Supporting systems of success

A new course in the Department of Electrical Science and Engineering (EECS) is not focused on machine learning or microprocessors, on cryptography or quantum computing. Instead, 6.C31 (Hack Yourself) draws student interest with an intriguing pitch: that by using computing tools to implement a personal toolkit of more than 60 positive habits, students can substantially improve their college experiences. 

A joint venture with the Experimental Study Group (ESG), developed by EECS Senior Lecturer Ana Bell, ESG Visiting Instructor Carter Jernigan, and ESG Senior Lecturer Paola Rebusco, Hack Yourself offers students the opportunity to upgrade their lives. 

“We wanted to build an MIT course that exposes students, in a structured way, to what we actually know about learning and thriving,” explain the course’s instructors. “That said, MIT students are skeptical unless they can see evidence. They want data. They want methods. So we built the class around research-backed ideas that students can explore themselves, and that’s where computational thinking comes in. The course treats well-being and learning almost like systems problems: How do you gather meaningful data, break a big question into manageable parts, test an intervention, and evaluate whether it worked?”

The course’s ambitious goals fall into two categories: personal interventions that students can deploy in their own lives, and analytical skills that they can use to test whether their new habits are working. “It’s an introduction to the data science pipeline and evidence-based thinking,” says Bell. “We start from the idea that good analysis begins with good data. Students spend time learning how to collect information thoughtfully: designing surveys, conducting interviews, thinking about bias and measurement. Those are human-centered skills that complement technical training and remain valuable no matter what technologies exist years from now.”

The data that students collect are as diverse as their goals. “A few patterns come up consistently — reducing smartphone use and getting more consistent sleep,” says Jernigan. “For juniors and seniors especially, a lot of interest centers on work habits, productivity, and the transition into the workplace, which naturally pulls them toward teamwork, leadership, and relationships. What’s interesting is that no single intervention emerged as universally ‘best.’” 

Many of the course’s assignments center on teamwork, using research-backed interventions for more-effective group collaboration. “Some are small interventions students can use anywhere — for instance, replacing default small talk with a prompt like, ‘Tell me something good going on in your life,’” explains Jernigan. “We also teach practices like the pre-mortem, where a group imagines a project has already failed, lists everything that could have gone wrong, and works backward to prevent those problems.” 

Even the class’s reading assignments are collaborative, using Perusall (a shared annotation app). “We polled students about which strategies were most personally impactful and which they’d recommend to others, and the answers varied a lot,” says Jernigan. “That’s part of the point. Different tools resonate with different people, so we help students build a broad toolkit.”

Part of that toolkit looks like a tiny photo album. The course’s “Intention cards” are professionally printed, playing-card-sized collectibles that the students accumulate as they move through the course. Each card distills a research-backed habit or strategy into a few words and a colorful AI-generated image as a tangible reminder of the skill. “Students earn a few each week and end up with more than 60 by the end of the semester,” says Jernigan.

As the students gain more cards, they store them in a booklet — but the instructors have noticed them carefully tending and tidying their collections. 

“It’s been interesting to watch the collection dynamic work in ways we didn’t fully design for,” Rebusco reports. “Students reorganize the cards, revisit old ones when they get new ones, and carry the booklets around. Without explicitly thinking about it, they end up engaging in spaced repetition and recall — the same learning principles we discuss in class.”

Junior Sarah Hopp has amassed a large collection of the cards. The course 6-3 Computer Science and Engineering major, from Stevens Point, Wisconsin, has found that the mementos work as both a memory tool and a sentimental souvenir. “With each card, I also tend to remember the class we talked about that intervention, and any positive memories I have from that class experience.” 

Hopp’s favorite class experience was an assignment, originally developed in 2001 by Laura King, in which the student is challenged to envision her “best possible future self” in close detail for 20 minutes, imagining completed life goals and successful outcomes, before writing that imagined future self a letter. “It gave me time to really think about what I want from my future and what I want it to look like,” says Hopp. “It also forced me to focus on the positives and the accomplishments I want to achieve, rather than thinking about the negatives and my fears for the future.”

“When I first signed up for Hack Yourself, I looked forward to having a psychologically insightful and reflective class environment that I figured would be a helpful break from my usual STEM-heavy course content,” says student Amitoj Singh, who is majoring in 6-3. “I wanted to build a habit of using strategies from class to improve my lifestyle in college.” One of the strategies Singh implemented stemmed from a class discussion on flow state (a state of full and joyous task absorption experienced when a person is immersed in challenging and creative work). After learning that even brief interruptions could significantly impact productivity and derail flow, Singh changed how he scheduled his time, creating longer and more coherent “blocks” for his desired goals. “I think one of the most powerful tools [the class uses] is the massive pool of research the lectures and content draw from,” says Singh, who is now working on “satisficing,” or learning when to deem an outcome satisfactory rather than chasing a perhaps-unattainable state of optimization.

The Hack Yourself curriculum is intentionally designed to steer students toward small experiments and goals that feel personal to them. “We’ve had students explore relationships, sports performance, music, motivation, and productivity,” says Jernigan. “They develop hypotheses, think about measurement and bias, design surveys or interviews, and create an analysis plan. They don’t run a large-scale study, but they learn how to think rigorously about their proposed questions and hypotheses, using the same kind of decomposition and iterative thinking we use in computing.” Part of that rigor focuses on the use of large language models — specifically, how to evaluate the quality of both the prompt and the response. “We ask students to engage critically with modern AI tools,” says Bell. “They can use generative AI to explore ideas or analyze data, but they’re required to reflect on the process and critique the quality of what comes back. The durable skill isn’t fluency with today’s AI — it’s knowing how to evaluate any tool that comes next.”

