the application of a large magnetic field some tens of thousands times stronger than the Earth’s magnetic field.

In the current work, the team reports creating the superhighway without any magnetic field.

Tonghang Han, an MIT graduate student in physics, is a co-first author of the paper. “We are not the first to discover this general phenomenon, but we did so in a very different system. And compared to previous systems, ours is simpler and also supports more electron channels.” Explains Ju, “other materials can only support one lane of traffic on the edge of the material. We suddenly bumped it up to five.”

Additional co-first authors of the paper who contributed equally to the work are Zhengguang Lu and Yuxuan Yao. Lu is a postdoc in the Materials Research Laboratory. Yao conducted the work as a visiting undergraduate student from Tsinghua University. Other authors are MIT professor of physics Liang Fu; Jixiang Yang and Junseok Seo, both MIT graduate students in physics; Chiho Yoon and Fan Zhang of the University of Texas at Dallas; and Kenji Watanabe and Takashi Taniguchi of the National Institute for Materials Science in Japan.

How it works

Graphite, the primary component of pencil lead, is composed of many layers of graphene, a single layer of carbon atoms arranged in hexagons resembling a honeycomb structure. Rhombohedral graphene is composed of five layers of graphene stacked in a specific overlapping order.

Ju and colleagues isolated rhombohedral graphene thanks to a novel microscope Ju built at MIT in 2021 that can quickly and relatively inexpensively determine a variety of important characteristics of a material at the nanoscale. Pentalayer rhombohedral stacked graphene is only a few billionths of a meter thick.

In the current work, the team tinkered with the original system, adding a layer of tungsten disulfide (WS2). “The interaction between the WSand the pentalayer rhombohedral graphene resulted in this five-lane superhighway that operates at zero magnetic field,” says Ju.

Comparison to superconductivity

The phenomenon that the Ju group discovered in rhombohedral graphene that allows electrons to travel with no resistance at zero magnetic field is known as the quantum anomalous Hall effect. Most people are more familiar with superconductivity, a completely different phenomenon that does the same thing but happens in very different materials.

Ju notes that although superconductors were discovered in the 1910s, it took some 100 years of research to coax the system to work at the higher temperatures necessary for applications. “And the world record is still well below room temperature,” he notes.

Similarly, the rhombohedral graphene superhighway currently operates at about 2 kelvins, or -456 degrees Fahrenheit. “It will take a lot of effort to elevate the temperature, but as physicists, our job is to provide the insight; a different way for realizing this [phenomenon],” Ju says.

Very exciting

The discoveries involving rhombohedral graphene came as a result of painstaking research that wasn’t guaranteed to work. “We tried many recipes over many months,” says Han, “so it was very exciting when we cooled the system to a very low temperature and [a five-lane superhighway operating at zero magnetic field] just popped out.”

Says Ju, “it’s very exciting to be the first to discover a phenomenon in a new system, especially in a material that we uncovered.”

This work was supported by a Sloan Fellowship; the U.S. National Science Foundation; the U.S. Office of the Under Secretary of Defense for Research and Engineering; the Japan Society for the Promotion of Science KAKENHI; and the World Premier International Research Initiative of Japan.

SPURS Fellowships offer time out to reflect, learn, and connect

Sixteen international mid-career urban planners and public administrators recently bid farewell to the MIT campus, having completed a 10-month exploration of North American education and culture designed to expand their professional networks and infuse their work with new insights as they return to influential positions in government agencies, private firms, and other organizations throughout the developing world.

Hailing from Argentina, Bhutan, China, Egypt, Honduras, India, Kosovo, Mexico, Nepal, Pakistan, Trinidad & Tobago, Yemen, and Zimbabwe, they comprise this year’s group of MIT Special Program for Urban and Regional Studies (SPURS) Fellows. Founded in the Department of Urban Studies and Planning in 1967, SPURS has drawn from 135 countries to host more than 750 mid-career individuals who are or will be shaping policy in their home countries. Along with admitting several fellows directly into SPURS, MIT has competed successfully to be among 13 U.S. universities that also host a larger group of fellows annually selected and funded by the U.S. Department of State’s Hubert H. Humphrey Fellowship Program.

