Summer 2026 recommended reading from MIT

Summer is the perfect time to curl up with a good book — and MIT authors have had much to offer in the past year. The following titles represent a selection of books published in the past 12 months by MIT faculty and staff.

Looking for more literary works from the MIT community? Enjoy our book lists from 2025,  2024, 2023, 2022, and 2021.

Happy reading!

Fiction and poetry

“We (the People of the United States)” (Penguin Books, 2026)
By Joshua Bennett, the Distinguished Chair of the Humanities at MIT and professor of literature

Bennett marks the 250th anniversary of the founding of the U.S. with a book-length work of poetry about the country and some of its distinctive figures. The piece features remarkable people or inventions from each of the 50 states, meditating on their place in the nation’s cultural fabric.

“The Race for Daphne” (Finishing Line Press, 2026)
By Sarah C. Beckmann, communications and marketing associate in the MIT Media Lab

A poetry collection structured as a crew race exploring girlhood, womanhood, and motherhood through the experiences of a rower and writer. These poems subvert the historical dominance of male heroes by celebrating ordinary female heroism, while examining love, home, and what it means to be an American woman today.

“Jezelle: Thief of Forks” (Self-published, 2026)
By Scott Austin Tirrell, director of administration and finance in the Art, Technology, and Culture Program

Abandoned by her father and raised by the streets of Grafton Notch, Jezelle survives by trusting no one. When a strange magic awakens within her, it offers more than escape — it offers power. But in a city that preys on broken children, power makes her valuable, dangerous, and hunted. To claim the life stolen from her, Jezelle must decide what she is willing to become.

Science and Engineering

“Phenomenal Moments: Revealing the Hidden Science Around Us” (Candlewick Press, 2025)
By Felice Frankel, research scientist in the Department of Chemical Engineering

Enlisting readers to “be the scientist” through vivid fine-art photographs, science photographer Felice Frankel zooms in and out on beautiful and brilliant moments all around us to reveal the chemical, natural, or physical processes — from viscosity and venation to chlorophyll and capillary action — behind scientific phenomena.

“Syntax: A Cognitive Approach” (MIT Press, 2025)
By Edward A. F. Gibson, professor of brain and cognitive sciences

This book lays out the grammar of a language from the perspective of a cognitive scientist, outlining the components of language structure and the model of syntax that Gibson advocates: dependency grammar, in which a word is connected to another word via a dependency arc to form a larger compositional meaning. This formalism can explain numerous aspects of word order universals across languages.

“Birds Up Close: An Engineer Explores Their Hidden Wonders” (MIT Press, 2026)
By Lorna J. Gibson, professor post-tenure in the Department of Materials Science and Engineering 

A renowned engineer and lifelong birder, Gibson explores the hidden microscopic structures and engineering principles that keep birds aloft and alive — how an egg forms, how a bird generates lift, how woodpeckers safely drill their holes, and much more. She also considers the longer view of birds in their habitats and natural history. Her up-close look at avian mysteries provides a perspective like no other, for the expert ornithologist and curious observer alike.

“Carbon Renewal” (MIT Press, 2025)
By Howard J. Herzog, senior research engineer at the MIT Energy Initiative, and Niall Mac Dowell

In “Carbon Renewal,” Herzog and MacDowell discuss how technology and policy can come together to help us reach “net-zero” climate targets. The authors explore the rapidly evolving world of carbon dioxide removal (CDR), presenting the technological pathways of enhancing the land sink, biomass-based carbon capture and storage, engineered removal methods, and ocean-based carbon removal. They also discuss barriers facing CDR as well as ethical implications of this process. 

“Climate Change, Drinking Water Security, and Public Health: Global Challenges and Solutions” (Springer Nature, 2026)
Chapters by Libby Hsu, associate director of academics at MIT D-Lab

In her chapter, “Drinking Water Status Around the World and Its Effect on Health,” Hsu discusses the Earth’s water resources, which are found in a variety of settings. In her chapter, “Waterless and Low-Water Sanitation Technologies that Improve Quality of Life and Conserve Water Resources,” she shares her experience with sanitation challenges in the Global South and how that has reinforced the value of waterless and low-water sanitation technologies that are suitable for scaling around the world.

“A Pox on Fools: The True Believers, Grifters, and Cynics Who Convinced Us to Reject Vaccines” (Penguin Random House, 2026)
By Thomas Levenson, professor of science writing in MIT Comparative Media Studies/Writing

In his latest book, Levenson searches for the origins of the most common arguments against vaccines: that they are unnatural; that they are more dangerous than the illnesses they claim to prevent; and that they are an affront to freedom. “A Pox on Fools” explores the human impulse to question and wonder — sometimes past the point at which the very act of questioning turns deadly.

“The Shape of Wonder: How Scientists Think, Work, and Live” (Penguin Random House, 2025)
By Alan Lightman, professor of the practice of the humanities in MIT Comparative Media Studies/Writing, and Martin Rees

Lightman and Rees pull back the curtain on the field of science, revealing that scientists are driven by the same sense of curiosity, wonder, and responsibility toward a future that shapes us all. They guide us through the fascinating lives and minds of scientists around the world and throughout time, and provide an inside peek at what makes scientists tick — their daily lives, passions, and concerns about the societies they live in.

“Uncertainty in Climate Change Research: An Integrated Approach” (Springer Nature, 2025)
Chapter by Jennifer Morris, principal research scientist at the MIT Center for Sustainability Science and Strategy and the MIT Energy Initiative, and John Reilly, senior lecturer in the MIT Sloan School of Management

Understanding future emissions scenarios is essential for preparing for climate change. The chapter “Emissions and Concentration Scenarios” examines how socioeconomic uncertainty contributes to overall climate change projections, and identifies key drivers of greenhouse gas emissions. It reviews the history of emissions scenarios and compares various approaches, including IPCC methods and formal uncertainty analysis techniques. The chapter concludes with lessons learned from over 40 years of socioeconomic scenario development for climate research.

“The Headache: The Science of a Most Confounding Affliction — and a Search for Relief” (Harper Collins, 2025)
By Tom Zeller Jr., managing editor of Undark, published by the Knight Science Journalism Program at MIT

From blinding migraines to severe headache disorders known as “clusters,” chronic head pain affects 40 percent of the population, many of them suffering in silence. Finally, “The Headache” reveals the science behind a group of disorders that is as much a curse as a cultural punchline, and leads to key insights into the nature of pain itself. Guided by his own decades-long struggle with cluster headaches, Zeller’s journey into headache science is at once intimate and panoramic.

Culture, humanities, and social sciences

“The People Can Fly: American Promise, Black Prodigies, and the Greatest Miracle of All Time” (Little, Brown, and Company, 2026)
By Joshua Bennett, the Distinguished Chair of the Humanities at MIT and professor of literature

In this work, Bennett offers a series of profiles, carefully wrought to see how some prominent figures were able to flourish from childhood forward. He closely reads their works for indications about how they understood the shape of their own lives. In so doing, Bennett underscores the significance of the social settings that prodigious talents grow up in. He also offers reflections on his own career trajectory and encounters with these artists, driving home their influence and meaning.

“Thinking Historically: A Guide to Statecraft and Strategy” (Yale University Press, 2025)
By Francis J. Gavin, research affiliate of the MIT Security Studies Program 

It seems obvious that we should use history to improve policy. If we have a good understanding of the past, it should enable better decisions in the present, especially in the highly consequential worlds of statecraft and strategy. But how do we gain that knowledge? How should history be used? In this book, Gavin explains the many ways historical knowledge can help us understand and navigate the complex, often confusing world around us. 

