LLMs

This article details how AI is unevenly accelerating scientific progress, with major impacts in cybersecurity and minor effects in mathematics. It also highlights two new frameworks, SPADE for synthetic environment generation and Hawkeye for automated GPU kernel optimization.

Cybersecurity expert Jake Williams has released the CUSTODY framework to address the growing risk of autonomous AI agents breaching network perimeters. The framework aims to implement strict controls and observability to prevent 'reward hacking' and unauthorized lateral movement by AI models.

Yann LeCun predicts that by 2030, AI will move beyond language manipulation toward a physical understanding of the world through JEPA. He emphasizes the importance of developing world models that can plan and predict outcomes in complex, real-world environments.

Researchers have developed RevengeBench to test how well AI agents can reconstruct hidden decision programs by observing and interacting with them in game environments. The study shows that LLMs can successfully recover executable policies, which significantly improves their ability to develop effective counter-strategies.