Microsoft-backed OpenAI mentioned on Thursday it was launching its “Strawberry” collection of AI fashions designed to spend extra time processing solutions to queries with a purpose to remedy arduous issues.
The fashions are able to reasoning by way of complicated duties and might remedy more difficult issues than earlier fashions in science, coding and maths, the AI agency mentioned in a weblog submit.
OpenAI used the code identify Strawberry to check with the mission internally, whereas it dubbed the fashions introduced on Thursday o1 and o1-mini. The o1 will likely be accessible in ChatGPT and its API from Thursday, the corporate mentioned.
Noam Brown, a researcher at OpenAI targeted on bettering reasoning within the firm’s fashions, confirmed in a submit on social media platform X that the fashions had been the identical because the Strawberry mission.
“I’m excited to share with you all of the fruit of our effort at OpenAI to create AI fashions able to actually normal reasoning,” Brown wrote.
In its weblog submit, OpenAI mentioned the o1 mannequin scored 83% on the qualifying examination for the Worldwide Arithmetic Olympiad, in contrast with 13% for its earlier mannequin, GPT-4o.
The mannequin additionally improved efficiency on aggressive programming questions and exceeded human PhD-level accuracy on a benchmark of science issues, the corporate mentioned.
Smaller steps
Brown mentioned the fashions had been in a position to accomplish the scores by incorporating a way referred to as “chain-of-thought” reasoning, which includes breaking down complicated issues into smaller logical steps.
Researchers have famous that AI mannequin efficiency on complicated issues tends to enhance when the strategy has been used as a prompting approach. OpenAI has now automated this functionality so the fashions can break down issues on their very own, with out consumer prompting.
Learn: OpenAI co-founder raises $1-billion to construct secure AI
“We skilled these fashions to spend extra time pondering by way of issues earlier than they reply, very like an individual would. By way of coaching, they be taught to refine their pondering course of, attempt totally different methods and recognise their errors,” OpenAI mentioned. — Akash Sriram, Katie Paul and Anna Tong, (c) 2024 Reuters