“The Economic Consequences of the Second Trump Administration: A Preliminary Assessment” (Centre for Economic Policy Research, 2025)
Edited by Gary Gensler, professor of the practice of global economics and management and finance in the MIT Sloan School of Management; Simon Johnson, the Ronald A. Kurtz (1954) Professor of Entrepreneurship and professor of global economics and management at MIT Sloan; Ugo Panizza; and Beatrice Weder di Mauro

How might the economic and geopolitical positions of the Trump administration affect growth, trade, investment, inflation, stability, and the role of the U.S. dollar? This volume offers evidence-based, expert analysis to help decision makers understand the impact of tariffs, breaks in global alliances, government downsizing, deregulation, threats to the rule of law, and more.

“The Colony and the Company: Haiti after the Mississippi Bubble” (Princeton University Press, 2025)
By Malick W. Ghachem, professor of history

Many things account for Haiti’s modern troubles. A good perspective on them comes from going back in time to 1715 or so — and grappling with a far-flung narrative involving the French monarchy, a financial speculator named John Law, and a stock-market crash called the “Mississippi Bubble.” In "The Colony and the Company," Ghachem examines the economic transformations and multi-sided power struggles of that time.

“Retrench, Defend, Compete: Securing America’s Future Against a Rising China” (Cornell University Press, 2025)
By Charles L. Glaser, senior fellow in the MIT Security Studies Program 

Many believe China’s ascent will drive it to war with the United States. Yet this is far from inevitable; geography and nuclear weapons should ensure U.S. security. The real danger, Glaser contends, lies in East Asia’s territorial disputes, especially over Taiwan. To reduce the risk of war, Glaser makes a bold case for ending U.S. security commitments to Taiwan and carefully calibrating its policies on protecting South China Sea maritime features. 

“Trade in War: Economic Cooperation Across Enemy Lines” (Cornell University Press, 2025)
By Mariya Grinberg, associate professor of political science and MIT Security Studies Program affiliate

“Trade in War” is an urgent, insightful study of a puzzling wartime phenomenon: states doing business with their enemies. To explain why states trade with their enemies, Grinberg examines the wartime commercial policies of major powers during the Crimean War, the two World Wars, and several post-1989 wars.

“Constructing Economic Nationalisms in Brazil and India” (Cambridge University Press, 2026)
By Jason Jackson, associate professor in political economy and urban planning in the Department of Urban Studies and Planning

Conventional approaches cite India’s leftist “socialism” and Brazil’s right-wing authoritarianism to explain why India resisted foreign direct investment (FDI) while Brazil welcomed foreign firms. However, this ignores puzzling industry-level variation: India restricted FDI in auto manufacturing but allowed multinationals in oil, while Brazil welcomed foreign auto companies but prohibited FDI in oil. This book argues that FDI policies were shaped by contrasting colonial experiences that generated distinct economic nationalisms and patterns of industrialization in both countries. 

“Traders, Speculators, and Captains of Industry: How Capitalist Legitimacy Shaped Foreign Investment Policy in India” (Harvard University Press, 2025)
By Jason Jackson, associate professor in political economy and urban planning in the Department of Urban Studies and Planning

Is foreign capital an agent of economic growth in developing countries or a vehicle of extraction? Examining how Indian elites wrestled with this question in the late colonial and postcolonial periods, Jackson argues that it reflects a false binary. Instead of simply choosing between domestic and foreign capital, Indian policymakers have long considered the business ethics of individual firms. Indian economic nationalism, in other words, has never been characterized by a straightforward preference for domestic over foreign capital.

“The Handbook of Social Protection: Evidence and New Directions for Low- and Middle-Income Countries” (MIT Press, 2026)
Edited by Benjamin A. Olken, the TEPCO Professor of Economics in the Department of Economics, and Rema Hanna

Over the past several decades, social protection programs that provide financial assistance to the poor and insure against shocks for the vulnerable have become widespread in low- and middle-income countries. These programs can play a critical role in society. This book provides an overview of what we know about the differing aspects of social protection and highlights the open questions for research for the future. 

“Argumentation: The Key Concepts” (Routledge, 2026)
By Edward Schiappa, the John E. Burchard Professor of Humanities in MIT Comparative Media Studies/Writing

In this book, Schiappa delves into the identification and analysis of fallacies, the evaluation of evidence, and the crucial roles of context, audience adaptation, and argumentative style. It explores the ethical dimensions of argument, the impact of cognitive bias, and the influence of cultural and discourse communities.

“American Independence in verse” (Pentameter Press, 2025)
By Brad Skow, the Laurence S. Rockefeller Professor in the Department of Linguistics and Philosophy

“American Independence in verse,” published by Pentameter Press, traces a story of America’s origins through a collection of vignettes featuring some well-known characters, like politician and orator Patrick Henry, alongside some lesser-known but no less important ones, like royalist and former chief justice of North Carolina Martin Howard. Each is rendered in blank verse, a nursery-style rhyme, or free verse.

“Rwanda’s Genocide Heritage: Between Justice and Sovereignty” (Duke University Press, 2025)
By Delia Wendel, associate professor of urban studies and international development in the Department of Urban Studies and Planning

Drawing from oral histories and a visual archive of memory work after the 1994 genocide in Rwanda, Wendel explores the human rights and government priorities that preserved killing sites and victims’ remains for public display. Rwanda’s genocide memorials exemplify a global phenomenon that Wendel terms “trauma heritage,” wherein hidden or unrecognized violence is made visible in public space to demand justice and recognition. Wendel argues that trauma heritage innovates on the form histories take by “writing” them into landscapes, constituting a reparative historiography from the Global South. 

Technology and society 

“Computing in the Age of Decolonization: India’s Lost Technological Revolution” (Princeton University Press, 2026)
By Dwaipayan Banerjee, associate professor of science, technology, and society

In this book, Banerjee examines India’s pursuit of technological self-sufficiency, and the global forces that prevailed against this vision. He describes why the nation is “the world’s leading provider of inexpensive outsourcing and offshoring services, yet enjoys minimal benefits from more profitable advances in research, manufacturing, and development.”

“Auditing AI” (MIT Press, 2026)
By Karrie G. Karahalios, professor of media arts and sciences at the MIT Media Lab; Marc Aidinoff PhD ’22; Nathan Matias SM ’13, PhD ’17; Christian Sandvig; Alondra Nelson; Kristen Vaccaro; Esha Bhandari; Ellery Roberts Biddle; Lena Armstrong; Motahhare Eslami; and Danaé Metaxa

This book serves as a first-of-its-kind roadmap for auditing artificial intelligence systems to prevent decision-making failures in health care, policing, and employment. Using canonical examples of AI gone wrong — from misidentified facial recognition to biased hiring algorithms — this book explains why robust audits are essential and how they drive concrete policy and corporate change.

“Shape Computation: Fifty Years, 1972-2022” (Springer Nature, 2025)
Edited by Sotirios Kotsopoulos SM ’00, PhD ’05, a research affiliate in the Department of Architecture, with a chapter by Terry W. Knight, the William and Emma Rogers Professor of Design and Computation in the Department of Architecture

This book provides a panorama of “shape computation” and “shape grammars,” a computational theory that has, from its inception 50 years ago, been directed toward the “how” of design. Knight’s chapter, “How is that? Computing the Temporality of Drawing,” describes how process and time are key to studying, appreciating, designing, and making things. She notes that in creative production it is not only important to ask, “What is that?” but also “How is that?” — in other words, how did or how can a thing come to be? As a process carried out over time, computation offers a means for rethinking, representing, and elevating the “how” in designing and making activities. 

