Machine learning

A study by MIT researchers monitored and compared the brain activity of participants using large language models, traditional search engines, and only their brains to write an essay on a given topic, reports Hessie Jones for Forbes. The study “found that the brain-only group showed much more active brain waves compared to the search-only and LLM-only groups,” Jones explains. “In the latter two groups, participants relied on external sources for information. The search-only group still needed some topic understanding to look up information, and like using a calculator — you must understand its functions to get the right answer. In contrast, the LLM-only group simply had to remember the prompt used to generate the essay, with little to no actual cognitive processing involved.”  

Researchers at MIT have found that ChatGPT users “showed minimal brain engagement and consistently fell short in neural linguistic, and behavioral aspects,” reports Kyle Wiggers for TechCrunch. “To conduct the test, the lab split 54 participants from the Boston area into three groups, each consisting of individuals ages 18 to 39,” explains Wiggers. “The participants were asked to write multiple SAT essays using tools such as OpenAI’s ChatGPT, the Google search engine, or without any tools.” 

Forbes contributor Tanya Fileva spotlights how MIT CSAIL researchers have developed a system called Air-Guardian, an “AI-enabled copilot that monitors a pilot’s gaze and intervenes when their attention is lacking.” Fileva notes that “in tests, the system ‘reduced the risk level of flights and increased the success rate of navigating to target points’—demonstrating how AI copilots can enhance safety by assisting with real-time decision-making.”

The New Yorker reporter Kyle Chayka spotlights a study by MIT researchers examining the impact of AI chatbot use on the brain. “The results from the analysis showed a dramatic discrepancy: subjects who used ChatGPT demonstrated less brain activity than either of the other groups,” explains Chayka. 

Researchers from MIT have found that “extended use of LLMs for research and writing could have long-term behavioral effects, such as lower brain engagement and laziness,” reports Theo Burman for Newsweek. “The study found that the AI-assisted writers were engaging their deep memory processes far less than the control groups, and that their information recall skills were worse after producing work with ChatGPT,” explains Burman. 

Researchers at MIT have found that “AI agents can make the workplace more productive when fine-tuned for different personality types, but human co-workers pay a price in lost socialization,” reports Kaustuv Basu for Bloomberg. The researchers concluded “found that humans using AI raised their productivity by 60%—partly because those workers sent 23% fewer social messages,” writes Basu. 

Sloan Lecturer Michael Schrage speaks with Fortune reporter Sheryl Estrada about prompt-a-thons, “structured, sprint-based sessions for developing prompts for large language models (LLMs).” The “prompt-a-thon process reframes prompting as a high-impact diagnostic and design discipline—engineered for fast, actionable insight,” explains Estrada. “It’s not just about using AI more effectively—it’s about thinking and collaborating more intelligently with it,” adds Schrage. 

Graduate student Alex Kachkine has developed a new technique that “uses AI-generated polymer films to physically restore damaged paintings in hours,” reports Benj Edwards for Ars Technica. “Kachkine's method works by printing a transparent ‘mask’ containing thousands of precisely color-matched regions that conservators can apply directly to an original artwork,” explains Edwards. “Unlike traditional restoration, which permanently alters the painting, these masks can reportedly be removed whenever needed. So it's a reversible process that does not permanently change a painting.” 

MIT researchers have found that ChatGPT use can lead to a decline in cognitive engagement, reports Robert B. Tucker for Forbes. “Brain regions associated with attention, memory, and higher-ordered reasoning were noticeably less active” in study participants, Tucker explains.

Researchers at MIT have found that the use of ChatGPT can “reduce activity in brain regions associated with memory and learning,” reports Eve Upton-Clark for Fast Company. “ChatGPT users felt less ownership over their essays compared to the other groups,” writes Upton-Clark. “They also struggled to recall or quote from their own essays shortly after submitting them—showing how reliance on the LLM bypassed deep memory processes.” 

Researchers at MIT have found that “people who used ChatGPT to write a series of essays suffered a ‘cognitive cost’ compared to others who used only their brains or a traditional search engine,” reports Ross Cristantiello for Boston.com. “The researchers found that as users relied on ‘external support’ more and more, their brain connectivity gradually scaled down,” explains Cristantiello. “Subjects who began the tests using ChatGPT before being told to use only their brains showed ‘weaker neural connectivity’ and ‘under-engagement’ of certain networks in their brains.”  

A study by MIT researchers finds that individuals who relied solely on ChatGPT to write essays had "lower levels of brain activity and presented less original writing,” reports Greta Cross for USA Today. "While these tools offer unprecedented opportunities for enhancing learning and information access, their potential impact on cognitive development, critical thinking and intellectual independence demands a very careful consideration and continued research," the researchers explain.

Noman Bashir, a fellow with the MIT Climate and Sustainability Consortium and a postdoc at CSAIL, speaks with Wired reporter Molly Taft about AI and energy consumption. Bashir explains that how quickly a model answers a question has a big impact on its energy use. “The goal is to provide all of this inference the quickest way possible so that you don’t leave their platform,” Bashir says. “If ChatGPT suddenly starts giving you a response after five minutes, you will go to some other tool that is giving you an immediate response.”

Prof. Daniela Rus, director of CSAIL, speaks with Newsweek reporter Marni Rose McFall about the impact of AI on entry level jobs. “We need a strong pipeline of talent that starts with entry-level roles, internships, and hands-on learning opportunities," says Rus. "These early experiences remain essential stepping stones, helping people build technical confidence, domain fluency, and problem-solving skills. And soon, the skills companies will be looking for in entry-level workers is how well they can make the most of AI tools."

Vijay Gadepally, a senior scientist at MIT Lincoln Lab, discusses users can help conserve energy while using AI tools, reports Nicolás Rivero for The Washington Post. Gadepally notes that users can save energy by asking the AI to be concise when you don’t need long answers, as models use more energy for each word they process. “People often mistake these things as having some sort of sentience,” says Gadepally. “You don’t need to say ‘please’ and ‘thank you.’ It’s okay. They don’t mind.”