Computer Science and Artificial Intelligence Laboratory (CSAIL)

FutureTech researcher Tamay Besiroglu speaks with New Scientist reporter Chris Stokel-Walker about the rapid rate at which large language models (LLMs) are improving. “While Besiroglu believes that this increase in LLM performance is partly due to more efficient software coding, the researchers were unable to pinpoint precisely how those efficiencies were gained – in part because AI algorithms are often impenetrable black boxes,” writes Stokel-Walker. “He also points out that hardware improvements still play a big role in increased performance.”

A number of MIT faculty and alumni – including Prof. Daniela Rus, Prof. Regina Barzilay, Research Affiliate Haddad Habib, Research Scientist Lex Fridman, Marc Raibert PhD '77, former Postdoc Rana El Kaliouby and Ray Kurzweil '70 – have been named key figures “at the forefront of Boston’s AI revolution,” reports Wyndham Lewis for Boston Magazine. These researchers are “driving progress and reshaping the way we live,” writes Lewis.

MIT researchers have developed a new technique that uses a large language model to allow robots to self-correct after making a mistake, reports Brian Heater for TechCrunch. “Researchers behind the study note that while imitation learning (learning to do a task through observation) is popular in the world of home robotics, it often can’t account for the countless small environmental variations that can interfere with regular operation, thus requiring a system to restart from square one,” writes Heater. “The new research addresses this, in part, by breaking demonstrations into smaller subsets." 

Boston Magazine spotlights MIT’s leading role in the AI revolution in the Greater Boston area. “With a $2 million grant from the Department of Defense, MIT’s Artificial Intelligence Lab combines with a new research group, Project MAC, to create what’s now known as the Computer Science and Artificial Intelligence Laboratory (CSAIL). Over the next three years, researchers lead groundbreaking machine-learning projects such as the creation of Eliza, a psychotherapy-based computer program that could process languages and establish emotional connections with users (a primordial chatbot, essentially).”

Boston Globe reporter Aaron Pressman spotlights the Perkins School for the Blind Hackathon held at MIT. “The students divided into 10 teams, named after colors, and picked one of eight challenges Perkins had crafted, such as assisting a blind person to navigate an indoor space or to pick up non-verbal cues in video conference conversations,” writes Pressman. “But before writing a line of code, the teams met with people with a disability relevant to the challenge they had selected.”

Prof. Pulkit Agrawal and graduate student Gabriel Margolis speak with The Economist’s Babbage podcast about the simulation research and technology used in developing intelligent machines. “Simulation is a digital twin of reality,” says Agrawal. “But simulation still doesn’t have data, it is a digital twin of the environment. So, what we do is something called reinforcement learning which is learning by trial and error which means that we can try out many different combinations.”

Prof. Mike Stonebraker co-founded DBOS, a serverless software platform, that aims to “put a database system at the bottom of the technology stack as close to the bare metal as possible where the operating system usually sits,” reports Ron Miller for TechCrunch. “Bare metal is a term used to describe the pure hardware layer where no software exists. Flipping the OS and the database is a bold and revolutionary idea,” explains Miller.

Prof. Thomas Malone speaks with Poets & Quants for Executives reporter Alison Damast about the executive education course he teaches with Prof. Daniela Rus that aims to provide senior-level managers with a better sense of how AI works. “We are certainly not trying to teach people to understand the details of how to write AI programs, though some of those in the course may know that already,” Malone says. “What we are trying to do is give them a sense of when it is easy and when it is hard to use AI technology at various times for different kinds of business applications.”

Prof. Daniela Rus, director of CSAIL, speaks with The Economist’s Babbage podcast about the history and future of artificial neural networks and their role in large language models. “The early artificial neuron was a very simple mathematical model,” says Rus. “The computation was discrete and very simple, essentially a step function. You’re either above or below a value.”  

Prof. Philip Isola and Prof. Daniela Rus, director of CSAIL, speak with Associated Press reporter Matt O’Brien about AI generated images and videos. Rus says the computing resources required for AI video generation are “significantly higher than for still image generation” because “it involves processing and generating multiple frames for each second of video.”

Prof. Daniela Rus, director of CSAIL, speaks with Boston Globe reporter Evan Sellinger about her new book, “The Heart and the Chip: Our Bright Future With Robots,” in which she makes the case that in the future robots and humans will be able to team up to create a better world. “I want to highlight that machines don’t have to compete with humans, because we each have different strengths. Humans have wisdom. Machines have speed, can process large numbers, and can do many dull, dirty, and dangerous tasks,” Rus explains. “I see robots as helpers for our jobs. They’ll take on the routine, repetitive tasks, ensuring human workers focus on more complex and meaningful work.”

Prof. Charles Stewart III and Ben Adida PhD ’06 speak with Fast Company reporter Spenser Mestel about how to restore the public’s faith in voting technology. Adida discusses his work launching VotingWorks, a non-profit focused on building voting machines. VotingWorks is “unique among the legacy voting technology vendors," writes Mestel. “The group has disclosed everything, from its donors to the prices of its machines.”

Prof. Erik Demaine speaks with Boston Globe reporter Cate McQuaid about how combining the art of origami with computer science has enhanced his work in both fields. “We get stuck on a science problem and that inspires a new sculpture, or we get stuck trying to build a sculpture,” says Demaine, “and that leads to new science.”

Researchers from MIT and elsewhere have developed an AI model that is capable of identifying 3 ½ times more people who are at high-risk for developing pancreatic cancer than current standards, reports Felice J. Freyer for The Boston Globe. “This work has the potential to enlarge the group of pancreatic cancer patients who can benefit from screening from 10 percent to 35 percent,” explains Freyer. “The group hopes its model will eventually help detect risk of other hard-to-find cancers, like ovarian.”

In a new study examining the potential impact of AI on jobs that employ computer vision, MIT researchers found, “a vast majority of jobs previously identified as vulnerable to AI are not economically beneficial for employers to automate at this time,” reports Catherine Thorbecke for CNN. “In many cases, humans are the more cost-effective way, and a more economically attractive way, to do work right now,” says Research Scientist Neil Thompson, director of the FutureTech Research Project at CSAIL. “What we’re seeing is that while there is a lot of potential for AI to replace tasks, it’s not going to happen immediately.”