Machine thinking laboratory, A definite start -up start -up harmonized by outstanding researchers from OpenAI revealed his first product -a tool called Tinker, which automates the creation of custom AI Frontier models.
“We believe [Tinker] It will help to enable researchers and programmers to experiment with models, and make border capabilities much more accessible to all people, “says Mira Murati, co -founder and general director of thinking machines, in an interview with the wired before the announcement.
Immense academic companies and laboratories are already adapting AI Open Source models to create recent variants optimized for specific tasks, such as solving mathematical problems, developing legal contracts or answering medical questions.
Usually, these works include acquisition and management of GPU clusters and the utilize of various program tools to ensure that enormous -scale training is stable and effective. Tinker promises that more companies, scientists and even hobbyists adapt their own AI models by automating a significant part of this work.
Basically, the band bet that helping people refine border models will be another great thing in artificial intelligence. And there is a reason to think that they may be right. The thinking laboratory is directed by scientists who played a basic role in creating chatgpt. Compared to similar tools on the market, Tinker is stronger and user -friendly, according to beta testers I talked to.
Murati says that the thinking of LAB machines is hoping to dismotiate the work related to tuning the most powerful AI models in the world and allows more people to examine external artificial intelligence boundaries. “We state that otherwise the possibility of the border is available to everyone, and this completely changes the game,” he says. “There are plenty of intelligent people and we need as many intelligent people as possible to conduct Frontier AI research.”
Tinker currently allows users to tune two Open Source models: Lama Meta and Qwen Alibaba. Users can write several code lines to utilize the API Tinker interface and start tuning through supervised learning, which means adjusting the model using data marked or by learning to strengthen, an increasingly popular method of tuning models, giving them positive or negative feedback based on their results. Users can then download their refined model and run it wherever they want.
The AI industry is carefully observing the premiere – partly due to the caliber of the band behind it.