Recipients of the Humphrey Fellowship have their travel to the United States, living expenses, and other costs fully financed by the U.S. State Department. Perhaps equally valuable — and some say unique among international fellowships — is a focus that frees all fellows to explore beyond classroom teachings to learn, and advance their professional development without the pressure of earning a degree.

“This is the best reward of my life, this year at MIT and Cambridge in general,” says Carina Arvizu-Machado of Mexico, former cities director for Mexico and Colombia at the World Resources Institute and Mexico’s former national deputy secretary of urban development and housing, who is sponsored by the Humphrey Fellowship. “I think this year of stepping back and stepping out of the active life that we have as professionals and being able to reflect, to learn, to exchange ideas — it’s very useful.”

Arvizu-Machado’s sentiments are echoed by many past and present fellows, says Bish Sanyal, MIT’s Ford International Professor of Urban Development and Planning and director of SPURS since 2004.

“The fellows mention that this one year has given them a real opportunity to reflect on what they have done in the past and what they are going to do in the future,” he says, adding that the value of developing professional networks with peers in other developing countries can’t be overstated. “Some have never met colleagues from another country before. The program provides the ideal setting to reflect on professional challenges, collectively, without political concerns which stifle frank deliberation in their home countries.”

While some SPURS Fellows might not be well-traveled before coming to MIT, they are nonetheless a uniformly “very highly motivated and politically powerful group,” Sanyal says — movers and shakers in their home countries in fields such as urban planning, economics, governance, and business development. Some notable alumni include the current managing director of the International Monetary Fund, a former CEO of the World Bank, former ambassadors to the United States from Colombia and Haiti, the corporate vice president of strategic programming of Banco de Desarrollo de América Latina or CAF (Latin America’s largest development bank), and a Nepalese Supreme Court justice.

“When the Ebola outbreak happened in Africa, the person who headed the Ebola response team in Liberia was a SPURS Fellow,” Sanyal says.

The benefits of having a such an accomplished and cosmopolitan group of people on campus flow both ways, says Allan Goodman, CEO of the Institute of International Education (IIE), which administers the Humphrey Fellowship for the state department.

“It really enriches MIT … and all the places that are participating,” Goodman says. “The undergraduate and graduate students interact with the fellows, and they wouldn’t ordinarily have that chance. You have a ready-made group of international consultants who are focused on the theme of your department.”

Each university participating in the Humphrey Fellowship program is assigned fellows based on a specific area of expertise. With SPURS housed within the Department of Urban Studies and Planning at MIT, the programmatic focus is on urban and regional planning. Sanyal remarks that this focus is deliberate and consistent regardless of whether fellows are sponsored by the U.S. Department of State or other agencies from the fellows’ home countries. One difference, however, is that Humphrey Fellows are required to be professionally affiliated for at least six weeks with U.S.-based organizations in their areas of work or interest — an engagement described as a cross between an internship and pro-bono consultancy that provides fellows the opportunity to develop professional relationships with U.S. practitioners.

Peter Moran, director of the Humphrey program at IIE, says the biggest value to fellows at MIT and other participating universities is the ability to step out of their past professional lives and reflect from a fresh perspective on their professional aspirations to serve their nations in an interconnected world. In the process, they also benefit from the relationships with other fellows and professional partnerships that last years after they return home.

“To say it broadens your perspective really undersells it,” he says. “The diversity of the fellows is remarkable. It’s a lot of the world … and we are putting them all around the table together.”

By continuing to put fellows from diverse corners of the world together for over 50 years, SPURS has sparked lasting partnerships between fellows, as well as among SPURS alumni, MIT faculty and students, and other professionals they encounter during their time in Cambridge.

Two factors are key to maintaining the high quality of the program, Sanyal says.

First, additional funding could strengthen the program, and, to that end, he envisions sponsoring financially sustainable relationships with over a dozen local, national, and international agencies as long-term partners.

The second challenge is to revise the program’s objective in a rapidly changing world. This is harder to surmount. When SPURS was established in 1967, Sanyal says, there was widely held public perception that the United States ought to look outward to help democratic nations of the world.

“I think the challenge now is that many countries, including the U.S., are looking inward,” Sanyal says, adding that this inward turn increases the importance that SPURS develops a diverse portfolio of funding sources.