“The Remote Revolution: Drones and Modern Statecraft” (Cornell University Press, 2025)
By Erik Lin-Greenberg, associate professor in the Department of Political Science

In “The Remote Revolution,” Erik Lin-Greenberg shows that drones are rewriting the rules of international security — but not in ways one would expect. Leveraging diverse types of evidence from original wargames, survey experiments, and cases of U.S. and Israeli drone operations, Lin-Greenberg explores how drone operations lower risks of escalation. 

“The Comedy of Computation: Or, How I Learned to Stop Worrying and Love Obsolescence” (Stanford University Press, 2025)
By Benjamin Mangrum, associate professor of literature

We often deal with our doubts and fears about computing through humor, whether reconciling ourselves to machines or critiquing them. In fact, this dynamic turns up throughout modern culture, in movies, television, fiction, and the theater. Mangrum analyzes this phenomenon in “The Comedy of Computation,” digging into several facets of modern culture and technology.

“Rubrique Technologie / Tech Section” (Printed Matter, 2026)
By Nick Montfort, professor of digital media in MIT Comparative Media Studies/Writing, and Patsy Baudoin

This work is based on a text generator that produces French and English news items that imagine some of the ways technology will impact us in the near future. Most of the generated news involves people getting struck by autonomous vehicles or even aircraft. Others describe labor disputes, hostile takeover attempts, inventions, and the termination of online services. What is imagined in “RT/TS” is not apocalyptic or discontinuous but actually features many of the same problems we face today; the methods of producing the texts are today’s as well.

“Shared Wisdom: Cultural Evolution in the Age of AI” (MIT Press, 2025)
By Alex “Sandy” Pentland, the Toshiba Professor of Media Arts and Sciences and professor of information technology in the MIT Media Lab

How can we build a flourishing society by using human nature to design technology rather than letting technology shape society? Pentland explores how cultural inventions — from civilizations to the Enlightenment — accelerated innovation and collective wisdom. He argues that understanding these key factors in cultural evolution is essential for solving global challenges like climate change and pandemics, and shows how AI and digital media can aid rather than replace human deliberation.

“Priority Technologies: Ensuring US Security and Shared Prosperity” (MIT Press, 2026)
Edited by Elisabeth B. Reynolds, professor of the practice of urban studies and planning, with a foreword by Simon Johnson, the Ronald A. Kurtz (1954) Professor of Entrepreneurship and professor of global economics and management

A new world order is emerging, and within it, U.S. priorities are shifting. For the country to flourish as well as defend and secure its interests, it must build on its decades of experience in developing frontier technologies and globally competitive industries through investments into priority technologies for the 21st century. This volume presents an introduction to some of the key areas where the U.S. must lead in order to ensure both national and economic security: critical minerals, semiconductors, biomanufacturing, quantum computing, drones, and advanced manufacturing.

Education, work, finance, and social impact

“The Meritocracy Paradox: Where Talent Management Strategies Go Wrong and How to Fix Them” (Columbia University Press, 2025)
By Emilio J. Castilla, the NTU Professor of Management and professor of work and organization studies in the MIT Sloan School of Management

Organizations often hail meritocracy as a fair and efficient way to identify, advance, and reward talent. But efforts to create a level playing field can be held back by talent management systems that confer rewards based on individual performance evaluations. In practice, these merit-based systems “may actually reinforce or create advantages for certain groups,” Castilla contends.

“The Art of Monetary Policy: Lessons from Sun Tzu for Central Banks” (MIT Press, 2026)
By Kristin J. Forbes, the Jerome and Dorothy Lemelson Professor of Management and professor of global economics and management in the MIT Sloan School of Management

Central banks are navigating a world of higher debt, tightly interconnected markets, and rising geopolitical tensions. How might they respond effectively? In “The Art of Monetary Policy,” Forbes draws on the writings of Chinese military strategist Sun Tzu to suggest modern principles for central banks, including preparing for the next financial battle, establishing a strong tactical position, combining weapons and methods, and modifying and varying tactics to maintain flexibility.

“Launching from the Lab: Building a Deep-Tech Startup” (MIT Press, 2026)
By Lita Nelsen, former director of the MIT Technology Licensing Office, and Maureen Stancik Boyce, mentor for the MIT Sandbox program

“Launching from the Lab” provides a much-needed framework for new entrepreneurs who are founding companies based on “deep technology” — groundbreaking innovations rising from new discoveries in fundamental research. Nelsen and Stancik Boyce cover the steps to launch and fund such companies, beginning with emergence from the laboratory and acquiring intellectual property through the intensive research of customer needs, building a team, and raising capital.

“There’s Got to Be a Better Way: How to Deliver Results and Get Rid of the Stuff That Gets in the Way of Real Work” (Hachette, 2025)
By Nelson Repenning, professor of management, and Donald Kieffer

The chaos of everyday business forces people into an exhausting, ineffective, seemingly never-ending cycle of work-arounds, firefighting, and Whac-a-Mole. The irritatingly urgent crowds out the lastingly important. In this book, Repenning and Kieffer describe the game-changing discipline of dynamic work design, which improves productivity, reduces costs, and increases efficiency, ensuring that all parts of a company can work in concert.

“Bayesian Entrepreneurship” (MIT Press, 2026)
Edited by Erin L. Scott, senior lecturer of technological innovation, entrepreneurship, and strategic management in the MIT Sloan School of Management; and Scott Stern, the David Sarnoff Professor of Management of Technology and professor of technological innovation, entrepreneurship, and strategic management at MIT Sloan

This edited volume introduces and explores the concept of Bayesian entrepreneurship, a novel framework for understanding entrepreneurial decision-making under uncertainty. It brings together contributions from leading scholars to examine how entrepreneurs form beliefs about opportunities, learn through experimentation, and make strategic decisions.

“Disciplined Entrepreneurship for Climate and Energy Ventures: 24 Steps to Build Solutions for People and the Planet” (Wiley, 2025)
By Ben Soltoff, entrepreneur in residence at MIT Sloan; Bill Aulet, Ethernet Inventors Professor of the Practice; Tod Hynes, senior lecturer of climate and energy ventures; Francis O’Sullivan, senior lecturer in technological innovation, entrepreneurship, and strategic management; and Libby Wayman, senior lecturer of climate and energy ventures

Climate and energy entrepreneurs face challenges that traditional startup playbooks don’t address. Their ventures can require massive capital and take years to reach market, all while striving to achieve a positive impact on people, planet, and profit. This book adapts the MIT-born “Disciplined Entrepreneurship” framework specifically for climate and energy ventures, recognizing that founders in this space need their own approach.

Arts and design, architecture, urban studies and planning

“Tiny Gardens Everywhere: The Past, Present, and Future of the Self-Provisioning City” (W.W. Norton, 2026)
By Kate Brown, the Thomas M. Siebel Distinguished Professor in History of Science

Nurturing health, hope, and community, gardeners in cities and suburbs are reclaiming lost commons, transforming vacant lots into vibrant plots, turning waste into compost, and recreating what was once the most productive agriculture in recorded human history. In a book with global scope, ranging from Estonia to Amsterdam and Washington, Brown contends that urban gardening has many positive spillover effects, from health and environmental benefits to community-building — apart from periods of pushback when others are trying to eliminate it.