Bell, Jernigan, and Rebusco hope that their students will leave with the tools to ask better questions — a lifelong skill. As they note, “We intentionally avoided making this another machine learning course. Students already get plenty of exposure to advanced technical tools elsewhere at MIT. In this course, we focus on a fundamental element: the data science pipeline itself. How do you collect useful data? How do you ask good questions? How do you design a small experiment before trying to boil the ocean? That mindset scales well, even as technology evolves.”

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

Shape-sensing sheet digitally tracks its movement as it bends and twists

MIT engineers have developed a flexible, shape-sensing sheet that digitally reconstructs its own form as it bends and twists. 

The sensor is similar in principle to some motion-capture garments, which track a person’s physical movements to mimic them in a virtual avatar. Those designs use rigid sensors that are stitched into a suit at strategic locations. 

In contrast, the team’s new design uses multiple soft optical fibers that zig-zag throughout a soft and stretchable sheet. When the sheet bends, the fibers bend in kind, changing the pattern of light that travels through them. The researchers developed an algorithm to interpret the bending light patterns on the fly, to create a digital image that moves the same way the sheet is moving. The team also showed that the design is resistant to damage: Even when a few fibers are cut or disconnected, the remaining fibers can still accurately reconstruct the sheet’s overall shape. 

“Our setup is targeting fully soft and stretchable sensing surfaces so it’s potentially safer for interacting with a human or a soft environment,” says Qifan Yu, a mechanical engineering graduate student at MIT. 

The team envisions that the new design could be fashioned into a soft and pliable garment that a user could comfortably wear to physically control a video game character or a tele-operated robot. The sheet could also be used in physical therapy settings, where it could wrap around a patient’s leg or arm to track and record their mobility and range of motion during each session. 

“In physical therapy, you often need to gather patient data on how they are moving and stretching their arm during rehabilitation, for example,” Yu says. “Surface shape sensing could be very useful in that way.” 

Yu and his colleagues present their design in a study appearing in the journal Advanced Intelligent Systems. The study’s co-authors at MIT are graduate student Nina Cao and assistant professor of mechanical engineering Kaitlyn Becker.

The shape of light

The team’s shape-sensing design is a tweak on the standard optical fiber. Also known as a waveguide, an optical fiber is designed to guide light efficiently through a fiber. Optical fibers consist of a core of transparent material, such as glass, wrapped in a dark, opaque cladding. When light is sent through one end of the fiber, the dark cladding traps the light within the glass core, which itself is extremely smooth, allowing the light to pass straight through the fiber with near-perfect efficiency. Waveguides are used in fiber optic cables for super fast, efficient data transmission and telecommunications. 

Scientists have also experimented with waveguides as simple shape sensors: When the fibers bend, they affect how much light can make it all the way through. The more a fiber bends, the less light comes out the other end. The amount of light that shines out, then, can be a measure of the degree the fiber bends. In this way, optical fibers have been used to sense simple curvatures. 

“People have used waveguides to sense the shape of a line, which they have applied to a robot arm to see how it curves,” Yu says. “But they haven’t been applied to surface shape sensing, to reconstruct the 3D shape of a surface. That’s what we’re trying to do here.”

Fiber tweaks

In their new design, the researchers fabricated their own optical fibers, with some twists. Like conventional waveguides, they made their fibers from a transparent core surrounded by dark cladding. Rather than hard glass, they used a clear, flexible rubber core and a cladding made from the same rubbery material, dyed black. 

And instead of keeping the core completely smooth, they intentionally roughed up one side. If one side of the fiber is rougher than the other, they reasoned, then when the fiber is bent one way, any light passing through would scatter off the rough surface and affect the total amount of light that makes it out the other end. This would be a different amount than would pass through if the fiber were bent at the same angle but toward the fiber’s smooth side. In this way, the half-roughened fiber should act as a bidirectional shape sensor. 

The team fabricated multiple bidirectional optical fibers and looked to embed them into a soft sheet of silicone, arranged so that the fibers would reconstruct the sheet’s shape as it bends and twists. To do so, the researchers carried out simulations of sheets embedded with different optical fiber patterns, from a straightforward checkerboard to criss-crossed, zig-zag arrangements. 

They simulated different ways to bend or twist the sheets, and measured the output of light from each sheet’s configuration of fibers. They converted these light measurements into estimates of how much each fiber must be bending, and combined these to construct an overall 3D shape of the sheet, which they compared to the original simulated sheet shape. From these simulations, they found that a particular spacing of zig-zagging fibers was closest to recreating the sheet’s original shape.

The researchers then fabricated a shape-sensing sheet with the same zig-zag pattern of optical fibers embedded into the sheet. They incorporated an LED at one end of each fiber and a light sensor at the other end, which they connected to an external circuit board to collect and amplify the light measurements. They also developed an algorithm to automatically convert the measurements from fibers into a reconstruction of the sheet’s 3D shape as a whole. 

In experiments, they showed that the algorithm smoothly created a virtual reconstruction of the sheet, almost in real time, as the researchers twisted it into different forms. For instance, when they folded the sheet diagonally, and then again in the opposite direction, the virtual twin mimicked the changing shapes, almost in real time. 

They also placed the sheet in different 3D-printed molds so they could precisely measure the difference between the digitally reconstructed sheet and the physical sheet laying over each mold. In these tests, they found the soft sensing sheet was more accurate than other designs.

“We use a metric that describes the distance between the actual surface and the reconstructed surface, and from that, we found our error was less than 0.4 centimeters,” Yu says. “Existing designs, which are based on rigid sensors, have errors of around 1 to 2 centimeters. So that’s respectable, and at least on par with existing technologies.”