As Arvizu-Machado prepared to return to Mexico this spring, she recounted myriad positive experiences enabled by her fellowship — from lectures she was invited to give and graduate courses she attended to practicing yoga with her undergraduate dorm mates.

“Most important, I think, is the people I’ve met,” she says. “This includes, foremost, the other fellows. They are just amazing people. They have become part of my family. But also, some of the faculty and the extended network which this fellowship allows you to have access to. I’m very grateful to be part of this program.”

One of Arvizu-Machado’s co-fellows, Tenzin Jamtsho, agrees that the opportunity for personal connections with other fellows as well as with faculty highly respected in their fields is the aspect of SPURS that will continue to resonate when he returns to his native Bhutan. Jamtsho, director of administration and finance at Bhutan’s Druk Gyalpo’s Institute (formerly the Royal Academy), who is sponsored by the Humphrey Fellowship, says he pursued the fellowship after colleagues at home told him it would be “life changing.” His actual experience at MIT affirmed this expectation.

Jamtsho says the MIT campus offers fellows a “free-flowing environment” for learning, with opportunities to take whatever classes they’re interested in. During his fellowship, Jamtsho says he came to appreciate different ways to approach challenges — viewing problems through a “systems lens,” which he calls “a valuable skill that I am taking back home.”

Also returning to Bhutan with Jamtsho are some less-tangible aspects of his time at MIT.

“I’ve been fortunate to interact with people who are very intelligent and passionate,” he says. “What I’m going to take home is the kindness and humility of these people.”

Study models how ketamine’s molecular action leads to its effects on the brain

Ketamine, a World Health Organization Essential Medicine, is widely used at varying doses for sedation, pain control, general anesthesia, and as a therapy for treatment-resistant depression. While scientists know its target in brain cells and have observed how it affects brain-wide activity, they haven’t known entirely how the two are connected. A new study by a research team spanning four Boston-area institutions uses computational modeling of previously unappreciated physiological details to fill that gap and offer new insights into how ketamine works.

“This modeling work has helped decipher likely mechanisms through which ketamine produces altered arousal states as well as its therapeutic benefits for treating depression,” says co-senior author Emery N. Brown, the Edward Hood Taplin Professor of Computational Neuroscience and Medical Engineering at The Picower Institute for Learning and Memory at MIT, as well as an anesthesiologist at Massachusetts General Hospital and a professor at Harvard Medical School.

The researchers from MIT, Boston University (BU), MGH, and Harvard University say the predictions of their model, published May 20 in Proceedings of the National Academy of Sciences, could help physicians make better use of the drug.

“When physicians understand what’s mechanistically happening when they administer a drug, they can possibly leverage that mechanism and manipulate it,” says study lead author Elie Adam, a research scientist at MIT who will soon join the Harvard Medical School faculty and launch a lab at MGH. “They gain a sense of how to enhance the good effects of the drug and how to mitigate the bad ones.”

Blocking the door

The core advance of the study involved biophysically modeling what happens when ketamine blocks the “NMDA” receptors in the brain’s cortex — the outer layer where key functions such as sensory processing and cognition take place. Blocking the NMDA receptors modulates the release of excitatory neurotransmitter glutamate.

When the neuronal channels (or doorways) regulated by the NMDA receptors open, they typically close slowly (like a doorway with a hydraulic closer that keeps it from slamming), allowing ions to go in and out of neurons, thereby regulating their electrical properties, Adam says. But, the channels of the receptor can be blocked by a molecule. Blocking by magnesium helps to naturally regulate ion flow. Ketamine, however, is an especially effective blocker.

Blocking slows the voltage build-up across the neuron’s membrane that eventually leads a neuron to “spike,” or send an electrochemical message to other neurons. The NMDA doorway becomes unblocked when the voltage gets high. This interdependence between voltage, spiking, and blocking can equip NMDA receptors with faster activity than its slow closing speed might suggest. The team’s model goes further than ones before by representing how ketamine’s blocking and unblocking affect neural activity.

“Physiological details that are usually ignored can sometimes be central to understanding cognitive phenomena,” says co-corresponding author Nancy Kopell, a professor of mathematics at BU. “The dynamics of NMDA receptors have more impact on network dynamics than has previously been appreciated.”

