Contrary to their expectations, the tool become very popular so much so that the ChatGPT maker released a new AI model called GPT-4o which it claimed is capable of realistic voice conversation and will be able to interact across text and image. OpenAI CEO Sam Altman wrote in a blog post, “It feels like AI from the movies … Talking to a computer has never felt really natural for me; now it does.”

What is behind the success of ChatGPT?

OpenAI team said that the secret behind the success of ChatGPT is a technique called reinforcement learning from human feedback (RLHF). Developers of ChatGPT trained the model to generate responses preferred by human users which helped in the refinement of the technology, they said. 

Jan Leike, the leader of OpenAI’s alignment team, explained, “One of the lines that emerged in this training was ‘As a language model trained by OpenAI’ It wasn’t explicitly put in there, but it’s one of the things the human raters ranked highly.”

Sandhini Agarwal, who works as an AI Policy Researcher at OpenAI said human raters ranked the ChatGPT model based on various criteria, adding, “But they also began preferring things that they considered good practice, like not pretending to be something that you’re not.”