The team will further optimize the sensing sheet, first by thinning it down. Currently, the optical fibers are 1 millimeter thick. Other fabrication processes could shave them down to tens of micrometers, which can be thinner than a strand of hair. Then, the researchers envision embedding many more fibers into a garment, to sense detailed changes in its shape and form, for instance to gauge a patient’s performance in physical therapy. 

“We hope to build tools that augment what a physical therapist can do and help them track and quantify their patients’ progress over time,” says Becker. “These sensing sheets could help a physical therapist to recall and compare more precisely how an evaluation went today versus two months and hundreds of appointments ago.”

This research was supported, in part, by MathWorks.

de MIT News https://ift.tt/af8hr95

Using AI to mitigate the growing environmental threat of data centers

The global building boom of power-hungry data centers is straining electrical grids, causing greater reliance on energy from polluting fossil fuels.

Christina Delimitrou, a newly tenured associate professor at MIT, is fighting this environmental threat by rethinking how the computer servers and networking equipment inside those data centers operate. 

She and her group apply machine learning to make large-scale data centers more efficient, secure, and reliable. They redesign outdated cloud computing systems, develop methods to manage shared hardware resources, and create streamlined server architectures. 

These advances allow data center operators to coax more computational power out of existing hardware.

“If data centers are not utilized to the best of their capabilities, then they will burn much more power than they need to meet growing user demand,” says Delimitrou, the KDD Career Development Associate Professor in Communications and Technology in the Department of Electrical Engineering and Computer Science (EECS) and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL). “There is a lot of bloating, especially on the software side of these systems. If we can remove that bloating in a way that doesn’t compromise performance, then we won’t need to build as many new data centers.”

She also harnesses AI to help programmers find and fix problems in cloud-based applications, like music-streaming services or video conferencing systems. This eliminates application downtime that hampers performance and drains computational resources.

“By managing resources more effectively in the cloud, the end user gets more predictable performance from the application running on their smartphone,” she adds.

Mathematical beginnings

Delimitrou grew up in a midsized town within the vast plains of northern Greece. Her early interest in math and science was sparked, in part, by the ancient history of her homeland, where Euclid and Pythagoras studied mathematical problems more than 2,000 years ago. 

“In Greece, there is a long tradition of geometry,” she says.

She also drew scientific inspiration from her parents. Her mother worked as a chemical engineer and her father as a pharmacist — and both encouraged their daughter’s innate curiosity.

Her early affinity for math led Delimitrou to study computer engineering at the National Technical University of Athens, even though she didn’t know much about the field. She quickly gravitated toward courses that focused on the applied science of engineering.

For her diploma thesis — a project all students complete during their fifth and final year of study — she studied resource management in a computer when multiple applications are running at once. 

“A lot of the challenges I was looking at then would get much harder if, instead of a single system, you had 100,000 of these systems. That was a problem that piqued my interest,” she says.

Seeking to make a bigger impact as a researcher, Delimitrou pursued a graduate degree at Stanford University. She began tackling inefficiencies in cloud computing systems and large-scale data centers, which was a rapidly growing area of research. 

Through that work, Delimitrou and her research mentor, Christos Kozyrakis, the Leonard Bosack and Sandy K. Lerner Professor of Engineering, realized many large computing systems were underutilized.

“You would expect, with all the demand for these systems, that they should be running close to 100 percent capacity. But we found that most were running at only about 15 percent capacity,” she says. “This is not a resource-efficient or sustainable way of scaling these systems.”

Applying AI

To push that utilization closer to 100 percent, she began investigating machine-learning solutions to streamline cumbersome computational processes. Machine learning could automate resource management operations in the cloud, identifying solutions that developers might miss on their own.

“Applying machine learning to solve a large-scale system problem was a novel approach at the time. It was a bit risky because people had not yet shown that these techniques would work,” Delimitrou says. “But empirical approaches require a lot of expertise, and the scale of the system is so large that it is difficult for users to manage. This is why machine learning is often the best solution.”

After earning her PhD, Delimitrou continued this line of work as an assistant professor at Cornell University.

One tool her group developed, Seer, uses deep learning to anticipate and prevent problems in web applications before they happen. This averts widespread slowdowns that may occur if a developer tries to fix a problem manually.   

As she delved deeper into cloud computing, Delimitrou observed that cloud applications were changing. Developers were now splitting applications into smaller pieces to spread across multiple servers, which increases the speed of deployment.

“But the servers were not built for this new style of application design. So, I rethought some of my earlier work to build machine-learning systems for this new class of applications,” she says.

To tackle these new challenges, she found herself collaborating more often with faculty members who had different software and hardware expertise. Those collaborations opened exciting new research areas.

A few years later, she decided to join MIT because of the opportunity to collaborate with researchers at the top of their fields in hardware and software engineering. She became an assistant professor in EECS in 2022.

Creative approaches

At MIT, Delimitrou also enjoys the teaching aspect of her role. One of her favorite courses to teach is 6.191 (Computation Structure), a popular undergraduate class with about 350 students each semester. 

While it’s challenging to keep the course material fresh when the field constantly evolves, she strives to inspire creativity in her students.

“I want the students to learn how to think and learn on their own. Part of that involves shifting away from formulaic assignments and making classes more open-ended. I’d rather give the students something to make them think more deeply,” she says.

In the lab, a creative mindset helps Delimitrou and her team identify novel solutions to problems in cloud computing that others might overlook.

For instance, she extended prior work on debugging problems in cloud applications to encompass not just errors in the code, but also security issues that can make user data vulnerable to hackers.

