The first product from Mira Murati's startup is revealed

Tinker, an API that lets developers alter frontier models without having to worry with complicated infrastructure, was unveiled by Thinking Machines Lab, an AI firm started by former OpenAI CTO Mira Murati and a number of eminent researchers. The specifics: Tinker facilitates the refinement of supervised and reinforcement learning techniques on models such as Alibaba's Qwen and Meta's Llama. With basic code, the models can be modified for specific uses such as chemistry data analysis, math problem solving, and other tasks. Tinker has been used by early adopters like as Princeton, Stanford, and Berkeley to create unique AI systems for research, scientific reasoning, and mathematical proofs. With plans to launch premium tiers in the upcoming weeks, the company is already accepting applications for early access at no initial cost. It takes a lot of resources to create AI models from scratch, thus most firms could be better off adapting pre-existing models. The future, according to Murati and colleagues, will not be determined by who can create the largest all-purpose AI, but rather by who can make it the simplest to produce a large number of hyper-specialized ones.