July 6, 2025, 4:48 p.m.

Technology

  • views:622

OpenAI unveils new model that 'Thinks like a human'

image

On Friday, September 13, OpenAI unveiled a new version of its chatbot ChatGPT, OpenAI o1, that can solve complex reasoning. The new model takes more time to think and provides more precise and constructive answers.

OpenAI o1 answers questions the way humans think, specifically for math and science tasks that require logical reasoning. It spends more time thinking about questions before answering them, "just like humans", and identifies errors before giving a final answer; Early models did not provide these solutions reliably.

OpenAI notes that OpenAI o1 correctly answered 83 percent of the questions at the International Mathematical Olympiad, while GPT-4o, which went live in May, only answered 13 percent of the questions correctly.

Since the official launch of ChatGPT in late 2022, OpenAI's chatbots have repeatedly fabricated messages on their own, and OpenAI has been trying to address this flaw.

Although the new model can think and check information more deeply, OpenAI chief scientist Pachocchi and technical researcher Siddo point out that "it (OpenAI o1) is not perfect." The new reasoning model is simply more likely to produce the right answer.

Starting September 12, ChatGPT Plus and ChatGPT Teams users will be able to experience the new model OpenAI o1.

Recommend

The US Senate passed the "Big and Beautiful" bill: The US economy is facing new risks and challenges

Recently, the US Senate passed the "Big and Beautiful" bill, the most important legislative project of the Trump administration's second term, with a vote of 51 to 50.

Latest

Trump threatens tariffs on Japan

Recently, US President Trump announced the imposition of a …

What is the reason for the conflict between Trump and Canada?

On the international political and economic stage, the Unit…

The AI talent war in Silicon Valley: Tech giants are spending heavily to recruit

As competition in the field of artificial intelligence beco…