She also uses AI to redesign software systems so they better fit the capabilities of existing hardware.

“One of the challenges when it comes to applying AI to these systems is that the AI is not interpretable,” she says. “A lot of the work we are doing now involves adding explainability into these AI tools so people can get useful feedback from the system.”

That not only helps developers ensure AI is giving the right answer, but also provides insights into how to design systems better in the future.

She finds that studying these large-scale cloud computing systems is becoming more challenging because tech companies that operate data centers use proprietary hardware, unlike the commodity equipment of early cloud computing days, as well as software systems that can’t be accessed by academic teams. A solution that works in the lab might not work in the real world.

To that end, Delimitrou and her group create clones of proprietary systems and applications. One tool they developed, called Ditto, mimics an application’s structure and performance characteristics, enabling a wide range of studies.

She expects her work to continue shifting as machine-learning models become more advanced, opening new possibilities to boost application performance and hardware efficiency.

“But you still have to use AI carefully. While it can greatly accelerate the application development side, we still need to audit it and be especially careful about how these models are applied so we don’t lose the ability to gain insights out of the solutions AI is giving,” she says.

Outside the lab, Delimitrou enjoys spending time with her husband and 1-year-old daughter.

While she doesn’t have as much time for hobbies these days, she also enjoys gardening and building an electrical toy train track for her daughter, as well as playing classical piano and painting nature scenes. She became interested in painting at Cornell, where she would often paint the many waterfalls near the campus. 

Whether she is working in the garden or painting, Delimitrou says she finds spending time outdoors to be a relaxing escape from the technical nature of her work, but also an important reminder of the role her research plays in sustainability. 

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

Margaret Hamilton, computing pioneer who led software development for the Apollo program, dies at 90

Margaret Hamilton, a profoundly influential computer scientist best known for leading the software engineering team at MIT’s Instrumentation Lab during NASA’s Apollo program, died on Sep. 30. She was 90.

A computing pioneer who authored over 130 publications, Hamilton helped to establish software engineering as a dedicated discipline. She worked at MIT from 1959 until the mid-1970s, after which she became a successful computing entrepreneur and CEO.

Her life’s work was recognized with many awards and honors, including the 2016 Presidential Medal of Freedom from President Barack Obama, whose citation noted: “Hamilton defined new forms of software engineering and helped launch an industry that would forever change human history. Her software architecture led to giant leaps for humankind, writing the code that helped America set foot on the moon.”

“To say Margaret Hamilton was a pioneer — to say she was ahead of her time — would be a dramatic understatement. She was a software engineer at a time when that field was in its infancy, and she not only developed advanced code herself but also led a team in using that nascent technology to develop one of the most complex systems humanity had ever achieved,” says Olivier de Weck, the Apollo Program Professor and interim head of the MIT Department Aeronautics and Astronautics. 

“The Apollo program still stands as one of our greatest testaments to the power of collaboration, ingenuity, and engineering, and Margaret Hamilton was a fundamental contributor to that program’s success. And for her that was just the beginning! She went on to become an entrepreneur and remained on the cutting edge of systems software throughout an extraordinary career, while acting as an advocate, mentor, and inspiration to millions.”

Early life and projects at MIT

Born in Paoli, Indiana, in 1936, Hamilton began studying mathematics at the University of Michigan in 1955 before transferring to Earlham College, where she earned a BA in mathematics with a minor in philosophy in 1958. She moved to Boston, Massachusetts, in 1959 with her husband while he pursued a law degree. 

Hamilton soon found a temporary position in the meteorology department at MIT, working with professor of meteorology Edward N. Lorenz SM ’43, ScD ’48 on weather prediction software. This was her first entry point into computer programming, and her work would go on to inform Lorenz’s future publications on chaos theory.

From there, Hamilton took a role as a programmer at MIT Lincoln Laboratory in 1961, working on the Semi-Automatic Ground Environment (SAGE) project, the United States’ first air defense system. Hamilton wrote software for the prototype AN/FSQ-7 computer (XD-1), used by the U.S. Air Force to search for potentially unfriendly aircraft. During this time, Hamilton began to take an interest in software reliability — a new and largely unexplored concept at the time.

In 1965 Hamilton was preparing to pursue graduate studies when her husband saw an advertisement in the newspaper: The Instrumentation Lab at MIT was seeking people to develop software to “send man to the moon.” The lab had won the contract from NASA to build the onboard flight software for the Apollo program. Intrigued by the challenge, Hamilton applied, and was hired as the first programmer for the Apollo project at MIT, as well as the first female programmer in the project. 

Hamilton worked first on the software for the uncrewed Apollo missions and then was promoted into leading the team developing the on-board flight software for the crewed missions. By 1968 she was assistant director in charge of the Command and Service Module team, and more than 400 people were working on Apollo’s software. 

“From my own perspective, the software experience itself (designing it, developing it, evolving it, watching it perform and learning from it for future systems) was at least as exciting as the events surrounding the mission,” Hamilton told MIT News in 2009. “There was no second chance. We knew that. We took our work seriously, many of us beginning this journey while still in our 20s. Coming up with solutions and new ideas was an adventure. Dedication and commitment were a given. Mutual respect was across the board. Because software was a mystery, a black box, upper management gave us total freedom and trust. We had to find a way, and we did. Looking back, we were the luckiest people in the world; there was no choice but to be pioneers.”

“Defensive” programming and priority-driven software take humans to the moon

Hamilton discovered a talent for leadership, as well as a keen instinct for problem-solving and critical thinking that would prove to save Project Apollo several times over. 

