Earlier, we reported that OpenAI is developing a new AI engine called Project Strawberry, designed to enhance ChatGPT with human-level reasoning abilities and improve its ability to solve complex math and programming challenges. According to a recent report by The Information, OpenAI is now preparing to integrate the Strawberry engine into ChatGPT.
According to two sources familiar with the development, OpenAI is expected to launch Strawberry for ChatGPT this fall. The Information has revealed that the AI company has already demonstrated Strawberry to U.S. national security officials. Additionally, the report notes that during an internal demo, Strawberry successfully solved the New York Times “Connections” word puzzle.
In addition, Strawberry is capable of conducting in-depth research, planning, and executing tasks, unlocking agentic capabilities. In our previous report, we highlighted that Strawberry’s approach closely resembles a technique called STaR (Self-Taught Reasoner), developed by Stanford researchers, which can greatly enhance reasoning abilities.
Moreover, OpenAI is reportedly developing its next-generation AI model, codenamed ‘Orion,’ which is expected to succeed GPT-4/4o and potentially become GPT-5. Strawberry is playing a key role in generating high-quality synthetic data to train Orion. However, the release date for Orion remains unknown.
0 Comments