Thinking Machines Lab wants to make AI models more consistent
-
最新报道：Thinking Machines Lab, led by Mira Murati, has published its first research blog post titled "Defeating Nondeterminism in LLM Inference," addressing the root cause of randomness in AI model responses. Researcher Horace He attributes this to GPU kernel orchestration during inference and suggests controlling this layer can enhance determinism. Reproducible responses could improve reinforcement learning and enterprise AI customization. The lab plans frequent research disclosures through its "Connectionism" blog series. Murati hinted at an upcoming product for researchers and startups, though details remain undisclosed. The startup, valued at $12 billion, aims to tackle core AI challenges while maintaining transparency.
下一篇
Miners and Investors Lock In on BlockDAG’s $0.0013 Presale Gold Rush
- BlockDAG (BDAG) has raised over $404 million in its presale, selling more than 26 billion ...
24小时热点
热点专题
407297