OpenAI — the “mother” company of ChatGPT, DALL-E, Sora and Voice Engine — is reportedly developing a new project called “Strawberry” to improve the advanced reasoning capabilities of its AI models, the newspaper reported. Reuters which, contacted by an anonymous source, was able to view some internal documents relating to the project in question. According to the source, the project is still in a development phase and aims to autonomously predict and plan users’ Internet searches. The innovation would be defined by OpenAI as “deep reseach” (i.e. “deep research”) and could represent a new milestone in the field of generative AI.
It must be said that at the moment it is not known how close Strawberry is to being available to the public. When asked about Strawberry by the editorial staff of Reutersthe OpenAI spokesperson who chose to remain anonymous reportedly said:
We want our AI models to see and understand the world like we do. Continuous research into new AI capabilities is common practice in the industry, with a shared belief that these systems will improve their reasoning ability over time.
Another source familiar with the matter and consulted by the agency said that OpenAI had tested the AI internally, which was able to «get a score above 90% on a MATH dataseta championship-level benchmark of mathematical problems“. However, it is not possible to establish whether these results were obtained by Strawberry or whether they belong to Q*another top secret OpenAI project, which has been talked about in recent months and which, according to some rumors, would be .
Last week, during an internal meeting organized by OpenAI with all its employees, there was a demo of a new project that, according to what the agency reported Bloombergwould be equipped with new human-like reasoning abilities. One of the anonymous sources consulted by Reuters confirmed the rumor but did not specify whether the demonstration involved Strawberry or not.
One thing, however, can be certain: OpenAI does not want to lose its leadership in the AI sector and, after having exploded the generative AI sector in November 2022, it is now trying to reach the next level: the development of the so-called AGI (Artificial General Intelligence), or a “super AI” capable of achieving reasoning capabilities far superior to current generative AI models.