There was the incident that became known as “the Lauren error”: One day, her daughter Lauren, then four years old, was playing with the command module simulator at the Instrumentation Lab when she somehow activated a pre-launch program, called P01, while the simulator was in midflight — which crashed the simulator altogether. Hamilton created a program add-on in the technical documentation warning users not to launch P01 during flight. 

She also proposed a software fix to prevent the error happening during a real mission, but she was overruled on the grounds that the highly trained astronauts were never going to make that same mistake. Yet during the Apollo 8 mission in 1968, that’s exactly what happened: Jim Lovell inadvertently launched P01 during the flight, causing the on-flight navigational data to vanish. Hamilton and her team were called in to solve the error, and after that her proposed changes were integrated into the program. This was an early example of “defensive” programming, the practice of building software that could anticipate or fix errors on its own. 

Hamilton’s most famous contribution to the Apollo program came during the pivotal Apollo 11 mission to land on the moon in July of 1969. Moments before the Eagle module was set to land on the lunar surface, the onboard computer raised the alarm. It had detected a 1202 error: The computer was overloaded due to a fault in a hardware switch, and it was possible the system would not be able to handle the complex landing procedure. 

But Hamilton and her team had engineered priority-driven software, able to shut down unnecessary background tasks in order to prioritize mission-critical tasks. Houston trusted Hamilton’s software and allowed the mission to proceed, and two men walked on the moon for the first time. 

Later career and legacy

Hamilton continued to work at the Instrumentation Lab into the 1970s. As the Apollo program wound down, the Instrumentation Lab spun out of MIT to become the independent Draper Laboratory.

“Margaret left an indelible mark on Draper, and we will be forever grateful,” says Jerry M. Wohletz SM ’97, PhD ’00, president and CEO at Draper. “Through her leadership and contributions to the development of the onboard flight software for NASA’s Apollo Guidance Computer, she helped ensure that Apollo astronauts landed safely on the lunar surface and safely returned home. We will honor her legacy at Draper forever.”

Hamilton went on to create her first software company, Higher Order Software, in 1976. The firm was based on Hamilton’s software engineering approach of error prevention and fault tolerance. 

A decade later, Hamilton founded another software company, Hamilton Technologies, “to provide products and services to modernize the planning, system engineering and software development process in order to maximize reliability, lower cost and accelerate time to market.” 

Hamilton Technologies’ flagship product is the Universal Systems Language (USL), a systems modeling language and methodology for engineering complex software systems that prioritize error prevention and defensive programming. 

Throughout her career, Hamilton worked to achieve recognition for software engineering as a dedicated discipline. 

“I fought to bring the software legitimacy so that it — and those building it — would be given its due respect, and thus I began to use the term ‘software engineering’ to distinguish it from hardware and other kinds of engineering, yet treat each type of engineering as part of the overall systems engineering process,” Hamilton told El Pais in 2018. “When I first started using this phrase, it was considered to be quite amusing. It was an ongoing joke for a long time. They liked to kid me about my radical ideas. Software eventually and necessarily gained the same respect as any other discipline.”

Among her many awards and honors, Hamilton was recognized with the NASA Exceptional Space Act Award for scientific and technical contributions in 2003; the Computer History Museum Fellow Award in 2017; the Intrepid Lifetime Achievement Award in 2019; and induction into the National Aviation Hall of Fame in 2022.

Later in life, Hamilton become an icon for women in science and technology, especially after a now-famous photo, showing her next to a printout of her MIT team’s Apollo code, began circulating online. In 2015, the Apollo software she helped to develop was added in its entirety to the code-sharing site GitHub. And in 2017, she became an official Lego Minifigure after a set originally designed by MIT science communicator Maia Weinstock, honoring her and several other women of NASA history, became available worldwide. 

“Margaret Hamilton has been an inspiration to generations of computer scientists and engineers. Hers was a career dedicated to preventing errors and what she called ‘handling the unknown,’” says de Weck. “She personified leadership by example, and established a practice of software engineering based on problem-solving and systems engineering that we all benefit from.” 

Hamilton is survived by her daughter, Lauren Hamilton; her son-in-law, Richard Selesnick; two grandsons; and four great grandchildren. A memorial service will take place in the spring in Cambridge, Massachusetts.

de MIT News https://ift.tt/ANkHzZ4

Turning the tide on Massachusetts shellfishing data

Until recently, when a Massachusetts shellfish warden needed to know how much rain had fallen overnight before deciding whether to close a shellfishing area, the process might look something like this: call a local resident who happened to own a rain gauge, ask for the measurement, write it down, and use that number to make the call. Heavy rain may cause a closure due to stormwater runoff that can carry pollutants from land to water. If the area closed, the public found out the same low-tech way: a paper notice at the boat ramp, a phone call, or word passed along at the marina.

It’s a system built on trust between neighbors, but it doesn’t scale particularly well. According to the Massachusetts Division of Marine Fisheries’ 2024 Annual Report, Massachusetts shellfisheries yielded an ex-vessel value of over $441 million. This number includes shellfish spread across 738 classified growing areas along the coast that fall under a tangle of overlapping state and municipal jurisdictions. Add in thousands of recreational permit holders, and the old system starts to show its age.

With support from MIT’s Abdul Latif Jameel Water and Food Systems Lab (J-WAFS), MIT Sea Grant researchers have spent the past several years trying to close the gap between the data that exist and the data people can actually use. The effort was proposed to J-WAFS by Michael Triantafyllou, the Henry L. and Grace Doherty Professor in Ocean Science and Engineering and director of MIT Sea Grant, under the title “Cloud-Based Data Applications for Streamlining Natural Resource Management.” 