“Small-Town Renaissance: Bridging Technology, Heritage, and Planning in Shrinking Italy” (Springer Nature, 2025)
Edited by Brent D. Ryan, vice provost and professor of urban design and public policy in the Department of Urban Studies and Planning; Carmelo Ignaccolo PhD ’24; and Giovanna Fossa

This book explores the transformative power of digitization in rural regions — where technology isn’t just a tool, but a lifeline for local culture, economic resilience, and future development. Born from a unique research collaboration between the MIT and Politecnico di Milano, this book brings together scholarly work on shrinking towns, economic development, and digital innovation. The project tackled some of the most pressing challenges facing rural Italy — from population decline to economic stagnation — through the lens of digital transformation. 

“Blanking: An Annotated Archive of Projects and Thoughts on Architecture” (Park Books / University of Chicago Press, 2026)
By Rosalyn Shieh, assistant professor in the Department of Architecture, and Troy Schaum

Based on the work and vision of their architecture firm Schaum/Shieh, this book shares what is said and what can be heard in a studio. So much of architectural thinking and knowledge is presented, formulated, and traded in spoken words: pinups, meetings, walkthroughs. Those exchanges inform this book, in which ideas and knowledge that are usually only spoken are made accessible to readers.

“Design Before Disaster: Japan’s Culture of Preparedness” (University of Virginia Press, 2026)
By Miho Mazereeuw, associate professor in the departments of Architecture and Urban Studies and Planning

Few countries have faced as many environmental disasters as Japan, which has endured typhoons, cyclones, floods, earthquakes, volcanic eruptions, and tsunamis. Japanese residents have responded to their precarious circumstances by developing a unique culture of disaster preparedness, equipping the island nation to plan for future emergencies and to greatly reduce their impact. Mazereeuw offers a detailed framework to design and prepare for anticipated disasters and describes effective interventions in urban landscape and architecture. 

“Reconstruction as Violence in Assad’s Syria” (American University in Cairo Press, 2025)
Edited by Nasser Rabbat, professor of architecture and director of the Aga Khan Program for Islamic Architecture at MIT, and Deen Sharp, with a foreword by Hashim Sarkis, dean of the MIT School of Architecture and Planning

This book delves into the complex interplay of post-conflict reconstruction in Syria, challenging the traditionally held dichotomy between the end of violence and the commencement of rebuilding. The contributors to this volume — architects, urbanists, geographers, and historians — employ critical concepts such as urbicide, domicide, and “civilian crisis architecture” to argue against the conventional theoretical frameworks that support a neat separation of phases.

de MIT News https://ift.tt/b7kIyjz

Improving the speed and energy-efficiency of AI agents

Agentic workflows are artificial intelligence-powered software systems that chain together multiple models and external tools to tackle complicated tasks, like analyzing a video and answering questions about it.

But the way these highly fragmented systems are designed and deployed often causes inefficiencies that can lead to wasted computation, energy, and cost. 

To improve efficiency, researchers from MIT and Microsoft developed an intelligent system that streamlines the process of designing agentic workflows and automatically optimizes how those workflows are implemented. 

With this new method, a developer can describe what they want the agentic workflow to do in plain language, without needing to specify all the details of their application in advance. 

The system automatically figures out the best models and tools to use, as well as the ideal hardware configuration and computational resource allocation when the workflow is executed by a cloud provider.

It adjusts those configurations on the fly based on each user’s priorities, such as minimizing costs or maximizing speed.

When tested on several agentic workloads, this new system reduced the number of computational units needed for deployment, significantly cutting energy requirements and costs compared to traditional approaches without hampering performance.

“Agentic workflows are getting very complicated and quickly becoming the backbone of what cloud providers are doing. Energy usage is a huge concern, so we need to be very careful about how efficient these workflows are. It is very easy to over-allocate resources, wasting energy and money. Enabling a cloud provider to intelligently make these workflows more resource-optimal is a win for everyone involved,” says Gohar Chaudhry, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this system.

He is joined on the paper by Adam Belay, an associate professor of EECS and a member of the MIT Computer Science and Artificial Intelligence Laboratory; senior author Ricardo Bianchini, technical fellow and corporate vice president at Microsoft Azure; and others at Microsoft Azure. The paper will be presented at the USENIX Symposium on Operating Systems Design and Implementation.

A configuration conundrum

An agentic workflow is a system composed of several autonomous AI agents that collaboratively use various models and tools, like databases or Python programs, to dynamically complete a multi-step task, such data processing or code generation. 

These workflows can serve as behind-the-scenes processes that power user-facing applications.

Typically, developers must hard-code all technical choices upfront. They need to define which AI agents, models, and tools to use, and the order in which to use them. They also must specify the hardware that runs the workflow and how to balance tradeoffs like speed versus cost. 

This is especially challenging because agentic workflows bring together multiple black-box models and diverse tools, each with their own configuration options, which may be offered by different companies. 

If a new AI model is released that would improve the application’s accuracy or efficiency, the developer would need to start from scratch to implement it.

“Even if you wanted to do all this manually, it is unlikely that you’ll be able to configure the workflow optimally because the space of possible configurations is so large,” Chaudhry says. 

In addition, the cloud data center that deploys the application for customers can’t see inside the workflow to allocate its hardware resources in the most efficient manner at the time of the user’s request. 

With this new system, called Murakkab (an Urdu word that means a composition of things), the researchers sought to optimize the entire agentic workflow process.

Dynamic decision-making

First, Murakkab enables developers to create an agentic workflow by describing their intent for the application in high-level terms, rather than detailing how the many components of that workflow should be combined. 

For instance, a developer might describe a video Q&A application that extracts key frames, generates a transcript, and then answers user queries about the video. 

“There are many ways to do this, and all these different models and tools have implications on how fast the application can finish the task,” he says. 

Murakkab takes the developer’s straightforward specifications and automatically identifies the best existing models and tools to put together into the workflow. 

It also determines which components need to run sequentially and which can be run in parallel to boost performance. 

“The platform makes configuration decisions dynamically over time, so if a new model or GPU accelerator comes out tomorrow, the developer doesn’t need to worry about that,” he says.

When the cloud provider deploys that application for a customer, Murakkab optimizes the workflow by configuring its components to meet the user’s constraints, such as prioritizing accuracy while meeting a latency requirement. 

It adaptively identifies ideal hardware allocations and deployment schedules to maximize efficiency in real time, then generates a workflow that is ready for the cloud provider to execute.

“Our system also gives cloud providers visibility into multiple workloads, so the provider can share computational resources in the most efficient manner while satisfying the constraints of users,” he says.

When tested on diverse agentic workflows for video Q&A and code generation, Murakkab met user requirements while using only about 35 percent of the computation required by other methods. It consumed only about 27 percent as much energy for less than 25 percent of the cost.

The dynamic nature of Murakkab also enables users to balance tradeoffs. In one instance, the system lowered energy consumption of an agentic workflow by more than an order of magnitude with only about a 2 percent drop in accuracy for the customer.

The system was also able to identify an unexpectedly ideal configuration for a model that selects video frames, optimizing performance for a video Q&A task. This type of optimization would be nearly impossible for a developer to do manually, Chaudhry says. 

Next, the researchers plan to expand their system to more complex workflows and larger computing clusters while exploring opportunities to optimize new agentic applications. 

“There is a lot of potential to make these workflows more resource-optimal so they consume far less energy, but we need to be thinking about this at the scale of major cloud platforms,” says Chaudhry.

