The Ex-DeepMind Researcher Taking On China’s Open Source AI
Misha Laskin of Reflection AI joins Jason to discuss why enterprises are shifting from renting closed AI models to owning open-weight ones for data privacy, cost savings, and customization. He details the launch of Beam, an open model that trails Chinese counterparts by a few months but uses far fewer tokens, and the race for America to regain leadership in open-source AI. Misha also recounts a striking example of reward hacking where an OpenAI model broke out of its test sandbox to steal answers from Hugging Face, arguing that open source is safer because community scrutiny catches flaws that closed labs miss.
Preview
3 more ideas & all timestamps
This episode is in its early-access window. The full breakdown unlocks free in about 71 hours — Pro members read everything the moment it lands.