From GPT-2 to GPT-4 and Beyond: The Remarkable Evolution of AI According to OpenAI’s CEO

In a recent conversation with Tobias Lütke, CEO of Shopify, Sam Altman, CEO of OpenAI, discussed the evolution of OpenAI’s Generative Pretrained Transformer (GPT) models, from GPT-2 to GPT-4, and what the future holds for these AI models.

Altman highlighted the significant improvements and capabilities each new version of the GPT model brought. “GPT-2 in 2019, people tried it and said this is like a dead-end direction. Scale that up, GPT-3, people are like, ‘Oh, okay.’ GPT-3.5 was like finally usable, and GPT-4, people are building entire companies around,” Altman said.

The CEO of OpenAI also shared insights into the predictability of a model’s intelligence based on its training. “We have figured out a computer algorithm that can really, truly learn. No tricks, no sleight of hand. It can actually learn,” Altman explained. He compared this period of discovery to the era when people were discovering fundamental physics, suggesting that we are living through one of the most important periods of human discovery.

Altman also touched on the future of these models, stating that the utility they provide will continue to increase with future versions like GPT-5, GPT-6, and GPT-7. “GPT-5, 6, and 7, I think will continue in future years on this trajectory of really increasing the utility they can provide,” he said.

As AI continues to evolve, Altman’s insights provide a fascinating glimpse into the journey of AI development and the potential of future AI models. With each new version of the GPT model, we are witnessing a remarkable evolution of AI that is reshaping our world.