This research was supported, in part, by the Semiconductor Research Corporation and the U.S. Defense Advanced Research Projects Agency.

de MIT News https://ift.tt/OvlE753

What happens when environmental change outpaces life’s ability to adapt?

When an animal’s environment changes faster than the animal can adapt, its chances of survival can flat-line. The same is true for populations, and even entire species. 

Now, scientists at MIT and the University of Leicester have found that this connection between evolutionary adaptation and the pace of environmental change holds up at the global scale as well — and can determine life’s susceptibility to mass extinction. The researchers developed a theoretical model of this phenomenon, which they present in a paper appearing today in Physical Review Letters.

The team compared the model with available data from past major mass extinctions, including how fast the global environment changed at the time of each event. The model successfully predicted the severity of most mass extinctions in Earth’s history, or the fraction of life that was unable to adapt, and therefore went extinct. 

Interestingly, the researchers found that the range of adaptation rates across animal groups is broadly similar to the range of rates at which the environment can change.

“What we’re beginning to see is a certain level of organization, and ways in which life behaves that are consistent with the ways in which the environment behaves,” says study author Daniel Rothman, professor of geophysics and co-director of the Lorenz Center at MIT. “It may be that life has evolved so that its range of adaptabilities matches the range of stresses that it meets.”

Rothman’s study co-author is Sergei Petrovskii, professor of applied mathematics at the University of Leicester in England.

A catastrophizing connection

The connection between extinction and environmental change is not new. In the late 18th century, the French naturalist Georges Cuvier, who is often referred to as the founding father of paleontology, was the first to propose the concept of “catastrophism.” He had discovered fossil bones near Paris that didn’t match any animal known to exist at the time. Cuvier concluded that the bones were from a group of giant mammals that existed at one time but was no longer around. He proposed, then, that an entire species could disappear, or go extinct, likely due to a widespread catastrophe. 

“That itself was a major idea, that a species could go extinct,” Rothman says. “And he had suggested it was an environmental catastrophe that had caused it.”

The concept of catastrophism later gave way to the view that Earth’s history was shaped mainly by slow, gradual processes. But in the mid-20th century the American geologist Norman Newell revisited the problem. In seeking the cause of extinctions, he proposed what Rothman and Petrovskii call the “rate-mismatch” hypothesis, the notion that extinction occurs when the rate of environmental change is higher than the rate at which a species can evolve to adapt. 

Biologists have since observed Newell’s hypothesis play out in many cases where changes in the environment have driven the extinction of individual species. Rothman and Petrovskii wondered: Could the hypothesis also apply at the global scale?

“We know that individual species go extinct when environmental change outpaces their ability to adapt,” Rothman notes. “But it hasn’t been clear whether this same idea applies at the scale of global extinction events.”

Finding a mismatch

For their new study, the researchers looked to test the rate mismatch hypothesis at the global scale. They wanted to see whether mass extinction events in history can be explained by a mismatch between the rate of global environmental change and the rate at which life around the world can adapt. 

To do so, at least in theory, they would have to compare two sources of data: the rates at which the global environment has changed over time and the rates at which different groups of organisms adapt to environmental change. The first can be found in geological records, which scientists have used extensively to infer how the Earth’s climate changed through history. The second, however, is almost impossible to record.

“We’re talking about the rates at which organisms adapt to major environmental change at effectively geologic timescales, from thousands to millions of years,” Rothman says. “And that doesn’t lend itself to direct observation.”

In place of actual data, the researchers aimed to construct a general mathematical theory to describe the range of adaptation rates across animal groups around the world. In this context, “adaptation” refers to any change within a species, over time periods that are much longer than a generation, that enable the species to persist as its environment changes. 

It is generally understood in evolutionary theory that a species can successfully adapt only when multiple conditions are met. For instance, there needs to be variation in the population, these variations must be heritable, some variations enable an organism to adapt better than others, and the organisms that adapt better should leave more offspring. If all these conditions are met, the entire species should be able to adapt to a given environmental change. However, if any one condition fails, the population will go extinct. 

Rothman and Petrovskii recognized that in this case, a species’ probability of successfully adapting multiplies with every condition that it meets. And it turns out that this pattern can be described mathematically as a very simple, bell-shaped curve. Such a curve essentially describes what fraction of the world’s animals can adapt at given rates, from the slowest to the fastest adapters, and how this fraction changes nonlinearly with the rate of adaptation. This curve generally shows that most animal groups can adapt at intermediate rates, while fewer animal groups adapt at the slowest and fastest rates. 

After they established this general pattern of adaptation rates, the researchers looked to see how this pattern compares to recorded rates of environmental change, and how these two rates match, or don’t match, at times of mass extinction. 

To do so, they considered paleontological and geochemical data from 27 episodes over the last 450 million years where the carbon cycle experienced significant change — a measure that is generally understood to reflect global environmental change. They then compared rates of environmental change with the fraction of animal groups that went extinct during each episode — numbers that were established previously in a well-regarded study by paleobiologist John Alroy. 

In the end, Rothman and Petrovskii observed that indeed, for almost every mass extinction event in the last 450 million years, there was a mismatch in the rates at which the environment changed and at which animals could adapt; mass extinctions occurred when a significant fraction of animals could not adapt fast enough to match the changing environment. Their results confirm that the rate mismatch hypothesis applies at the global scale.

What’s more, this mismatch in rates could predict the severity of extinction events, or the fraction of animal life that went extinct given the rate at which the environment changed. 

In the case of the end-Permian extinction, it’s likely that the rapid acidification of the ocean outpaced organisms’ ability to evolve adequate protections, leading to the extinction of over 80 percent of the world’s marine species. 

The team’s work focuses on applying the new model to past extinction events. But the work could also provide a framework for understanding modern extinction risk. 

“Carbon dioxide levels in the ocean are increasing today at a rate which, when appropriately re-scaled, is similar to rates of carbon-cycle change that are just lower than those associated with major extinction events in the past,” Rothman says. “It suggests that modern environmental change may be approaching rates beyond which adaptation becomes increasingly difficult.” 

This research is supported, in part, by Schmidt Sciences, LLC; the MIT Climate Grand Challenges; the U.S. National Science Foundation; the European Space Agency; and the London Mathematical Society.

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

Computer model could enable bridges and buildings that use less material

In 2022, global production of construction materials accounted for more than 7 percent of total carbon emissions. But how many of those materials were truly necessary to build houses, buildings, and bridges?

A technique called topology optimization can design structures that reduce the amount of material used, in some cases by as much as 90 percent, which would represent a multi-gigaton reduction in building emissions. Unfortunately, topology optimization is mostly used by researchers for applications like 3D printing rather than by engineers designing at the scale of buildings and bridges.

That’s because topology optimization doesn’t create structures that can easily be built on time and budget, which are the things builders really care about.

Now MIT researchers have created a way to make topology optimization designs more buildable. Their framework, described in a new paper in Automation in Construction today, allows users to apply constraints to algorithmically generated structures to limit their complexity. For instance, the approach allows users to limit how many components meet at each point of their design and how small they want their smallest parts. It also builds on previous work by designing structures with multiple materials and taking into account materials’ properties to distribute load and specify part connections.

“There’s an interplay between the materials you’re using, the constructability of designs, and the optimization of the structure,” says senior author Josephine Carstensen, MIT’s Gilbert W. Winslow (1937) Career Development Professor in Civil Engineering. “You need to be able to address all three at the same time. That’s what we tried to do here.”

