ChatGPT maker OpenAI developing new breakthrough reasoning technology
OpenAI, led by Sam Altman, is currently working on a new groundbreaking reasoning technology for its large language models (LLMs). This new technology, codenamed "Strawberry," is expected to enhance the reasoning capabilities of OpenAI's AI models significantly.
According to a report by Reuters, Strawberry is a highly confidential project within OpenAI, previously known as Q*. It has been recognized as a major breakthrough within the organization, with the company hoping that it will revolutionize the capabilities of their AI models.
Advancements in AI
OpenAI has already demonstrated Q* demos to some of its staff, showcasing the LLMs' ability to answer complex science and math questions that are currently beyond the capacity of existing models in the market.
The Strawberry project aims to enable OpenAI's AI to not only generate answers to queries but also to autonomously navigate the internet for in-depth research purposes. This includes the ability to plan ahead and perform what OpenAI terms as "deep research."
Post-Training Models
Strawberry introduces a specialized method for processing AI models after being pre-trained on extensive datasets. It involves a unique 'post-training' technique for OpenAI's generative AI models, allowing them to be adapted and improved for specific tasks even after their initial training on generalized data.
Long-Horizon Tasks
OpenAI intends to leverage Strawberry for long-horizon tasks (LHT) that require AI models to plan ahead and execute a series of actions over an extended period. This includes conducting autonomous web research with the support of a 'computer using agent' (CUA) capable of taking actions based on its findings.
For more information, you can refer to the article on OpenAI's Five-Tier system to gauge AI progress towards human surpassing abilities.
Additionally, Apple is set to get an OpenAI board observer seat to integrate ChatGPT in iOS 18. Read more here.
