Mira Murati’s Secret AI Lab Unveils First Product
▼ Summary
– Thinking Machines Lab has launched Tinker, a tool that automates the creation of custom frontier AI models, cofounded by prominent ex-OpenAI researchers.
– Tinker aims to make frontier AI capabilities more accessible by simplifying the fine-tuning process for businesses, researchers, and hobbyists.
– The tool allows users to fine-tune open source models like Meta’s Llama and Alibaba’s Qwen using supervised or reinforcement learning methods.
– Beta testers report that Tinker is more powerful and user-friendly compared to similar tools on the market.
– The AI industry is closely monitoring the launch due to the high caliber of the team, which includes key contributors to ChatGPT’s development.
Thinking Machines Lab, a well-funded startup established by former OpenAI researchers, has officially launched its inaugural product. Named Tinker, this new platform automates the process of building custom frontier artificial intelligence models. According to Mira Murati, the company’s cofounder and CEO, the tool aims to empower a broader range of individuals. “We believe Tinker will help empower researchers and developers to experiment with models, and will make frontier capabilities much more accessible to all people,” she explained in a recent discussion.
Currently, large corporations and academic institutions often refine open-source AI models to develop specialized versions. These variants might be designed for particular functions, such as tackling mathematical equations, generating legal documents, or responding to healthcare inquiries. This type of work generally demands significant resources, including managing clusters of GPUs and utilizing specialized software to maintain stable and efficient large-scale training operations. Tinker seeks to simplify this by automating a substantial portion of the workload, potentially enabling smaller businesses, independent researchers, and even enthusiasts to undertake their own model fine-tuning projects.
The team at Thinking Machines Lab is placing a significant bet on the idea that simplifying access to frontier model fine-tuning represents the next major evolution in artificial intelligence. Their confidence is not without foundation; the lab’s leadership includes individuals who were instrumental in the development of ChatGPT. Early feedback from beta testers indicates that Tinker is more powerful and user-friendly than other comparable tools currently available.
Murati emphasized the company’s goal to demystify the complexities involved in working with the world’s most advanced AI systems. “We’re making what is otherwise a frontier capability accessible to all, and that is completely game changing,” she stated. She further highlighted the importance of broadening participation, noting, “There are a ton of smart people out there, and we need as many smart people as possible to do frontier AI research.”
At present, the Tinker platform supports fine-tuning for two major open-source models: Meta’s Llama and Alibaba’s Qwen. Users can initiate the fine-tuning process with just a few lines of code via the Tinker API. The system accommodates both supervised learning, which involves training the model with labeled datasets, and reinforcement learning. Reinforcement learning is a growingly popular technique that adjusts a model’s behavior by providing feedback on its outputs. Once the fine-tuning is complete, users have the freedom to download their customized model and deploy it in their preferred environment.
The introduction of Tinker is being closely monitored across the AI sector, largely due to the impressive credentials and proven track record of the team driving its development.
(Source: Wired)