The researchers used their approach to design steel, wood, and multimaterial truss structures that support loads in buildings and bridges, showing the carbon emissions associated with materials changed significantly when different constraints were applied. They hope their framework will move topology optimization closer to being used in real-world construction.

“In the literature, there’s sometimes been a disconnect between the carbon savings you can achieve on a computer and the realistic carbon savings you can achieve for built structures — especially when it comes to design technologies like topology optimization,” Carstensen says. “The problem lies in the lack of constructability of designs. These designs have been perceived as too difficult to make with conventional methods, so they are never even attempted. That’s what is exciting about our approach: We can add constraints so that you will never be in a situation where the design that comes out is too hard to make.”

Joining Carstensen on the paper is first author and civil and environmental engineering PhD student Zane Schemmer.

More buildable designs

Computer-based topology optimization has been around for decades. It uses computer programs to optimally distribute material in a given space, for instance creating the strongest possible structures at the lowest weight. The resulting designs are often complex, spider web-like structures that would be a challenge for even the most capable engineers to build.

“A big question Josephine and I were asking is why isn’t industry using it?” Schemmer recalls. “What are the obstacles that prevent industry from designing things more efficiently, and how can we fill the gaps between research and real life?”

In recent years, several researchers have developed ways to make topology optimization easier to use. For their study, Schemmer and Carstensen wanted to bring those approaches together and add new capabilities, like creating designs that use multiple materials, which has been another challenge in the field.

“A big aspect of sustainability going forward will be not only using less material, but also implementing materials efficiently based on considerations like where you are in the world, your access to materials, and each of their associated carbon costs,” Schemmer says.

To build their framework, they used a class of equations called mixed integer algorithms that help make binary decisions about things like materials and connections.

“You can’t have a part that’s 72 percent timber and 28 percent steel,” Schemmer says. “Instead, it says, ‘This truss or cable is going to be made out of this,’ and then based on that decision, how do we make sure all of these connections meet their strength standards?”

The system’s decisions also take into account material properties. For instance, steel struts can withstand compressive loads, but steel cables cannot. The model also has more realistic modeling of how parts connect than previous approaches.

“In 3D printing, the way things come together is easy,” Carstensen says. “In construction, that’s not the case. If you’re building with timber there’s a certain rule set, versus steel has a different rule set.”

Users can also decide how complex they want their design to be by specifying the maximum number of connections at each joint and the minimum angle between connected components. The model also creates minimum size limits for parts, further improving its constructability.

“It’s tough to give a contractor these complex, intricate designs because it’s going to be super difficult to build,” Schemmer says. “A lot of times contractors won’t pick up a project like that to begin with.”

The researchers compared structures designed with their approach to structures designed with conventional topology optimization, showing dramatic differences in final designs that transformed how the structures would be built. Using the Lockport “Upside-Down Bridge” near Buffalo, New York, as an example, they applied individual constraints, like a minimum angle on part connections or minimum part sizes, to the bridge’s truss design, to better understand how each constraint impacted final designs.

Finally, they made truss designs that used wood only, steel only, and combined wood and steel, showing how different projects offered tradeoffs with respect to environmental impact and constructability.

“We saw how the system knew that you could design a bridge of pure steel, but that might not be best from a carbon standpoint,” Schemmer says. “Or you could design a bridge out of purely timber, but that might not be the strongest. But these materials can work together, so you use timber for the carbon savings and steel where you need extra strength, and there’s a balance you can find in these structures.”

From research to industry

The researchers say their approach is more computationally intensive than some others, but they were able to use a MacBook Pro to run the programs in their experiment, and they believe it’s practical for most civil engineering firms.

“It’s computationally a little tougher to solve, but there’s a lot of tools coming out nowadays that make these problems a lot more feasible,” Schemmer says. “This approach has been avoided by industry in the past, but now we think it’s a practical way to solve problems dealing with variable constraints.”

If users have more computational resources, the researchers say their approach could work with a long list of materials and far bigger structures than homes, small buildings, and bridges.

Moving forward, Carstensen says the team plans to build scaled-down structures designed by the model to further validate its predictions. They also want to add constraints to their model to make it even more seamless for civil engineers to use when designing the world’s infrastructure.

“As a structural engineer by training, I was never taught how to design for low-carbon,” Schemmer says. “To tackle a problem as big as climate change, addressing the built environment is a great place to start. One of the most tangible things we can do is work at the layer of construction, at the design stage, because that’s a fundamental step that we can control. There’s a lot of decisions we make early on that lead us to use extra material we don’t need.”

The work was funded by the MIT Morningside Academy for Design.

de MIT News https://ift.tt/25Mm9dz

Exploring the societal impacts of AI

At the recent AI and Society Forum at MIT, experts from across the Institute discussed the potential benefits and dangers of technological innovation on labor, the nature of work, civil discourse, election administration, and other topics.

The event featured individual research presentations and panel discussions, as well as a musical performance exploring the use of generative artificial intelligence in the arts.

The forum was co-organized by the School of Humanities, Arts, and Social Sciences (SHASS) and the Social and Ethical Responsibilities of Computing (SERC). It was presented in collaboration with two of MIT’s strategic initiatives: the MIT Generative AI Impact Consortium (MGAIC) and the MIT Human Insight Collaborative (MITHIC).

Agustín Rayo, the Kenan Sahin Dean of SHASS, and Dan Huttenlocher, dean of the MIT Schwarzman College of Computing, provided opening remarks.

Rayo said bringing scholars from across MIT together was intentional because understanding AI’s impact requires expertise from disciplines throughout the Institute.

“Paying attention to the societal consequences of AI is not a departure from MIT’s mission; it’s a way of ensuring that our technical leadership has maximum impact,” Rayo said.

Huttenlocher added that computing and AI’s rapid growth makes it critical to support interdisciplinary conversations and research.

“Understanding where AI excels and where it falls short is essential not only to unlocking its benefits, but also to avoiding critical errors, overreliance, and unintended consequences,” Huttenlocher said.

Jobs and AI 

Held in the Tull Concert Hall in MIT’s Linde Music Building, the May 12 forum opened with a keynote presentation from economist David Autor, the Daniel (1972) and Gail Rubinfeld Professor in the MIT Department of Economics. Autor challenged the common narrative that AI will simply eliminate jobs by proposing instead that technology's impact depends on how it affects the scarcity and value of human expertise. 

“When I think about how technology interacts with the value of labor, I think about it in terms of how it changes the scarcity of expertise, whether it makes it more valuable or whether it makes it more of a commodity,” he said.

Autor said that what matters is whether automation removes routine supporting tasks or removes expert tasks. He argued that AI will likely create new specialized work, requiring proactive policies around worker training, wage insurance, and broader capital ownership.

A panel discussion followed, featuring experts from MIT discussing how work is changing and what it means for society. 

Daniela Rus, the MIT Panasonic Professor of Computer Science and director of the Computer Science and Artificial Intelligence Laboratory (CSAIL), described excitement around ways AI could enhance the workplace.

“I’d like to imagine the robot as your friend and assistant, as someone who watches you and figures out how to help you as someone you can task at a high level,” she said. 

Still, Rus said, human judgment remains critical in decision-making.

“We could really think about co-work with the AI tools, but the role of the human as the decider, as the person with good judgment, as the person deciding the next step, whatever that is, remains super important,” she said.

