Based in the Bay Area, Thinking Machines Lab was co-founded by Mira Murati, the former chief technology officer of OpenAI. Indeed, it’s already garnered a record-breaking $2 billion in seed funding! The lab investment represents a total valuation of $1.2 billion. For Murati, it’s the crowning achievement to date as she attempts to transform the landscape of artificial intelligence.
We’re committed to developing AI models that provide reliable, uniform responses. This work addresses a key barrier for researchers and industry. Led by an all-star team of former OpenAI researchers, the lab is strategically positioning itself to build safer, more reliable AI technologies. Horace He, a researcher at the lab, authored a recent blog post titled “Defeating Nondeterminism in LLM Inference,” which sheds light on the root causes of randomness in AI models.
In his recent blog post, he makes the case that GPU kernels are the biggest contributor to nondeterminism during AI model inference. This realization might entail developing more reliable and uniform outputs in generative AI tools. It will serve the public good and help improve the lab’s research culture.
The lab plans to introduce its first product in the next few months. This creative new resource is specifically designed for researchers and startups working on creating their own state of the art AI/ML technology. By producing reproducible AI responses, Thinking Machines Lab hopes to improve the quality of reinforcement learning training as well.
Maxwell Zeff, a senior reporter at TechCrunch who covers AI, recognized the importance of Murati’s new endeavor. She has produced for major outlets, including Gizmodo, Bloomberg News, and MSNBC. His commentary sheds light on why having a stable AI model is becoming critically important to our evolving tech landscape.
In addition to these goals, Thinking Machines Lab hopes to spread practical knowledge about AI and machine learning through its regular blog series “Connectionism.” This initiative has allowed us to broadcast the research findings to a much larger audience. In doing so, it invigorates our scientific community.
“Our goal is not only to benefit the public but also to improve our own research culture.” – Thinking Machines Lab
At the same time, artificial intelligence is developing at a breakneck pace. At Thinking Machines Lab, researchers are at the forefront of this work and have the potential to disrupt and reshape how AI is developed and applied across industries. For researchers and practitioners, we all hope that the new-found religion on reproducibility will wipe away years of pain caused by randomness in AI outputs.