With their model, the scientists simulated how different doses of ketamine affecting NMDA receptors would alter the activity of a model brain network. The simulated network included key neuron types found in the cortex: one excitatory type and two inhibitory types. It distinguishes between “tonic” interneurons that tamp down network activity and “phasic” interneurons that react more to excitatory neurons.

The team’s simulations successfully recapitulated the real brain waves that have been measured via EEG electrodes on the scalp of a human volunteer who received various ketamine doses and the neural spiking that has been measured in similarly treated animals that had implanted electrode arrays. At low doses, ketamine increased brain wave power in the fast gamma frequency range (30-40 Hz). At the higher doses that cause unconsciousness, those gamma waves became periodically interrupted by “down” states where only very slow frequency delta waves occur. This repeated disruption of the higher frequency waves is what can disrupt communication across the cortex enough to disrupt consciousness.

A very horizontal chart plots brain rhythm frequency over time with colors indicating power. Bars along the top indicate the dose of ketamine. After the dose starts more gamma frequency power appears. After the dose gets even higher, the gamma waves periodically stop and then resume.

A spectrogram of brain rhythm frequencies over time predicted by the team’s model. After a first, moderate dose of ketamine, gamma brain rhythm power (warmer colors) emerges. Then as the dose increases, the gamma rhythms become periodically interrupted, leaving only very low-frequency waves, and then resume.

Image courtesy of Adam, Kopell, McCarthy, et. al.


But how? Key findings

Importantly, through simulations, they explained several key mechanisms in the network that would produce exactly these dynamics.

The first prediction is that ketamine can disinhibit network activity by shutting down certain inhibitory interneurons. The modeling shows that natural blocking and unblocking kinetics of NMDA-receptors can let in a small current when neurons are not spiking. Many neurons in the network that are at the right level of excitation would rely on this current to spontaneously spike. But when ketamine impairs the kinetics of the NMDA receptors, it quenches that current, leaving these neurons suppressed. In the model, while ketamine equally impairs all neurons, it is the tonic inhibitory neurons that get shut down because they happen to be at that level of excitation. This releases other neurons, excitatory or inhibitory, from their inhibition allowing them to spike vigorously and leading to ketamine’s excited brain state. The network’s increased excitation can then enable quick unblocking (and reblocking) of the neurons’ NMDA receptors, causing bursts of spiking.

Another prediction is that these bursts become synchronized into the gamma frequency waves seen with ketamine. How? The team found that the phasic inhibitory interneurons become stimulated by lots of input of the neurotransmitter glutamate from the excitatory neurons and vigorously spike, or fire. When they do, they send an inhibitory signal of the neurotransmitter GABA to the excitatory neurons that squelches the excitatory firing, almost like a kindergarten teacher calming down a whole classroom of excited children. That stop signal, which reaches all the excitatory neurons simultaneously, only lasts so long, ends up synchronizing their activity, producing a coordinated gamma brain wave.

A network schematic shows the model arrangement of three different types of neurons in a cortical circuit.

A schematic of the brain network model. Tonic Inhibitory neurons (blue) use GABA to inhibit the other neuron types. Pyramidal excitatory neurons stimulate each other and the phasic inhibitory neurons (red), which, in turn, inhibit the excitatory neurons.

Image courtesy of Adam, Kopell, McCarthy, et. al.


“The finding that an individual synaptic receptor (NMDA) can produce gamma oscillations and that these gamma oscillations can influence network-level gamma was unexpected,” says co-corresponding author Michelle McCarthy, a research assistant professor of math at BU. “This was found only by using a detailed physiological model of the NMDA receptor. This level of physiological detail revealed a gamma time scale not usually associated with an NMDA receptor.”

So what about the periodic down states that emerge at higher, unconsciousness-inducing ketamine doses? In the simulation, the gamma-frequency activity of the excitatory neurons can’t be sustained for too long by the impaired NMDA-receptor kinetics. The excitatory neurons essentially become exhausted under GABA inhibition from the phasic interneurons. That produces the down state. But then, after they have stopped sending glutamate to the phasic interneurons, those cells stop producing their inhibitory GABA signals. That enables the excitatory neurons to recover, starting a cycle anew.