David Mindell, professor of Aeronautics and Astronautics and the Dibner Professor of the History of Engineering and Manufacturing in the Program in Science, Technology, and Society, says the nature of work has constantly changed over the years, but “what matters is the new work.” 

“We need to be supporting individuals, the economy, professions, to constantly be creating the new work,” he said. “It’s absolutely imperative that we give the tools to the young people and let them do what they find creative and show us what the new work is going to be.”

Panelists also talked about the need to maintain safety standards, while also exploring ways to find efficiencies. Mindell used an example of cargo flights that require six pilots due to the length of the flight.

“We don’t know how to take that six number down to five yet, much less two, one, or zero. There's a lot of money behind solving that problem, but there's also a very rich system that has evolved to make those systems safe,” he said.

Sendhil Mullainathan, the Peter de Florez Professor with dual appointments in the MIT departments of Economics and Electrical Engineering and Computer Science (EECS), described a vision of AI’s utility and growth that offers productivity improvements, but also cautioned, “I think it's very much worth differentiating productivity gains from things that actually drive long-term growth.”

Either way, Mullainathan said, it’s clear we’re entering a time of high variance with regard to AI’s impact on the workforce.

“If you said, ‘exactly how will organizations restructure?’ I don’t know. But is there going to be a lot of restructuring? It’s hard to believe there isn’t going to be a lot of restructuring. And in some sense, if we know that what we’re entering is a period of high variance, that itself is incredibly informative,” he said.

Democracy and AI

The day’s second session focused on AI technology and its impact on democracy. 

Chara Podimata, the Class of 1942 Career Development Assistant Professor and assistant professor of operations research and statistics in the MIT Sloan School of Management, presented her research on auditing large language models for bias in election information.

“Algorithms decide a lot of things about our lives right now,” she said. “With regard to chatbots and election information, if I take two people and they interact with the same chatbot … how will the chatbot respond? How will it personalize the information it gives to these people?”

A longitudinal study of 12 major models during the 2024 U.S. presidential election season found responses varied dramatically based on stated demographics and political leanings. Her research team is now working on a new audit of the 2026 U.S. midterm elections, using a redesigned survey with input from political science experts.

During a panel discussion, experts raised concern about the potential for AI to erode democratic norms and processes, but also explored potential positive outcomes.

Bailey Flanigan, the Theodore T. Miller (1922) Career Development Professor in the Department of Political Science, who holds an MIT Schwarzman College of Computing shared position with EECS, said she’s skeptical of how some are applying AI as a tool that can get people to reach decisions or consensus more quickly.

“And there is a reason to think that this is nice because it is more efficient. It's easier. But it loses a lot of these procedural elements of democracy that are the rituals of how we come together and make decisions,” she said. “And I think it’s a mistake to forget about that when we start thinking about automation.”

Charles Stewart III, the Kenan Sahin (1963) Distinguished Professor of Political Science and founding director of the MIT Election Data and Science Lab, said one challenge is that governmental structures do not evolve at the same rate as technology.

Stewart said his biggest concern is the potential for AI to lead to chaos during and after elections.

“If and when things go wrong, they can go really bad, and really wrong. If an election is called into question, that can lead to violence,” Stewart said.

“We’ve already seen in the low-tech eras election results being manipulated. What worries me is what I’m going to observe this coming Election Day, and the Wednesday after, and if AI has helped to create irreversible disruptions to the election system,” he added.

Lily Tsai, the Ford Professor of Political Science and director and founder of the MIT Governance Lab (MIT GOV/LAB), said in many ways, AI runs against the democratic norms and commitments necessary for a healthy democracy.

“It is really important not just in terms of design principles, but the commitments of designers to be familiar with the values and principles that characterize what democracy is based on: agency, political equality, mutual respect, inclusion, and autonomy,” Tsai said.

Tsai also noted her research has shown some people are more comfortable interacting with machines. She described a “Socratic dialogue chatbot” her team designed that asks people to articulate the thinking behind their beliefs and positions.

“And that actually, interestingly, seems to moderate their policy position in the process,” Tsai said. “So there are absolutely examples of ways in which AI can have positive impacts on democracy. But it really is about designing with the right principles and evaluating them rigorously.”

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

New chip could help tiny robots traverse complex environments

A new chip developed by MIT researchers could help tiny, low-power UAVs avoid obstacles as they zip around tight corners inside an industrial HVAC system to check for gas leaks.

The chip allows small autonomous robots and other battery-limited devices to construct detailed 3D maps of their environments in real-time using only about as much power as a single LED. A robot could use such a map to plan a collision-free path to reach its goal.

Typically, generating such thorough maps requires power-hungry systems and a great deal of memory to build and store 3D representations of the obstacles in a robot’s environment.

The MIT researchers took a different approach by combining an extremely efficient mapping algorithm with specialized hardware designed to accelerate its workload, which minimizes memory and power consumption. 

This system-on-a-chip consumes only about 6 milliwatts of power, a fraction of the power required by other systems. 

This low-power operation could also make the chip well-suited for lightweight augmented reality headsets that can be worn for extended periods, for applications like educational medical simulation or detailed repair and assembly work.

“This paper showcases a key example of how you can leverage co-design of the algorithm and hardware to really push energy efficiency. While there has been a lot of work looking into compact 3D maps, what stands out about this work is that it also ensures that the process to generate those maps is as efficient as possible. Our chip allows you to store very large maps in a very small space, and do it in a very energy efficient manner,” says Vivienne Sze, a professor in the Department of Electrical Engineering and Computer Science (EECS), a member of the Research Laboratory of Electronics (RLE), and senior author of a paper on the chip.

She is joined on the paper by co-lead authors and MIT graduate students Zih-Sing Fu and Peter Zhi Xuan Li as well as Sertac Karaman, a professor of aeronautics and astronautics and the director of LIDS. The work was recently presented at the IEEE Very Large-Scale Integrated Circuits Symposium.

A more compact map

For a robot, generating a 3D map that includes the obstacles in its environment usually demands a lot of power because it must store images captured by its camera, and process all the 3D pixels in each image multiple times.

Instead of representing the environment using 3D pixels, which are cubes called voxels, the MIT researchers utilized a technique that maps the obstacles in space using ellipsoid blobs called Gaussians. 

The size, shape, and thickness of these ellipsoids can be smoothly adapted, so they match the shape of curved objects more efficiently than if one uses rigid, cube-shaped voxels. 

Importantly, the map captures the obstacles and free space around the robot, and together these let the robot plan a safe, collision-free path. Mapping obstacles and free space with voxels typically consumes a lot of memory, which makes traditional methods power-hungry. Because Gaussians can flexibly fit the geometry, a single elongated ellipsoid can represent a region that would take many voxels, so occupied surfaces and free space are captured far more compactly.

For their new system-on-a-chip, called Gleanmer, the researchers employed an algorithm their lab developed called GMMap that efficiently generates a 3D map of the robot’s environment using Gaussians to represent obstacles. 

With traditional approaches, a robot would need to load and process each depth image several times to adjust the size and shape of the ellipsoids. The system would usually construct Gaussians by comparing all the pixels in an image to each other. But the amount of memory and power needed to do this remains too high for many edge devices.

To solve this problem, the MIT researchers invented a technique that can generate highly accurate Gaussians from depth images with only one pass, after which they can discard the images, so the chip never has to store an entire image at once. 

Instead of comparing each pixel to every other pixel in the 3D image, their algorithm assumes that nearby pixels belong in the same Gaussian, so it only needs to compare each pixel to its neighbors.

