news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Source: Global Times

[Global Times reporter Li Meng] Reuters recently quoted an insider as saying that artificial intelligence (AI) company OpenAI is developing a new AI model codenamed Strawberry. Details about "Strawberry" have never been disclosed before. But OpenAI has been trying to "enable AI models to see and understand the world like humans" and have advanced reasoning capabilities.

Reuters said that the "Strawberry" project is underway. Even within OpenAI, the project is kept strictly confidential. According to several foreign media, the "Strawberry" project was previously called "Q*", which was the reason that triggered the "palace fight" at OpenAI and led to the sudden dismissal of CEO Altman. The US technology website Techreport said that the "Q*" project is controversial. In early June, about 13 employees resigned from OpenAI, including core members such as co-founders. Relevant personnel said that OpenAI has deviated from its original principles and no longer prioritizes safety. In addition, industry experts and OpenAI's internal scientists have also questioned whether AI will be "out of control", believing that it will become more powerful than humans.

Documents show that the goal of the Strawberry project is to not only allow large AI models to generate answers to user queries, but also to plan ahead and browse the Internet autonomously and reliably, so as to conduct what OpenAI calls "deep research." More than a dozen AI researchers interviewed believe that this is a problem that AI models have not yet solved.

Reuters quoted researchers as saying that reasoning is the key to AI achieving human or superhuman intelligence. OpenAI CEO Altman previously said that "the most important progress in the field of AI will revolve around reasoning capabilities." Companies such as Google, Meta, Microsoft, and most academic laboratories conducting AI research are also trying different technologies to improve the reasoning of AI models.

When asked about the "Strawberry" technology, an OpenAI spokesperson did not directly answer the relevant questions, but said in a statement: "It is common practice in the industry to continuously research new AI capabilities, and the system's reasoning ability will improve over time." After learning about the project, Tesla CEO Musk mocked on social media: "We originally thought that the AI ​​doomsday would be the 'Paperclip Maximizer' (a hypothetical theory about the threat of AI), but we didn't expect it to be 'Forever Strawberry Fields' (a rock song)."