Results 2141 - 2150 of 23899
We’re delighted to share that Miller fellow and Simons Institute Quantum Pod postdoc Ewin Tang has been awarded the 2025 Maryam Mirzakhani New Frontiers Prize for “developing classical analogs of quantum algorithms for machine learning and linear algebra, and for advances in quantum machine learning on quantum data.”
This month, we held a joint workshop with SLMath on AI for Mathematics and Theoretical Computer Science. It was unlike any other Simons Institute workshop I have been to. Over half the participants were mathematicians. But what really set it apart was its afternoons of hands-on tinkering. After lunch on the first three days, participants received a worksheet from the organizers. We opened up our laptops in the Calvin Lab auditorium and did the exercises side by side, with a fleet of TAs among us.
Greetings from the Simons Institute, where we are in the final week of a yearlong research program on Large Language Models and Transformers.
On April 10, Simons Institute Science Communicator in Residence Anil Ananthaswamy sat down with Sasha Rush, an associate professor at Cornell Tech working on natural language processing and machine learning, with a focus on deep learning text generation, language modeling, and structured prediction. This episode of Polylogues explores a significant shift in the last year in how large language models are trained and used.
The leading AI companies are increasingly focused on building generalist AI agents — systems that can autonomously plan, act, and pursue goals across almost all tasks that humans can perform. Despite how useful these systems might be, unchecked AI agency poses significant risks to public safety and security, ranging from misuse by malicious actors to a potentially irreversible loss of human control. In his Richard M. Karp Distinguished Lecture this month, Yoshua Bengio (IVADO / Mila / Université de Montréal) discussed how these risks arise from current AI training methods.
In March, the Simons Institute hosted a Workshop on Quantum Memories. This specialized workshop explored recent progress around robust quantum information storage in physical systems. We’re delighted to share one of our favorite talks from the workshop: “A Local Automaton for the 2D Toric Code,” presented by Shankar Balasubramanian (MIT).
Goal-conditioned reinforcement learning (GCRL) is a powerful way to control an AI agent’s behavior at runtime. That said, popular goal representations, e.g., target states or natural language, are either limited to Markovian tasks or rely on ambiguous task semantics. We propose using automata to represent temporal goals and guide GCRL agents. Automata balance the need for formal temporal semantics with ease of interpretation: if one can understand a flow chart, one can understand an automaton. On the other hand, automata form a countably infinite concept class with Boolean semantics, and subtle changes to the automaton can result in very different tasks, making them difficult to condition agent behavior on. To address this, we observe that all paths through an automaton correspond to a series of reach-avoid tasks and propose a technique for learning provably correct embeddings of "reach-avoid derived" automata, guaranteeing optimal multi-task policy learning. Through empirical evaluation, we demonstrate that the proposed pretraining method enables zero-shot generalization to various task classes and accelerated policy specialization without the myopic suboptimality of hierarchical methods.
In many modern engineering domains, the presence of "observer effects" creates interdependence between measurement and underlying quantities of interest. In such settings, decisions (also called actions or inputs) both impact the underlying system state and determine what information is observed. Accounting for this dual role is crucial for designing reliable algorithms for learning and decision making, for applications ranging from robotics to personalized recommendation systems. In this talk, I will discuss recent work in the setting of partially observed dynamical systems with linear state transitions and bilinear observations. Inspired by the rich line of work on learning and control for linear systems, our goal is to understand how much (and which) data is necessary for reliable decision-making.
First, we will consider learning from observations when the dynamics are unknown. The identification procedure involves heavy tailed and dependent covariates. Nevertheless, we provide finite data error bounds and a sample complexity analysis for inputs chosen according to a simple random design. Second, we consider the optimal control problem with the objective of minimizing a quadratic cost. Despite the similarity to standard linear quadratic Gaussian (LQG) control, neither does the separation principle (SP) hold, nor is the optimal policy affine in the estimated state. Under certain conditions, the SP-based controller locally maximizes the cost instead of minimizing it, and instability can result from a loss of observability. I will conclude with a discussion of open questions. Based on joint work with Yahya Sattar, Sunmook Choi, Yassir Jedra, and Maryam Fazel.
Today, many expect AI to tackle complex problems by performing reasoning—commonly interpreted as large language models generating sequences of tokens that resemble chains of thought. Yet historically, AI reasoning had a very different meaning: executing symbolic algorithms that performed logical or probabilistic deduction to derive definite answers to questions about knowledge. In this talk, I show that such old-fashioned ideas are very relevant to reasoning with large language models today. In particular, I will demonstrate that integrating symbolic reasoning algorithms directly into the architecture of language models enables state-of-the-art capabilities in controllable text generation and alignment.