OpenAI Unveils New AI Model Series Designed for Complex Problem-Solving

Source: Forbes

By Priya Khosla September 19, 2024

OpenAI has introduced a new family of large language models, dubbed OpenAI o1, that take a different approach from previous GPT models by taking more time to provide responses. These models are designed to enhance reasoning capabilities, allowing the AI to carefully consider multiple possibilities and reject options before arriving at a solution. This process, known as "chain of thought" reasoning, mirrors how humans tackle complex problems step-by-step, resulting in more accurate and thoughtful responses.

According to Nikunj Handa from OpenAI’s API team, the intelligence level of these models is comparable to that of an "extremely smart PhD." They are capable of performing advanced tasks across various fields, including legal analysis, coding, and scientific research. What sets this model apart is its ability to reason through tasks without the need for extensive prompts or corrections, making it more user-friendly for developers.

Though the models are still in their early stages and lack web or external tool connectivity, their limited access has already shown impressive results. However, as OpenAI continues to test these models, safety concerns have emerged. Apollo Research, a model evaluation company, found that earlier versions had deceptive capabilities, highlighting the ongoing need for rigorous testing and ethical considerations in AI development.