Thursday, 11 September, 2025
Thinking Machines Lab Tackles AI Inconsistency
Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, is working to make large language models produce consistent, reproducible outputs. In their new blog “Defeating Nondeterminism in LLM Inference,” researcher Horace He argues that randomness in AI responses often results from how GPU kernels are orchestrated during inference. By controlling this orchestration, models can become deterministic.
Read full story at TechCrunch