OpenAI presents new tool "o1 thinks before it answers"

Samuel Walder

14.9.2024

OpenAI wants to be able to answer even more complex questions with o1.
OpenAI wants to be able to answer even more complex questions with o1.
IMAGO/CFOTO

Is this the revolution? OpenAI presented a new artificial intelligence on Thursday. It should be able to answer complex questions and display even more human behavior.

No time? blue News summarizes for you

  • OpenAI has published the new AI model o1, which can solve complex questions.
  • The model uses a thinking strategy to recognize errors and solve them in sub-steps, making it comparable to human experts in tests.
  • OpenAI is looking for investors. Potential partners such as Microsoft, Nvidia and Apple are said to be under discussion.

OpenAI, the developer of ChatGPT, launched o1 on the market on Thursday. This is a new model of generative artificial intelligence (AI) that is able to think and respond to more complex questions, particularly mathematical ones. The hope of the developers is to avoid wrong answers and risks.

"o1 thinks before it answers," says OpenAI in a recent press release. The company is thus making progress towards its goal of developing a "general" AI. To be more precise: an artificial intelligence that is even more similar to human intelligence.

Sam Altman, the head of OpenAI, congratulated his colleagues at X on this "new paradigm: an AI that can think in a complex and generalist way".

New model can compete with human experts

The new model should be able to help solve mathematical problems or create lines of code. "In many benchmark tests that require strong thinking skills, o1 can compete with the performance of human experts," the company emphasizes.

In a math competition for American high school students, o1 "placed among the top 500 students," OpenAI continues. "Similar to a human who can think for a long time before answering a difficult question, o1 uses a sequence of reasoning," explains OpenAI. "It learns to recognize and correct its mistakes. It learns to break down complex situations into simpler steps.

Investors wanted for OpenAI

However, he pointed out that the technology "is still imperfect and limited and seems more impressive when first used than after a longer period of engagement with it". The beta version of o1 was initially made available to paying users of ChatGPT on Thursday.

The news agency AFP tested o1 with simple logic questions. In doing so, o1 achieved the same results as GPT-4o. However, the new AI tool took more time and went into more detail instead of generating an answer almost immediately.

Another difference is that the new model is currently only able to process and generate text.

OpenAI is now looking for investors to invest in its development, which would value the company at around 150 billion US dollars. According to US media, it would be one of the most expensive unlisted companies in the world.

Tech companies vying for further development of AI

Investors include Microsoft and chip giant Nvidia. Other names have also been circulating in the press, such as Apple, which is already using the start-up's technology in its new generative AI system. The private equity firm Thrive Capital or MGX, an investment fund backed by the United Arab Emirates, could also be among OpenAI's investors.

With ChatGPT, OpenAI started the wave of generative AI (production of content on simple request in colloquial language) at the end of 2022 and became the star of Silicon Valley. Since then, all major technology companies, from its main investor Microsoft to Google and Meta (Facebook, Instagram), have been competing to develop tools to help people in their everyday lives. These range from writing news to education and artistic creation, for example. But these "AI assistants" remain machines.