OpenAI Launches New Super AI With 'Reasoning' Capabilities

On September 12, OpenAI launched a series of new artificial intelligence (AI) models that are capable of spending more time thinking to provide more accurate and beneficial responses to users.

The new models, known as OpenAI o1-Preview, are designed to tackle complex tasks and more difficult problems in science , coding, and mathematics — areas where previous models have often been criticized for failing to provide consistent answers.

OpenAI's new AI model has the ability to 'reason' and can even solve Olympic math problems. Photo: AFP

OpenAI o1-Preview was trained to refine its thought process, test different approaches, and detect errors before coming up with a final answer.

Sam Altman, CEO of OpenAI, has called the new models “capable of complex reasoning,” although they may still have shortcomings and limitations.

Accordingly, OpenAI is still working to improve the capabilities of AI models to overcome the "illusion" problem - a phenomenon when chatbots create convincing but inaccurate content.

OpenAI researcher Jerry Tworek said the new model would be less prone to the “illusion” problem, but it still doesn’t completely solve the problem.

OpenAI o1-Preview models are known to have successfully solved difficult tasks in physics, chemistry, and biology at the level of PhDs.

In particular, in the field of mathematics and coding, this OpenAI o1-Preview achieved an accuracy rate of up to 83% when solving the questions in the International Mathematical Olympiad entrance exam. This number is much higher than the 13% rate of the previous GPT-4o model.

Sam Altman, CEO of OpenAI, has described the new models as “capable of complex reasoning.” Photo: AFP

According to OpenAI, the new reasoning capabilities could help healthcare researchers annotate cell sequencing data, while the tool could help physicists develop complex formulas.

OpenAI also said the new AI models have passed rigorous jailbreak tests and are resistant to attempts to bypass security barriers.

Safety measures have also been stepped up, including recent agreements with the US and UK AI Safety Institute, which has been granted early access to these models for testing and evaluation.

Source: https://laodong.vn/cong-nghe/openai-ra-mat-sieu-ai-moi-voi-kha-nang-lap-luan-1393825.ldo