Antidepressant connection?

The model makes another prediction that might help explain how ketamine exerts its antidepressant effects. It suggests that the increased gamma activity of ketamine could entrain gamma activity among neurons expressing a peptide called VIP. This peptide has been found to have health-promoting effects, such as reducing inflammation, that last much longer than ketamine’s effects on NMDA receptors. The research team proposes that the entrainment of these neurons under ketamine could increase the release of the beneficial peptide, as observed when these cells are stimulated in experiments. This also hints at therapeutic features of ketamine that may go beyond antidepressant effects. The research team acknowledges, however, that this connection is speculative and awaits specific experimental validation.

“The understanding that the subcellular details of the NMDA receptor can lead to increased gamma oscillations was the basis for a new theory about how ketamine may work for treating depression,” Kopell says.

Additional co-authors of the study are Marek Kowalski, Oluwaseun Akeju, and Earl K. Miller.

The work was supported by the JPB Foundation; The Picower Institute for Learning and Memory; The Simons Center for The Social Brain; the National Institutes of Health; George J. Elbaum ’59, SM ’63, PhD ’67; Mimi Jensen; Diane B. Greene SM ’78; Mendel Rosenblum; Bill Swanson; and annual donors to the Anesthesia Initiative Fund.

All in the family

It’s no news that companies use money to influence politics. But it may come as a surprise to learn that many family-owned firms — the most common form of business in the world — do not play by the same rules. New research by political science PhD candidate Sukrit Puri reveals that “family businesses depart from the political strategy of treating campaign donations as short-term investments intended to maximize profitmaking.”

Studying thousands of such firms in India, Puri finds that when it comes to politics, an important influence on political behavior is ethnic identity. This in turn can make a big impact on economic development.

“If family businesses actually think about politics differently, and if they are the most common economic actors in an economy, then you break channels of accountability between a business and the government,” says Puri. “Elected officials may be less likely to deliver effective policies for achieving economic growth.”

Puri believes his insights suggest new approaches for struggling economies in some developing countries. “I’d like to get governments to think carefully about the importance of family firms, and how to incentivize them through the right kinds of industrial policies.”

Pushing past caricatures

At the heart of Puri’s doctoral studies is a question he says has long interested him: “Why are some countries rich and other countries poor?” The son of an Indian diplomat who brought his family from Belgium and Nepal to the Middle East and New York City, Puri focused on the vast inequalities he witnessed as he grew up.

As he studied economics, political science, and policy as an undergraduate at Princeton University, Puri came to believe “that firms play a very important role” in the economic development of societies. But it was not always clear from these disciplines how businesses interacted with governments, and how that affected economic growth.

“There are two canonical ways of thinking about business in politics, and they have become almost like caricatures,” says Puri. One claims government is in the pocket of corporations, or that at the least they wield undue influence. The other asserts that businesses simply do governments’ bidding and are constrained by the needs of the state. “I found these two perspectives to be wanting, because neither side gets entirely what it desires,” he says. “I set out to learn more about how business actually seeks to influence, and when it is successful or not.”

So much political science literature on business and politics is “America-centric,” with publicly listed, often very large corporations acting on behalf of shareholders, notes Puri. But this is not the paradigm for many other countries. The major players in countries like South Korea and India are family firms, big and small. “There has been so little investigation of how these family businesses participate in politics,” Puri says. “I wanted to know if we could come up with a political theory of the family firm, and look into the nature of business and politics in developing economies and democracies where these firms are so central.”

Campaign donation differences

To learn whether family businesses think about politics differently, Puri decided to zero in on one of the most pervasive forms of influence all over the world: campaign donations. “In the U.S., firms treat these donations as short-term investments, backing the incumbent and opportunistically switching parties when political actors change,” he says. “These companies have no ideology.” But family firms in India, Puri’s empirical setting, prove to operate very differently.

Puri compiled a vast dataset of all donations to Indian political parties from 2003 to 2021, identifying 7,000 unique corporate entities donating a cumulative $1 billion to 36 parties participating in national and state-level elections. He identified which of these donations came from family firms by identifying family members sitting on boards of these companies. Puri found evidence that firms with greater family involvement on these boards overwhelmingly donate loyally to a single party of their choice, and “do not participate in politics out of opportunistic, short-term profit maximizing impulse.”