The original aim of the proposal was to build a web-based system to help decision-making among managers and their stakeholders in order to support productivity of shellfisheries, safe human consumption, and vitality of coastal economies. The team wanted the platform to be scalable enough to serve the entire decision-making pipeline, from the state agency managing hundreds of growing areas down to an individual permit holder deciding whether it’s worth the drive to the water that day. 

The result is a platform called Seashell. The team describes it as sitting at the intersection of big data, natural resource management, and what’s often called the blue economy — the sustainable use of ocean and coastal resources for economic growth — a phrase increasingly common in state and federal policy circles, and one Massachusetts has staked real ground on. Food security is a large part of the equation too. Many of these towns don’t just issue permits for sport; these areas are open so residents have shellfish to eat. When permit holders can trust a platform to tell them, in real time, whether an area is safe to harvest, that’s not just convenience. It’s local food supply staying reliable.

An old bottleneck, a new pipeline

The project’s origins trace back to 2020, when MIT Sea Grant Research Scientist Carolina Bastidas and her colleagues built a shellfish restoration webpage and noticed that propagation data was scattered and often misrepresented. Propagation isn’t a minor detail; it’s how many coastal towns support food security for residents, a fact that came into sharper focus during the Covid-19 pandemic, when recreational shellfishing demand rose sharply enough that some towns limited permit sales just to protect their beds.

That early work led to an informal working group with the Massachusetts Division of Marine Fisheries (DMF) and towns of Falmouth, Mashpee, and Nantucket well before any formal funding existed. “We already had engagement,” says Ben Bray, geospatial applications developer at MIT. “We already had a very real starting point.” 

By the time J-WAFS awarded the team a seed grant in 2022, letters of support from DMF and all three towns were already in hand, each describing the same underlying problem: a lack of an efficient, shared way to track and communicate shellfishing area status.

Bray’s path to this work goes back two decades. He joined MIT Sea Grant in 2005 to help build an internal grant management system, and the job pulled him toward an idea he’d recently encountered in Thomas Friedman’s book “The World Is Flat”: that data tend to sit stranded on disconnected “islands,” reachable only through the right application programming interface (API).

The underlying architecture reflects that same philosophy of connecting, rather than replacing, existing systems. Seashell runs on a PostgreSQL database and Amazon Web Services cloud infrastructure, designed from the outset to be exportable — towns or states that want to run it on their own servers can do so. Rather than asking DMF or municipalities to abandon the tools they already use, the system pulls in data through APIs from a long list of sources: National Oceanic and Atmospheric Administration tide stations, satellite feeds for sea surface temperature and chlorophyll-a, local weather networks, and DMF’s own geographic information system.

Built with towns, not for them

If Seashell has a defining trait, it’s that the towns using it had a hand in shaping nearly every piece of it. In 2023, the research team ran a regional assessment of the Massachusetts shellfishing community that drew responses from more than 312 people, 122 of whom volunteered for follow-up focus groups. In 2025, dedicated focus groups for both the public and administrative interfaces — described by Bastidas as involving anywhere from one to a handful of participants at a time — drove a final round of revisions before launch.

What the team learned sometimes cut against their own assumptions. “How little the end user wants to see and interact with” the data was the biggest surprise, Bray says. Focus group participants didn’t want a comprehensive dashboard of every growing area in the state; they wanted their own location, loaded automatically, with the fewest clicks possible. More simply, users wanted answers to questions such as: Is an area open? Is it high or low tide? What’s the water temperature? That preference reshaped the public interface around geolocation and minimal menu navigation, particularly for mobile use, allowing permit holders to check the system on their way out the door.

The towns themselves turned out to have sharply different needs, shaped largely by how many people they have on staff and who’s showing up at their docks. Many of these departments run on a skeleton crew with one or two staff, managing the entire permitting and enforcement operation. “All the organizations we work with are all understaffed,” Bray says. “I didn’t realize that going into this.”

That imbalance shaped something called the “Ways to Water” feature. Towns popular with seasonal and vacationing residents who buy a permit without knowing the local geography asked for a tool that could route users to the nearest shellfishing access points. They also wanted the tool to show parking availability, boat ramps, and travel time. This detail captures the blue economy dimension of the project well. Both tourism and traditional shellfishing for food security call for differing groups to share the same water and the same system, without necessarily sharing the same local knowledge.

Other requests were smaller, but no less telling. One shellfish officer asked simply whether the interface could be offered in different languages, a request the team is now addressing through built-in translation. It’s the kind of feature that doesn’t show up in a funding proposal’s list of objectives, but that turned out to matter a great deal to the people actually using the tool.

The relationship with DMF followed a different shape than the one with the towns. Where individual towns had fewer resources but could move quickly once they decided to adopt something, DMF moved more deliberately. “They had more ability than they had power,” Bray says of DMF staff, describing an agency with existing institutional constraints that limited the design approach the researchers could take. That distinction ultimately reshaped the project’s strategy: rather than building one system meant to unify DMF and every town under a single workflow, the team shifted toward treating towns and DMF as separate entities with distinct needs, while still syncing DMF’s official closure data into the platform automatically. Bray describes the process as investing effort where towns have the flexibility to adopt something new, rather than waiting on slower-moving state processes to catch up.

Rules, real-time data, and respecting jurisdiction

The system’s handling of jurisdiction is arguably its core technical achievement. DMF manages the baseline layer of the state’s 738 growing areas; individual towns manage sub-areas on top of that layer, closing sections for reasons ranging from water quality to family-designated harvesting zones reserved for local residents. Early on, the team assumed DMF would want to manage its polygons directly inside Seashell. DMF preferred to keep working in an existing ArcGIS system. So the MIT team adjusted, syncing Seashell to DMF’s existing records via API instead, and building in automatic alerts whenever DMF’s underlying data changed. It’s a small design decision with an outsized effect: DMF keeps full control of its own data while towns still benefit from updated, synced data.

