Asimov: Building An Omniscient RL Oracle with ReflectionAI’s Misha Laskin

Superintelligence, at least in an academic sense, has already been achieved. But Misha Laskin thinks that the next step towards artificial superintelligence, or ASI, should look both more user and problem-focused. ReflectionAI co-founder and CEO Misha Laskin joins Sarah Guo to introduce Asimov, their new code comprehension agent built on reinforcement learning (RL). Misha talks about creating tools and designing AI agents based on customer needs, and how that influences eval development and the scope of the agent’s memory. The two also discuss the challenges in solving scaling for RL, the future of ASI, and the implications for Google’s “non-acquisition” of Windsurf.

Chapters:

– Misha Laskin Introduction

– Superintelligence vs. Super Intelligent Autonomous Systems

– Misha’s Journey from Physics to AI

– Asimov Product Release

– What Differentiates Asimov from Other Agents

– Asimov’s Eval Philosophy

– The Types of Queries Where Asimov Shines

– Designing a Team-Wide Memory for Asimov

– Leveraging Pre-Trained Models

– The Challenges of Solving Scaling in RL

– Training Agents in Copycat Software Environments

– When Will We See ASI?

– Thoughts on Windsurf’s Non-Acquisition

– Exploring Non-RL Datasets

– Tackling Problems Beyond Engineering and Coding

– Where We’re At in Deploying ASI in Different Fields

– Conclusion

Asimov: Building An Omniscient RL Oracle with ReflectionAI’s Misha Laskin

Kapitelklicken springt direkt hin

Shownotes