3 Questions: Modeling adversarial intelligence to exploit AI’s security vulnerabilities
MIT CSAIL Principal Research Scientist Una-May O’Reilly discusses how she develops agents that reveal AI models’ security weaknesses before hackers do.
Topic
Computer Science and Artificial Intelligence Laboratory (CSAIL)
Download RSS feed: News Articles / In the Media / Audio
Toward video generative models of the molecular world
Starting with a single frame in a simulation, a new system uses generative AI to emulate the dynamics of molecules, connecting static molecular structures and developing blurry pictures into videos.
Explained: Generative AI’s environmental impact
Rapid development and deployment of powerful generative AI models comes with environmental consequences, including increased electricity demand and water consumption.
Teaching AI to communicate sounds like humans do
Inspired by the human vocal tract, a new AI model can produce and understand vocal imitations of everyday sounds. The method could help build new sonic interfaces for entertainment and education.
Images that transform through heat
The Thermochromorph printmaking technique developed by CSAIL researchers allows images to transition into each other through changes in temperature.
A new computational model can predict antibody structures more accurately
Using this model, researchers may be able to identify antibody drugs that can target a variety of infectious diseases.
Ecologists find computer vision models’ blind spots in retrieving wildlife images
Biodiversity researchers tested vision systems on how well they could retrieve relevant nature images. More advanced models performed well on simple queries but struggled with more research-specific prompts.
MIT affiliates receive 2025 IEEE honors
Five MIT faculty and staff, along with 19 additional alumni, are honored for electrical engineering and computer science advances.
MIT researchers introduce Boltz-1, a fully open-source model for predicting biomolecular structures
With models like AlphaFold3 limited to academic research, the team built an equivalent alternative, to encourage innovation more broadly.
Study reveals AI chatbots can detect race, but racial bias reduces response empathy
Researchers at MIT, NYU, and UCLA develop an approach to help evaluate whether large language models like GPT-4 are equitable enough to be clinically viable for mental health support.
Lara Ozkan named 2025 Marshall Scholar
The MIT senior will pursue graduate studies in the UK at Cambridge University and Imperial College London.
MIT affiliates named 2024 Schmidt Sciences AI2050 Fellows
Five MIT faculty members and two additional alumni are honored with fellowships to advance research on beneficial AI.
Teaching a robot its limits, to complete open-ended tasks safely
The “PRoC3S” method helps an LLM create a viable action plan by testing each step in a simulation. This strategy could eventually aid in-home robots to complete more ambiguous chore requests.
AI in health should be regulated, but don’t forget about the algorithms, researchers say
In a recent commentary, a team from MIT, Equality AI, and Boston University highlights the gaps in regulation for AI models and non-AI algorithms in health care.
Researchers reduce bias in AI models while preserving or improving accuracy
A new technique identifies and removes the training examples that contribute most to a machine-learning model’s failures.