The administrative interface goes a step further, pulling in real-time environmental data and using it to recommend closure decisions based on rules that towns configure themselves. It’s a direct stand-in for the old rain-gauge phone call, and it’s built to shrink the gap between when DMF issues a closure notice and when the public actually learns about it.

A rare bipartisan corner of ocean policy

Seashell officially launched at the 2026 Northeast Aquaculture Conference and Exposition in Portland, Maine, this past January. The team is scheduled to present at the Massachusetts Shellfish Officers Association’s meeting this October, betting that broader town-level adoption will reduce the confusion that crops up when neighboring towns operate on entirely different systems. Some towns’ shellfish officers are using Seashell while the next town over might still be working on paper.

Bray sees something durable in the subject matter itself, apart from the technology. “Shellfishing is one of the few things, in terms of the ocean, in terms of cultural communities, that is bipartisan,” he says. Bastidas points to the layered value of the resource, not just its economic footprint, but its role in food security and in traditions, including Indigenous shellfishing practices, that predate Massachusetts’ current regulatory system by centuries.

Looking ahead, the team is exploring an extension into commercial fisheries and landing-data reporting, building on related, town-driven effort. More broadly, the researchers see little about the underlying framework that’s specific to shellfishing. The same architecture, they argue, could serve any natural resource management problem with a geospatial and jurisdictional dimension: wildlife management, water quality, coastal permitting.

For Bray, who led his first funded research project through this J-WAFS grant, the work has doubled as a lesson in institutional collaboration as much as software design. The project helped him understand how fast and how much a large state agency can move versus a small town, and where to put the effort to balance both sides. “You have to figure out what an organization is capable of providing, and work within that,” he says, “while still trying to push things forward.”

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

Planning system ensures a robot’s flight path will remain collision-free

Uncrewed aerial vehicles (UAVs) could fly deep into the heart of a raging wildfire, avoiding sudden flare-ups and dodging falling tree branches to gather critical information for rescuers. But in this dangerous situation, the UAV could easily be damaged or destroyed by falling debris.

The UAV — and the rescuers who rely on it — would benefit from a new, autonomous navigation system that is guaranteed to avoid collisions, even in completely unfamiliar surroundings. 

Developed by MIT researchers, this method plans a flight path for a UAV that eludes unknown obstacles that may move in unpredictable ways. It charts an efficient course through an unmapped environment that is mathematically proven to be safe from collisions.

Many popular navigation systems can only offer formal safety guarantees when the environment is static, or when the obstacles are known in advance.

By enabling the UAV to avoid moving obstacles while it is discovering its environment, this trajectory planner, which the researchers call “SANDO” (for “Safe AutoNomous trajectory planning for Dynamic unknOwn environments”) could be especially useful for applications like search and rescue missions into collapsed buildings, mine explorations through networks of hidden tunnels, or package delivery across a crowded neighborhood.

“In the hardest possible environment, where the UAV has no map of the area and there are unknown obstacles moving around, we established a mathematical guarantee of safety. The only thing the planner needs to know is the top speed the obstacles could reach. Given that, you could use it in any environment, without a map, and you know the UAV is not going to crash into anything,” says Kota Kondo SM ’23, PhD ’26, who recently earned his doctorate in aeronautics and astronautics at MIT and is lead author of a paper on this new system.

He is joined on the paper by Jesús Tordesillas PhD ’22, an assistant professor at Comillas Pontifical University in Madrid; MIT graduate students Juan Rached, Lili Sun, and Yixuan Jia; and senior author Jonathan P. How, a Ford Professor of Aeronautics and Astronautics and a principal investigator in the Laboratory for Information and Decision Systems (LIDS) and the Aerospace Controls Laboratory (ACL) at MIT. The research appears in the IEEE Transactions on Robotics.

Safety first

Trajectory planners use images and data from a UAV’s onboard cameras and sensors to chart a flight path that reaches the vehicle’s goal. 

Most existing planners are either designed for unknown static environments, where the obstacles don’t move, or they loosely avoid dynamic obstacles without providing a formal guarantee that the robot won’t crash. 

Formal safety guarantees are important in high-stakes situations, such as if a UAV were delivering medical supplies to the site of a remote natural disaster. But trying to compute every possible crash in a dynamic environment would take too long for real-world deployments.

“In an unknown dynamic environment, you don’t have many assumptions to rely on. In those types of environments, researchers haven’t yet been able to mathematically guarantee that a trajectory is safe,” Kondo explains.

The MIT researchers used a rigorous mathematical approach to develop SANDO, their safe trajectory planner. They theoretically proved the algorithm always computes trajectories which are guaranteed to avoid collisions with moving obstacles in unknown environments.

SANDO starts by mapping out a safety corridor through the robot’s environment. This corridor is a series of connected regions of 3D space the robot can travel through, which are guaranteed not to contain any obstacles. 

But unlike other systems, SANDO creates a time-sensitive safety corridor that considers the possible future movements of dynamic obstacles. It employs a special module that detects, groups, and monitors dynamic obstacles to estimate where they will move next.

While the system doesn’t know exactly where an obstacle will move in the future, it uses that obstacle’s maximum velocity to compute how far it could possibly go in a certain timespan. It puts a sphere around the obstacle that captures the farthest distance it could travel in all directions.

SANDO builds the safety corridor around these spheres to ensure the UAV will not collide with a moving object.

