Machine learning

Prof. Danielle Li and graduate student Lindsey Raymond speak with NPR hosts Wailin Wong and Adrian Ma about how generative artificial intelligence could impact the workplace based on their research examining how an AI chatbot affected workers at customer contact centers. “A lot of what customer service is, is about managing people's feelings 'cause people come, they're tired or whatever,” says Li. “And so in some sense there's kind of this sort of human soft skills component that these technologies are able to capture in a way that prior technologies couldn't.”

Institute Prof. Daron Acemoglu and Prof. Aleksander Mądry join GBH’s Greater Boston to explore how AI can be regulated and safely integrated into our lives. “With much of our society driven by informational spaces — in particular social media and online media in general — AI and, in particular, generative AI accelerates a lot of problems like misinformation, spam, spear phishing and blackmail,” Mądry explains. Acemoglu adds that he feels AI reforms should be approached “more broadly so that AI researchers actually work in using these technologies in human-friendly ways, trying to make humans more empowered and more productive.”

MIT researchers have developed a new method to make chatbots more factual, reports Gerrit De Vynck for The Washington Post. “The researchers proposed using different chatbots to produce multiple answers to the same question and then letting them debate each other until one answer won out,” explains Vynck. “The researchers found using this ‘society of minds’ method made them more factual.”  

MIT researchers have developed a new machine-learning technique that can identify which pixels in an image represent the same material, which could help with robotic scene understanding, reports Kyle Wiggers for TechCrunch. “Since an object can be multiple materials as well as colors and other visual aspects, this is a pretty subtle distinction but also an intuitive one,” writes Wiggers.

Researchers from MIT and McMaster University have used artificial intelligence to identify a new antibiotic that can fight against a drug-resistant bacteria commonly found in hospitals and medical offices, reports Ken Alltucker for USA Today. The researchers believe the AI “process used to winnow thousands of potential drugs to identify one that may work is an approach that can work in drug discovery,” writes Alltucker.

Researchers from MIT and elsewhere have used artificial intelligence to develop a new antibiotic to address Acinetobacter baumannii, a bacteria known for infecting wounds, lungs and kidneys, reports Harland-Dunaway for The World.

Using a machine-learning algorithm, researchers from MIT and McMaster University have discovered a new type of antibiotic that works against a type of drug-resistant bacteria, reports Brenda Goodman for CNN. Goodman notes that the compound “worked in a way that stymied only the problem pathogen. It didn’t seem to kill the many other species of beneficial bacteria that live in the gut or on the skin, making it a rare narrowly targeted agent.”

Researchers from MIT and McMaster University used a machine-learning algorithm to identify a new antibiotic that can treat a bacteria that causes deadly infections, reports Maya Yang for The Guardian. The researchers used an “AI algorithm to screen thousands of antibacterial molecules in an attempt to predict new structural classes. As a result of the AI screening, researchers were able to identify a new antibacterial compound which they named abaucin,” writes Yang.

The Economist spotlights several MIT studies that use artificial neural networks (ANN) to examine the human brain, including work by former postdoctoral researcher Daniel Yamins and his colleagues training ANNs to pick out objects from photographs, reports The Economist.

Prof. Marzyeh Ghassemi speaks with NPR host Kate Wells about a decision by the National Eating Disorders Associations to replace their helpline with a chatbot. “I think it's very alienating to have an interactive system present you with irrelevant or what can feel like tangential information,” says Ghassemi.

Prof. Gilbert Strang received a standing ovation after delivering his last lecture after over 60 years of teaching at MIT, reports Saleen Martin for USA Today. “Teaching has been a wonderful life,” wrote Strang in the comments section of his last lecture on YouTube. “I am so grateful to everyone who likes linear algebra and sees its importance. So many universities (and even high schools) now appreciate how beautiful it is and how valuable it is. That movement will continue because it is right.”

Boston.com reporter Eli Curwin spotlights how after 63 years of teaching and over 10 million views of his online lectures, MIT Prof. Gilbert Strang received a standing ovation after delivering his last lecture. Prof. Michel X. Goemans, head of the Department of Mathematics, notes that Strang “has had a tremendous impact on the teaching of mathematics to tens of thousands of students at MIT through his lectures, to countless of students at other academic institutions through his textbooks, and to millions of people all over the globe.”

A study conducted by graduate student Aspen Hopkins and colleagues trained a version of a GPT neural network on the board game Othello “by feeding in long sequences of move in text form”, reports George Musser for Scientific American. “Their model became a nearly perfect player,” writes Musser.

Prof. David Autor speaks with Greg Rosalsky of NPR’s Planet Money about the potential benefits and downsides of AI, sharing his hope that with the right policies in place to help prepare workers AI could be harnessed to help “reinstate the middle class.” Says Autor: "Basically, the middle-skilled workers of the future could be people who have foundational skills in healthcare, in the trades, in travel and services. Then, with the help of AI, they could get really good at these jobs.”

Wired reporter Caitlin Harrington writes that a study by researchers from MIT and Stanford highlights the impact of generative AI tools on workers and raises a “provocative new question: Should the top workers whose chats trained the bot be compensated?”