Puri believes there are sociological explanations for this unexpected behavior. Family firms are more than just economic actors, but social actors as well — embedded in community networks that then shape their values, preferences, and strategic choices. In India, communities often form around caste and religious networks. So for instance, some economic policies of the ruling Bharatiya Janata Party (BJP) have hurt its core supporters of small and medium-sized businesses, says Puri. Yet, these businesses have not abandoned their financial support of the BJP. Similarly, Muslim-majority communities and family firms stick with their candidates, even when it is not in their short-term economic best interest. Their behavior is more like that of an individual political donor — more ideological and expressive than strategic.

Engaged by debate

As a college first-year, Puri was uncertain of his academic direction. Then he learned of a debate playing out between two schools of economic thought on how to reduce poverty in India and other developing nations: On one side, Amartya Sen advocated for starting with welfare, and on the other, Jagdish Bhagwati and Arvind Panagariya argued that economic growth came first.

“I wanted to engage with this debate, because it suggested policy actions — what is feasible, what you can actually do in a country,” recalls Puri. “Economics was the tool for understanding these trade-offs.”

After graduation, Puri worked for a few years in investment management, specializing in emerging markets. “In my office, the conversation each day among economists was just basically political,” he says. “We were evaluating a country’s economic prospects through a kind of unsophisticated political analysis, and I decided I wanted to pursue more rigorous training in political economy.”

At MIT, Puri has finally found a way of merging his lifelong interests in economic development with policy-minded research. He believes that the behavior of family firms should be of keen concern to many governments.

“Family firms can be very insular, sticking with old practices and rewarding loyalty to co-ethnic partners,” he says. There are barriers to outside hires who might bring innovations. “These businesses are often just not interested in taking up growth opportunities,” says Puri. “There are millions of family firms but they do not provide the kind of dynamism they should.” 

In the next phase of his dissertation research Puri will survey not just the political behaviors, but the investment and management practices of family firms as well. He believes larger firms more open to outside ideas are expanding at the expense of smaller and mid-size family firms. In India and other nations, governments currently make wasteful subsidies to family firms that cannot rise to the challenge of, say, starting a new microchip fabricating plant. Instead, says Puri, governments must figure out the right kind of incentives to encourage openness and entrepreneurship in businesses that make up its economy, which are instrumental to unlocking broader economic growth.

After MIT, Puri envisions an academic life for himself studying business and politics around the world, but with a focus on India. He would like to write about family firms for a more general audience — following in the footsteps of authors who got him interested in political economy in the first place. “I’ve always believed in making knowledge more accessible; it’s one of the reasons I enjoy teaching,” he says. “It is really rewarding to lecture or write and be able to introduce people to new ideas.” 

Ultrasound offers a new way to perform deep brain stimulation

Deep brain stimulation, by implanted electrodes that deliver electrical pulses to the brain, is often used to treat Parkinson’s disease and other neurological disorders. However, the electrodes used for this treatment can eventually corrode and accumulate scar tissue, requiring them to be removed.

MIT researchers have now developed an alternative approach that uses ultrasound instead of electricity to perform deep brain stimulation, delivered by a fiber about the thickness of a human hair. In a study of mice, they showed that this stimulation can trigger neurons to release dopamine, in a part of the brain that is often targeted in patients with Parkinson’s disease.

“By using ultrasonography, we can create a new way of stimulating neurons to fire in the deep brain,” says Canan Dagdeviren, an associate professor in the MIT Media Lab and the senior author of the new study. “This device is thinner than a hair fiber, so there will be negligible tissue damage, and it is easy for us to navigate this device in the deep brain.”

In addition to offering a potentially safer way to deliver deep brain stimulation, this approach could also become a valuable tool for researchers seeking to learn more about how the brain works.

MIT graduate student Jason Hou and MIT postdoc Md Osman Goni Nayeem are the lead authors of the paper, along with collaborators from MIT’s McGovern Institute for Brain Research, Boston University, and Caltech. The study appears today in Nature Communications.

