Thinking Machines Lab wants to make AI models more consistent
SMRTR summary
Thinking Machines Lab is working to make AI models produce more consistent responses, a problem they believe is solvable by controlling GPU kernel coordination during inference. Researcher Horace He suggests this consistency could improve reinforcement learning and benefit enterprises and scientists. The company, led by former OpenAI CTO Mira Murati, plans to launch its first product in coming months while maintaining open research practices.
SMRTR provides this summary for quick context. The original article belongs to TechCrunch.
Read the original article