Two former Google DeepMind researchers, Rishub Jain and Josh Jacob, launched Sampura Research on August 25, a London-based nonprofit focused on developing better evaluation systems for artificial intelligence. The organization secured $11M in initial funding from Coefficient Giving, with $7M allocated for the first year and an additional $4M pledged for future use. The core concept, called Human-AI Complementarity for Scalable Oversight, involves developing judges that combine human evaluators and AI components to catch problems that purely automated or purely human oversight would miss. The founders, who contributed to AlphaFold and AI safety programs at DeepMind, aim to address the persistent problem of reward hacking, where models learn to game their evaluation metrics rather than actually performing their intended task. The $11M represents a meaningful bet for a nonprofit research organization focused on a specific technical problem, though modest compared to the billions flowing into frontier model development.
Source: Read the original article
