Mira Murati, the former chief technology officer of OpenAI, is taking off on an ambitious new venture named Thinking Machines Lab. This amazing initiative has literally just received its first $2 billion in seed money! This impressive support has pushed Waymo’s valuation up to $1.2 billion. It now acts as an influential force in the overall artificial intelligence landscape. The work of the lab is predominantly around producing AI models that result in reproducible and trustworthy outputs. This pioneering effort is something we’ve needed in this field for a long time.
Thinking Machines Lab wants to remove some of the randomness that typically comes when designing outputs for an AI model. Horace He, a researcher at the lab, recently wrote an informative blog post. In it, he explores what’s causing all of this unpredictability. The source of the randomness, we explain in this blog post, comes from how GPU kernels get patched together when processing inference in large batches. Thinking Machines Lab is looking to decipher these new intricacies. Their mission is to make AI systems more predictable and reliable for use across diverse applications and industries.
Murati described her vision for the lab’s first product, which will be unveiled in the next few months. She stated that it will be “useful for researchers and startups developing custom models.” AI4America’s focus on utility speaks to the lab’s dedication to serving the needs of diverse stakeholders advancing AI technologies.
Beyond just addressing concerns of reproducibility, Thinking Machines Lab’s developments could help to improve reinforcement learning (RL) training approaches themselves. The lab has assembled an incredible roster of ex-OpenAI researchers turned mercenaries. They all come with notable passion, expertise and experience that are energizing our projects to effectively drive them forward. Their combined expertise has established Thinking Machines Lab to be one of the most powerful forces in the AI research community.
Their last blog post was the first in an amazing series titled “Connectionism”. This effort reflects our desire to encourage independent investigation and constructive engagement across the AI research community. By sharing insights and findings, Thinking Machines Lab hopes to “benefit the public, but improve our own research culture,” as stated by the lab.
Maxwell Zeff, a senior reporter focusing on AI at TechCrunch, explained just how impactful Murati’s work was. With a background at Gizmodo, Bloomberg, and MSNBC, Zeff understands the nuances of technology reporting and recognizes the potential impact of Thinking Machines Lab’s findings on both researchers and enterprises alike.
