Meet Project 'Strawberry': OpenAI's initiative to enhance AI's reasoning capabilities
OpenAI is working on a new project codenamed 'Strawberry,' according to internal documents reviewed by Reuters. The project aims to enhance the reasoning capabilities of AI models, enabling them not only to answer queries but also to autonomously navigate the internet for 'deep research.' When asked about the project and the details reported in the documents, an OpenAI spokesperson said, "We want our AI models to see and understand the world more like we do."
'Strawberry' project: A leap toward advanced reasoning
The details of the 'Strawberry' project, previously known as Q*, remain confidential within OpenAI. The project is seen as a significant advancement in AI technology, with demos earlier this year showcasing its prowess to answer complex science and math questions, beyond current commercial models. In the internal all-hands meeting held on Tuesday, OpenAI showed a demo of a research project that it claimed to have human-like reasoning skills, according to Bloomberg. It remains unclear if 'Strawberry' was the project demonstrated.
Reasoning crucial to achieve super-human-level intelligence
According to a source familiar with the project, 'Strawberry' could significantly enhance the reasoning abilities of AI models. This involves a unique method of processing an AI model after it has been pre-trained on extensive datasets. Researchers agree that reasoning is crucial for AI to achieve human or super-human-level intelligence. They believe that advancements in reasoning could unlock AI's potential to make significant scientific discoveries and develop new software applications.
Altman foresees AI progress centered around reasoning
OpenAI CEO Sam Altman has previously stated that future progress in AI will be centered around reasoning ability. This aligns with the objectives of the 'Strawberry' project, which reflects a shared industry belief that AI systems can be improved in reasoning over time.
