OpenAI Introduces New AI Model Capable of Human-Like Reasoning

OpenAI has launched a groundbreaking artificial intelligence model, the first in its new OpenAI o1 series, which promises to tackle complex problems in science, coding, and math with human-like reasoning abilities.

Released as a preview on Thursday, the o1 model is set to receive regular updates and improvements. It will gradually become accessible to most ChatGPT users.

“We’ve trained these models to think through problems more thoroughly before responding, much like a human would,” OpenAI stated on its website. “Through this training, they learn to refine their thought processes, experiment with different strategies, and recognize their mistakes.”

OpenAI highlighted the model's potential applications, such as assisting healthcare researchers in annotating cell sequencing data and helping physicists generate complex mathematical formulas for quantum optics.

Noam Brown, a research scientist at OpenAI, emphasized the model's capabilities. “While the current o1 model thinks for seconds, future versions aim to think for hours, days, or even weeks. This will increase inference costs, but the potential benefits, such as breakthroughs in cancer treatment or battery technology, are worth it,” he posted on X.

The energy demands of advanced AI systems, including the new o1 model, were a topic of discussion at a meeting between senior White House officials and top US tech executives, including OpenAI CEO Sam Altman, Google’s Ruth Porat, and Anthropic CEO Dario Amodei. While AI technology holds promise for solving significant issues like cancer and climate change, it also presents challenges, such as the substantial electricity requirements that could exacerbate global warming.

The new OpenAI model currently lacks some features that make ChatGPT widely useful, such as web browsing and file uploads. However, it represents a significant advancement in complex reasoning tasks.

In tests, the o1 model performed on par with PhD students in physics, chemistry, and biology. It also excelled in a qualifying exam for the International Mathematics Olympiad, solving 83% of the problems correctly.