August 7, local time,OpenAIChief Executive OfficerSam AltmanA photo of strawberries was released, hinting at the mysterious "Strawberry" project, sparking heated discussions across the Internet.
As OpenAI seeks to raise more funding, its researchers are trying to launch a new AI product that can solve tough problems better than existing ones, The Information reports.
OpenAI is said to have demonstrated the technology to U.S. national security officials this summer, and a person with direct knowledge of the meetings said they had not been previously reported (IT Home Note: Reuters first broke the news last month).
Two insiders revealed that the team aims to launch the new AI project code-named Strawberry (formerly known as Q*) as part of a chatbot (presumably referring to ChatGPT) as early as this fall.
Strawberry can solve math problems it has never seen before, something that existing AI has a hard time doing. It is said that although it was trained to solve problems involving programming, it can answer any technical questions.
Sources said that the "Strawberry" project aims to enhance the reasoning ability of OpenAI's models and its ability to handle complex scientific and mathematical problems, so that large models can not only generate answers to queries, but also plan in advance to browse the Internet autonomously and reliably and conduct "deep research" as defined by OpenAI.