OpenAI is reportedly working on a secretive project code-named "Strawberry," aiming to push the boundaries of AI reasoning abilities, according to sources familiar with internal documents reviewed by Reuters. The project, previously known as Q*, represents a significant leap forward in AI technology within the Microsoft-backed startup.
Details about "Strawberry" are closely guarded within OpenAI, with the project purportedly focusing on enabling AI models to autonomously navigate and conduct deep research on the internet. This capability aims to surpass current AI limitations, where models typically struggle with advanced reasoning tasks.
Earlier this year, OpenAI reportedly showcased demos of its Q* prototypes, demonstrating impressive capabilities in solving complex science and math problems, which commercially available models struggle to handle.
OpenAI's spokesperson commented broadly on the project, highlighting their commitment to advancing AI capabilities in understanding and interacting with the world more intuitively, akin to human reasoning.
The development of "Strawberry" underscores OpenAI's ongoing research into enhancing AI reasoning, leveraging specialized processing techniques post-pre-training on extensive datasets. The exact timeline for "Strawberry's" public availability remains undisclosed, reflecting the project's current status as a work in progress within the organization.