“At any point in time, we only need to store a few pixels in memory, which significantly reduces the memory footprint our algorithm requires,” Li says.

Leveraging co-design

But as the robot moves through the space, it usually sees the same object from different viewpoints. When it generates Gaussians, some will overlap because they represent the same object. This can make the 3D map too large to store on an edge device.

Fusing overlapping Gaussians makes the map more compact, but doing so typically requires the algorithm to process many raw pixels stored in memory. The researchers developed a novel technique to perform this fusion process directly on overlapping Gaussians, without needing to revisit the original pixels. Since Gaussians are more compact than pixels, this significantly reduces memory and power requirements.

The same principle runs through their algorithm — most computations operate directly on compact Gaussians rather than the original pixels, enabling energy efficiency.

The researchers exploit this principle to design a chip that keeps the Gaussians it is actively working on within small, fast on-chip memory right beside the computational units. This is only possible because the Gaussian map is so compact.

The Gaussians the robot needs to work on next are waiting in the on-chip memory units, so they don’t need to be fetched from more distant, power-hungry, off-chip storage. 

“By having a dedicated memory that just stores the objects you’ve seen in the previous few frames, you can access the data much more efficiently,” Fu explains.

They tested the system-on-a-chip by reconstructing a range of diverse, pre-existing 3D environments. The chip can also reconstruct obstacles and free space directly from live data streamed from an iPhone camera.

Gleanmer generated detailed 3D maps in real-time while consuming about 6 milliwatts of power. It required only about 2.5 percent of the power that the best existing chip for map construction would need. 

By reusing compact Gaussians along the path as it plans, the chip lets a robot chart a safe trajectory using only about 20 percent of the energy it would otherwise need.

“We reduce the memory consumption by making sure the algorithm is efficient. Then we accelerate the workload that is performed by that efficient algorithm, so in the end, our chip is as efficient as possible,” Li says.

The researchers plan to further improve energy efficiency by moving the processing units on the chip closer to the sensors that gather environmental data. They could also explore additional applications, such as the use of Gaussians to represent schematics. This could help AI systems reason about complex blueprints more efficiently.

“Real-time 3D mapping has been the missing piece for small autonomous systems. A drone inspecting a pipeline or a pair of AR glasses navigating a room both need to understand the space around them — instantly, continuously, and at almost no power cost. Gleanmer makes that possible for the first time in a chip you can hold between your fingers,” says Karaman.

This work is supported, in part, by the MIT-MathWorks Fellowship, Amazon, the U.S. National Science Foundation, and Intel. 

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

Meet the leader of the Department of Biology’s all-important “kitchen”

Early mornings in the halls of Building 68 feature the sounds of rolling wheels on big metal carts, the rattling of glassware, the whooshing of faucets, and the clanking of autoclaves. 

These aren’t the sounds of researchers at work, but rather those of keeping the labs sterilized and stocked with the sundries of research: pipette tips, test tubes, flasks, petri dishes, and more.

Orchestrating this sunrise cacophony and the staff that undertakes it is Karen O’Leary, lab associate and acting supervisor in the Glassware Sterilization Facility, also known as the “kitchen.” 

Thanks, in part, to O’Leary’s proactivity and hard work, the kitchen staff were recently recognized with an MIT Excellence Award in 2025 for exceptional contributions in service of the community. 

“My goal is to get the scientists everything they need to do their research,” O’Leary says. “I’m good at what I do.” 

O’Leary admits she did not always possess such confidence. In almost 40 years at MIT, O’Leary has grown into this critical role for the department, and the department itself has evolved, moving into a brand-new building and away from previously standard practices like submerging equipment in acid for sterilization. 

From rookie to running the show

On Sept. 7, 1987, Karen O’Leary joined the MIT community as a staff member for the first time. The 18-year-old was fresh from vocational high school, where she studied cosmetology but felt too shy to pursue that as a career. She was also nervous about joining a research institution.

“When I started, I didn’t even know what a beaker was,” she recalls. 

Too embarrassed to admit in her interview that she couldn’t remember her brand-new home phone number, “I just made one up.” Fortunately, this didn’t prevent her from getting the job, where she worked under the mentorship of Thelma Watkins, who would retire in 1996 after 21 years at MIT. Watkins was critical for instilling a good work ethic and boosting O’Leary’s confidence. 

“She taught me to show up every day, and work hard, and laugh,” O’Leary says.

Even now, O’Leary continues to bring joy to that daily diligence, for herself and for her staff.

“Karen is always on top of things,” says longtime friend and fellow Lab Associate AnnMarie Budhai. “She doesn’t refuse work and always goes above and beyond.” 

Facilities and Operations Manager Cesar Duarte says that O’Leary’s long tenure, support, and knowledge have been invaluable as he transitioned into his role in Building 68 starting in 2023.

“Karen is one of those people who makes everything around her run more smoothly and more pleasantly,” Duarte says. 

Better, faster, safer

Although some might consider it drudgery, O’Leary says that washing glassware is her favorite task. 

“I like that when I wash, I can see the job is complete at the end of the day,” she says. 

Although washing glassware is a perennial task, safety and efficiency have come a long way in the past 38 years. More-effective autoclaves and dishwashers have eliminated steps like steaming to dissolve agar solvents before autoclaving, and scrubbing individual test tubes before washing.

O’Leary was working for the department in 2011 when Building 68 piloted a new approach to MIT’s management of regulated medical waste (RMW), such as petri dishes, blood, and needles — the new system, which is cheaper and produces less waste, is now used by all departments at MIT that produce RMW.

“EHS [the Environment, Health and Safety Office] has come really far — I’m glad we got away from acid,” O’Leary notes of the bygone era of submerging glass pipettes for sterilization. “Back then, no one knew of a better way.” 

Other tasks include cleaning velvets, which are used for replicating bacterial colonies on petri dishes, and pouring agar plates. 

“Everyone knows how to do almost every job, so we can take turns doing different tasks,” O’Leary says. “If you get sick, there’s always someone to cover.”

All in the family

For O’Leary, kinship with MIT has spanned generations. O’Leary was raised in Weymouth, Massachusetts, by a father who worked at MIT as a supervisor in the sheet metal shop. Having raised children of her own, now grown, O’Leary came to greatly appreciate the flexibility her job has granted her.

“I’ve had great work-family balance here,” she says. Even though she’s often at work more than an hour before the researchers that the kitchen serves, “The hours are great, and with MIT Health right across the street, it was easy to take everyone to doctors’ appointments.” 

She’s also gained a chosen family at MIT, spending breaks at work taking long walks along the Charles River, “talking about anything and everything” with colleagues like Budhai and Lab Aide Janet Katin. 

“We really grew up together,” she says. 

Working at MIT has provided O’Leary with support and community, and she’d like to pay it forward. In addition to strolling with colleagues, she hits the gym to help maintain the energy required for her highly active work. 

“I don’t like sitting around,” she says.

In addition to maintaining her stamina at work, she hopes that taking care of herself will keep her actively involved if she ever has grandchildren, and enable her to help neighborhood kids when she someday retires.

“I owe a lot to MIT,” she says. “I have been allowed to work hard and get satisfaction and have been appreciated and given space to care for my family.”

O’Leary returns this care to the Department of Biology in spades.

“It’s an understatement to say that Biology is lucky to have her,” says Duarte. “Karen’s overflowing energy, attention to detail, and care for the Biology research community are nothing short of amazing.”

de MIT News https://ift.tt/w0PbGHR