San Francisco - OpenAI's ChatGPT took the world by storm since its launch in November 2022. The San Francisco-based artificial-intelligence company was caught off guard by the popularity of the chatbot, which has become one of the most popular internet apps ever.

The firm has been working to catch up and capitalize on its success, according to an interview with with four people who helped build the technology that appeared in Technology review.

Jan Leike, the leader of OpenAI’s alignment team, said the success of ChatGPT has been "overwhelming" and that the team has been "trying to catch up."

John Schulman, a co-founder of OpenAI, said he was surprised by the chatbot's "mainstream popularity," which he did not expect to reach such levels.

ChatGPT is a fine-tuned version of GPT-3.5, a family of large language models that OpenAI released months before the chatbot.

The initial version called GPT-1 (Generative Pre-trained Transformer), was released in 2018. It was a significant improvement over previous language models, but still had limitations in its ability to understand context and produce coherent responses.

OpenAI continued to refine and improve the model, releasing GPT-2 in 2019 and GPT-3 in 2020. These models were capable of producing remarkably human-like responses and had a wide range of applications, from language translation to chatbots like ChatGPT.

Liam Fedus, a scientist at OpenAI who worked on ChatGPT, said the conversational data had a big positive impact on the chatbot.

Since its release, OpenAI has updated ChatGPT several times, using a technique called adversarial training to stop users from tricking the chatbot into behaving badly.

OpenAI has also signed a multibillion-dollar deal with Microsoft and announced an alliance with Bain, a global management consulting firm. The buzz around ChatGPT has set off a gold rush around large language models, with companies and investors worldwide getting into the action.

Sandhini Agarwal, who works on policy at OpenAI, said the team was surprised by how much people began using the chatbot.

"We work on these models so much, we forget how surprising they can be for the outside world sometimes," she said. Leike said she would love to understand better what's driving the virality of ChatGPT. "Like, honestly, we don’t understand. We don’t know," she said.

Despite the team's puzzlement, OpenAI is working to push the technology forward, watching how millions of people are using it and trying to fix the worst problems as they come up.

As ChatGPT continues to captivate users around the world, OpenAI is poised to remain at the forefront of AI development.