Deep in the brain

Dagdeviren’s lab has previously developed wearable ultrasound devices that can be used to deliver drugs through the skin or perform diagnostic imaging on various organs. However, ultrasound cannot penetrate deeply into the brain from a device attached to the head or skull.

“If we want to go into the deep brain, then it cannot be just wearable or attachable anymore. It has to be implantable,” Dagdeviren says. “We carefully customize the device so that it will be minimally invasive and avoid major blood vessels in the deep brain.”

Deep brain stimulation with electrical impulses is FDA-approved to treat symptoms of Parkinson’s disease. This approach uses millimeter-thick electrodes to activate dopamine-producing cells in a brain region called the substantia nigra. However, once implanted in the brain, the devices eventually begin to corrode, and scar tissue that builds up surrounding the implant can interfere with the electrical impulses.

The MIT team set out to see if they could overcome some of those drawbacks by replacing electrical stimulation with ultrasound. Most neurons have ion channels that are responsive to mechanical stimulation, such as the vibrations from sound waves, so ultrasound can be used to elicit activity in those cells. However, existing technologies for delivering ultrasound to the brain through the skull can’t reach deep into the brain with high precision because the skull itself can interfere with the ultrasound waves and cause off-target stimulation.

“To precisely modulate neurons, we must go deeper, leading us to design a new kind of ultrasound-based implant that produces localized ultrasound fields,” Nayeem says. To safely reach those deep brain regions, the researchers designed a hair-thin fiber made from a flexible polymer. The tip of the fiber contains a drum-like ultrasound transducer with a vibrating membrane. When this membrane, which encapsulates a thin piezoelectric film, is driven by a small electrical voltage, it generates ultrasonic waves that can be detected by nearby cells.

“It’s tissue-safe, there’s no exposed electrode surface, and it’s very low-power, which bodes well for translation to patient use,” Hou says.

In tests in mice, the researchers showed that this ultrasound device, which they call ImPULS (Implantable Piezoelectric Ultrasound Stimulator), can provoke activity in neurons of the hippocampus. Then, they implanted the fibers into the dopamine-producing substantia nigra and showed that they could stimulate neurons in the dorsal striatum to produce dopamine.

“Brain stimulation has been one of the most effective, yet least understood, methods used to restore health to the brain. ImPULS gives us the ability to stimulate brain cells with exquisite spatial-temporal resolution and in a manner that doesn’t produce the kind of damage or inflammation as other methods. Seeing its effectiveness in areas like the hippocampus opened an entirely new way for us to deliver precise stimulation to targeted circuits in the brain,” says Steve Ramirez, an assistant professor of psychological and brain sciences at Boston University, and a faculty member at B.U.’s Center for Systems Neuroscience, who is also an author of the study.

A customizable device

All of the components of the device are biocompatible, including the piezoelectric layer, which is made of a novel ceramic called potassium sodium niobate, or KNN. The current version of the implant is powered by an external power source, but the researchers envision that future versions could be powered a small implantable battery and electronics unit.

The researchers developed a microfabrication process that enables them to easily alter the length and thickness of the fiber, as well as the frequency of the sound waves produced by the piezoelectric transducer. This could allow the devices to be customized for different brain regions.

“We cannot say that the device will give the same effect on every region in the brain, but we can easily and very confidently say that the technology is scalable, and not only for mice. We can also make it bigger for eventual use in humans,” Dagdeviren says.

The researchers now plan to investigate how ultrasound stimulation might affect different regions of the brain, and if the devices can remain functional when implanted for year-long timescales. They are also interested in the possibility of incorporating a microfluidic channel, which could allow the device to deliver drugs as well as ultrasound.

In addition to holding promise as a potential therapeutic for Parkinson’s or other diseases, this type of ultrasound device could also be a valuable tool to help researchers learn more about the brain, the researchers say.

“Our goal to provide this as a research tool for the neuroscience community, because we believe that we don’t have enough effective tools to understand the brain,” Dagdeviren says. “As device engineers, we are trying to provide new tools so that we can learn more about different regions of the brain.”

The research was funded by the MIT Media Lab Consortium and the Brain and Behavior Foundation Research (BBRF) NARSAD Young Investigator Award.