“In the real world, obstacles are going to move, so the safety corridor you create at one point won’t be useful as things move into the corridor. But because we consider this time component, we can now guarantee safety into the future,” Kondo says.

The system uses a heat-map based planner to identify “hot” regions of the environment with many obstacles and guides the UAV away from these dangerous areas. This helps the robot chart a more efficient course around danger zones.

Fast reactions

Once it has established a collision-free safety corridor, SANDO optimizes the trajectory within that corridor to find the fastest path to reach the goal. 

As the robot travels, SANDO adjusts the safety corridor and reformulates the trajectory to ensure the robot’s path remains collision-free until it reaches its goal. 

The researchers employed a few tricks to make the optimization easier to solve so the UAV can rapidly recalculate trajectories using its onboard computer, quickly reacting to sudden changes.

“The most difficult part of developing SANDO was the math,” Kondo says. “When you try to guarantee safety, you need to be rigorous and ensure your theory covers every possible case, even edge cases. Once we had that mathematical guarantee, it was very easy to fly the UAVs.”

In simulations, SANDO reached the robot’s goal faster than several state-of-the-art systems while completely avoiding collisions in all environments. 

SANDO also avoided all dynamic obstacles in 12 test flights with a real UAV, using the robot’s onboard computer and sensors to rapidly replan safe trajectories. 

In the future, researchers could make SANDO more computationally efficient and combine the system with machine-learning models that allow the user to give instructions to a robot using plain language.

A central challenge in autonomous flight is that a path that is safe when it is planned may become unsafe as the environment changes. SANDO addresses this challenge with time-varying safe flight corridors and hard-constrained trajectory optimization that supports frequent onboard replanning. Its combination of spatiotemporal planning, formal safety analysis, and hardware validation provides a practical approach to autonomous flight in complex dynamic environments,” says Fei Gao, an associate professor at Zhejiang University in China, who was not involved with this research.

This research is funded, in part, by the Defense Science and Technology Agency of Singapore.

de MIT News https://ift.tt/Lu31ZPX

Supercomputing researchers document evolution of AI hardware

As artificial intelligence transforms industries and national security, understanding the latest hardware capabilities is important for maintainind technological advantage.

AI accelerators — specialized systems designed to speed up capabilities such as neural networks, deep learning, and machine learning — have been a major area of development for nearly a decade. Since 2018, team from the Lincoln Laboratory Supercomputing Center (LLSC) has been conducting the Lincoln AI Computing Survey (LAICS, pronounced "lace"). Six papers later, LAICS continues to summarize current commercial AI accelerators and compare their peak performance and peak power.

"About eight years ago, we saw a sharp rise in the number of research AI accelerators described in research papers and commercial accelerators being announced, and we started to get questions about them from government sponsors of the laboratory's work. That was motivation enough to start the survey," says Albert Reuther, a staff member at the LLSC, which operates and optimizes the high-performance computing systems used by thousands of laboratory research staff.

Although AI accelerators are frequently used for processes such as machine learning, they also can enable other parallel applications, such as modeling the functions of molecules and speeding up simulations of fluid dynamics — processes that are very computationally expensive.

AI accelerator technology can come in a number of forms: central processing units (CPUs), graphics processing units (GPUs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and dataflow accelerators. Each type of accelerator has slightly different capabilities. CPUs can be used for general-purpose computing, while ASICs can perform only very specific tasks. Dataflow accelerators, FPGAs, and GPUs are more flexible and can be configured for a variety of workloads. Efficiency and performance vary across the different types of accelerators depending on how they are designed. The goal of LAICS is to survey the technologies currently on the market and compare them to find the best accelerators for certain needs.

Led by Reuther, the LAICS team includes LLSC members Michael Jones, Peter Michaleas, Jeremy Kepner, and Vijay Gadepally. The team also collaborates with researchers across Lincoln Laboratory, including in the Advanced Technology Division and Intelligence, Surveillance, and Reconnaissance and Tactical Systems Division, to learn how accelerators support research and development for their missions.

The first paper in their series studied 57 accelerators, while the latest one looked at more than 120 accelerators. The main metrics the team uses to compare accelerators are the peak performance and power; then they sort accelerators by whether they're on a chip, card, or system. All the data in the papers are drawn from public sources, which can be challenging because some companies prefer to keep their performance and power data private. To keep up to date on the latest in the field, Reuther runs daily news and citation searches that report new technical press articles, company announcements, and industry presentations.

"It continues to surprise me how each year another five to 10 startups get funded and announced, and then release new AI accelerators," Reuther says. "One might think that the landscape is saturated enough, but then another batch of innovative accelerators is introduced."

In addition to summarizing the performance versus peak power of the current accelerators, each paper explores a new aspect of the field. For example, the paper published in 2022 investigated sources of performance increases, finding that they stem from smaller, denser transistor designs and the use of lower numerical precision (i.e., calculating fewer significant digits). The latest paper examined different architectural choices available, analyzing how the addition of certain components, such as more cores per processor or parallel performance, would change the system.

Reuther plans to continue the survey for the foreseeable future, stating that, in just the past few months, six new startups have announced their first AI accelerators.

"AI and the hardware it runs on are such hot topics, and it is important for Lincoln Laboratory to be an unbiased technical advisor for choosing and pursuing the right technologies," Reuther says. "Our AI accelerator surveys have helped many sponsors and government colleagues gain a better understanding of the AI accelerator landscape and make better research and acquisition decisions about them. This survey has also been very valuable to determine which GPUs we should consider for upcoming LLSC system purchases so it not only benefits our sponsors and mission programs, but also benefits all LLSC users."

The full set of papers and the accompanying datasets can be found here.

de MIT News https://ift.tt/RvFEA56