Frontier AI systems solve up to 9 of 10 problems. Open-weight entrant solves none.
Earlier this month, the Simons Institute convened a group of theoretical computer scientists from academia and industry to develop a set of...
Greetings from Berkeley, where our fall research programs on Spectral Theory Beyond Graphs, and on Pseudorandomness & High-Dimensional Expansion are...
Hsin-Yuan (Robert) Huang asks whether there is a path forward for working with LLMs in a way that does not undermine the sacred human act of doing research.
In this talk, Alyosha Efros argues that models trained on surface (scraped) data will not transcend mimicry and pastiche. He suggests that to achieve full intelligence, we need deep data, mined by the models through curiosity-driven exploration, driven by intrinsic motivation rather than rewards.
This month, we sat down for a conversation with the newest member of our leadership team, Senior Scientist Jason Lee. Jason is an Associate Professor of EECS and Statistics at UC Berkeley. His research interests are in the theory of machine learning, optimization, and statistics. He has been participating in Simons Institute programs since we opened our doors in Fall 2013, when he was a graduate student.
In his talk from the workshop on The Role of TCS in Modern Machine Learning, Nikhil Bansal (University of Michigan) describes a bound for the Komlós problem that improves upon the long-standing bound due to Banaszczyk.
Greetings from Berkeley! This has been a dizzying summer, one in which we witnessed an inflection point in AI’s capability to autonomously solve major open problems in math and theoretical CS. Like many of you, I feel a mix of amazement at this rapid acceleration and uncertainty about what it means for the future of TCS research as a profession. Scientifically, though, there is much to celebrate: the field has advanced by leaps and bounds this summer, whether done the old-school way, AI-assisted, or AI-driven.
Simons Institute Director Venkatesan Guruswami has been honored this summer with two 2026 Test of Time Awards for papers that transformed coding theory and pseudorandomness, and emerged from a common algebraic core.
As of July 1, our friend and colleague Nikhil Srivastava is the Simons Institute’s Interim Associate Director. He succeeds Sampath Kannan, who returns to UPenn this fall, after two highly successful years as Associate Director of the Simons Institute. Taking Nikhil’s place as Senior Scientist is Berkeley Statistics and EECS faculty member Jason Lee.
First Proof has released the results of its second batch benchmark, assessing the ability of AI systems to autonomously solve naturally occurring mathematical research problems. Their solutions were evaluated by thirty expert mathematicians in a gathering last week at Harvard’s Center for Mathematical Sciences and Applications.
Given points in n-dimensional space drawn independently from an arbitrary (unknown) distribution and labeled red or blue so that the colors can be separated by an (unknown) intersection of k half-spaces, can you efficiently compute (learn) a rule to separate the colors? Santosh Vempala (Georgia Tech) describes a simple algorithm that improves on known results.
Not all convex functions have finite minimizers; some can only be minimized by a sequence as it heads to infinity. In this talk from our recent workshop on The Role of TCS in Modern Machine Learning, Robert Schapire (Microsoft Research) presented joint work with Miro Dudík and Matus Telgarsky aiming to develop a theory for understanding such minimizers at infinity.