Helping robots grasp the unpredictable

When robots come across unfamiliar objects, they struggle to account for a simple truth: Appearances aren’t everything. They may attempt to grasp a block, only to find out it’s a literal piece of cake. The misleading appearance of that object could lead the robot to miscalculate physical properties like the object’s weight and center of mass, using the wrong grasp and applying more force than needed.

To see through this illusion, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) researchers designed the Grasping Neural Process, a predictive physics model capable of inferring these hidden traits in real time for more intelligent robotic grasping. Based on limited interaction data, their deep-learning system can assist robots in domains like warehouses and households at a fraction of the computational cost of previous algorithmic and statistical models.

The Grasping Neural Process is trained to infer invisible physical properties from a history of attempted grasps, and uses the inferred properties to guess which grasps would work well in the future. Prior models often only identified robot grasps from visual data alone.

Typically, methods that infer physical properties build on traditional statistical methods that require many known grasps and a great amount of computation time to work well. The Grasping Neural Process enables these machines to execute good grasps more efficiently by using far less interaction data and finishes its computation in less than a tenth of a second, as opposed seconds (or minutes) required by traditional methods.

The researchers note that the Grasping Neural Process thrives in unstructured environments like homes and warehouses, since both house a plethora of unpredictable objects. For example, a robot powered by the MIT model could quickly learn how to handle tightly packed boxes with different food quantities without seeing the inside of the box, and then place them where needed. At a fulfillment center, objects with different physical properties and geometries would be placed in the corresponding box to be shipped out to customers.

Trained on 1,000 unique geometries and 5,000 objects, the Grasping Neural Process achieved stable grasps in simulation for novel 3D objects generated in the ShapeNet repository. Then, the CSAIL-led group tested their model in the physical world via two weighted blocks, where their work outperformed a baseline that only considered object geometries. Limited to 10 experimental grasps beforehand, the robotic arm successfully picked up the boxes on 18 and 19 out of 20 attempts apiece, while the machine only yielded eight and 15 stable grasps when unprepared.

While less theatrical than an actor, robots that complete inference tasks also have a three-part act to follow: training, adaptation, and testing. During the training step, robots practice on a fixed set of objects and learn how to infer physical properties from a history of successful (or unsuccessful) grasps. The new CSAIL model amortizes the inference of the objects’ physics, meaning it trains a neural network to learn to predict the output of an otherwise expensive statistical algorithm. Only a single pass through a neural network with limited interaction data is needed to simulate and predict which grasps work best on different objects.

Then, the robot is introduced to an unfamiliar object during the adaptation phase. During this step, the Grasping Neural Process helps a robot experiment and update its position accordingly, understanding which grips would work best. This tinkering phase prepares the machine for the final step: testing, where the robot formally executes a task on an item with a new understanding of its properties.

“As an engineer, it’s unwise to assume a robot knows all the necessary information it needs to grasp successfully,” says lead author Michael Noseworthy, an MIT PhD student in electrical engineering and computer science (EECS) and CSAIL affiliate. “Without humans labeling the properties of an object, robots have traditionally needed to use a costly inference process.” According to fellow lead author, EECS PhD student, and CSAIL affiliate Seiji Shaw, their Grasping Neural Process could be a streamlined alternative: “Our model helps robots do this much more efficiently, enabling the robot to imagine which grasps will inform the best result.” 

“To get robots out of controlled spaces like the lab or warehouse and into the real world, they must be better at dealing with the unknown and less likely to fail at the slightest variation from their programming. This work is a critical step toward realizing the full transformative potential of robotics,” says Chad Kessens, an autonomous robotics researcher at the U.S. Army’s DEVCOM Army Research Laboratory, which sponsored the work.

While their model can help a robot infer hidden static properties efficiently, the researchers would like to augment the system to adjust grasps in real time for multiple tasks and objects with dynamic traits. They envision their work eventually assisting with several tasks in a long-horizon plan, like picking up a carrot and chopping it. Moreover, their model could adapt to changes in mass distributions in less static objects, like when you fill up an empty bottle.

Joining the researchers on the paper is Nicholas Roy, MIT professor of aeronautics and astronautics and CSAIL member, who is a senior author. The group recently presented this work at the IEEE International Conference on Robotics and Automation.