We didn't think ChatGPT was very good when we released it, OpenAI's chief scientist says

• OpenAI's chief scientist told MIT Technology Review that he initially thought ChatGPT wasn't very good.
• Ilya Sutskever said he worried people would find the chatbot "boring."

OpenAI's chief scientist admitted that he didn't think ChatGPT was very good before its record-breaking launch.

Ilya Sutskever, OpenAI's cofounder and chief scientist, told MIT Technology Review that he was initially unimpressed by the chatbot's inability to answer questions accurately, and was taken by surprise by its explosive popularity.

"I will admit, to my slight embarrassment … when we made ChatGPT, I didn't know if it was any good," said Sutskever.

"When you asked it a factual question, it gave you a wrong answer. I thought it was going to be so unimpressive that people would say, 'Why are you doing this? This is so boring!'" he added.

Sutskever told MIT tech review that the true draw of ChatGPT was convenience, not accuracy, comparing the first time people used it to a "spiritual experience."

"That first time experience is what hooked people," he said. "You go, 'Oh my God, this computer seems to understand.'"

ChatGPT was an instant record-breaking success when it launched late last year, reaching 100 million users in just two months and sparking a global AI gold rush as tech companies scrambled to build their own models.

It seems that no one was more surprised by that than OpenAI itself. OpenAI president and cofounder Greg Brockman told Forbes earlier this year that the startup's staff did not think the chatbot was particularly useful and were taken aback by its sudden popularity.

The AI startup was reportedly on the fence about when to release ChatGPT, and was building alternative models before ultimately deciding to launch it.

OpenAI did not immediately respond to a request for comment from Insider, made outside normal working hours.