Mira Murati's Stealth AI Lab Launches Its First Product
12 hours ago
- #AI
- #OpenSource
- #Startup
- Thinking Machines Lab, cofounded by OpenAI researchers, launched Tinker, a tool automating custom frontier AI model creation.
- Tinker aims to make frontier AI capabilities more accessible to researchers, developers, and hobbyists by simplifying fine-tuning processes.
- The tool supports fine-tuning Meta's Llama and Alibaba's Qwen via supervised or reinforcement learning, with users able to download their models.
- Thinking Machines Lab raised $2 billion in seed funding, valuing the startup at $12 billion, with a team including OpenAI veterans like Mira Murati and John Schulman.
- Tinker's reinforcement learning functionality is praised by beta testers for enabling specialized tasks and simplifying the fine-tuning process compared to starting from scratch.
- The company plans to eventually charge for Tinker's API, currently free, and is implementing measures to prevent misuse of open-source models.
- Thinking Machines Lab publishes fundamental research on model training, supporting tools like Tinker, and advocates for openness in AI model development.
- Murati hopes Tinker will counter the trend of commercial AI models becoming increasingly closed, promoting broader access to frontier AI research.