Sam Altman Compares OpenAI’s o1 Model to GPT-2 for Reasoning

In a recent interview, OpenAI CEO Sam Altman described the company’s latest AI model, o1, as being at the ‘GPT-2 stage’ of reasoning development. Altman explained, “I think of this as like we’re at the GPT-2 stage of these new kinds of reasoning models.” He emphasised that while the model is still early in its development, significant improvements are expected in the coming months.

Altman said that users will notice o1 rapidly improving as OpenAI moves from the o1-preview model to the full release. “Even in the coming months, you'll see it get a lot better as we move from o1-preview to o1, which we shared some metrics for in our launch blog post,” he said.

"You will see it reach the GPT-4 equivalent over the coming years."

He highlighted the steep improvement curve of the o1 model, comparing it to previous jumps in AI capabilities. “It's a very significant step forward, and one of the fun things about these new paradigms is that the improvement curve is really steep,” Altman noted.

Altman expressed confidence in the model's future performance, stating that tasks it currently struggles with will be solvable in a few months. "The things that the model just can't solve right now, in a few months they'll be able to solve, and a few months after that, even more," he added.

He also predicted that users will discover new ways to leverage the o1 model, similar to the gradual adoption of GPT-3.5 and the eventual success of ChatGPT. "It took people a while to figure out how to use ChatGPT, and it took us a while to build," Altman said, hinting at similar potential for o1's applications.

Altman added that it will take users some time to fully grasp how to utilise ChatGPT with the updated model. “It'll  take users a while to figure out how to use it, and this is quite different from the GPT models.”