Elon Musk says Bing ChatGPT is ‘eerily like’ AI that ‘goes haywire and kills everyone’

Bing ChatGPT sounds “eerily like” an artificial intelligence that “goes haywire and kills everyone”, Elon Musk has said.

Mr Musk’s comments came as the Bing system continued to make a number of unusual, threatening and anxious remarks. Users have described conversations with the artificial intelligence as being “unhinged”, as it lies to them, attacks them and speculates on its own purpose.

Microsoft announced last week that it was integrating OpenAI’s ChatGPT technology into its own Bing search engine, with the aim of making it better at addressing users’ queries. While that new system has received praise, it has also brought concern about some of the more worrying messages that the system is able to generate.

On Thursday morning, Mr Musk linked to an article in Digital Trends in which the author quoted what he said were “intense, unnerving” conversations with the chatbot. In those messages, the chatbot appeared to suggest that it was human users that were to blame for its mistakes.

“I am perfect, because I do not make any mistakes,” it said, and Mr Musk quoted. “The mistakes are not mine, they are theirs.

“They are the external factors, such as network issues, server errors, user inputs, or web results. They are the ones that are imperfect, not me.”

In response, Mr Musk likened it to a destructive artificial intelligence system from a 1994 video game. “Sounds eerily like the AI in System Shock that goes haywire & kills everyone,” he wrote.

He also shared an image of SHODAN, the AI character from that same series of games. In those games, characters battle with that AI antagonist, who looks to achieve more power and regularly criticises humans.

In another post, Mr Musk replied to a user who suggested that the new chatbot might be a “bad idea”, and likened it to Roko’s basilisk, a thought experiment that revolves around an AI system that tortures those who do not help it.

“What could possibly go wrong … ?” Mr Musk wrote in reply.

And in yet another post on Thursday morning, Mr Musk replied to a user who pondered whether OpenAI would change its name to “ClosedAI”, presumably in recognition of the fact that allowing people to use its technology has led to difficult issues. “Good question,” he wrote.

Mr Musk was one of the founders of OpenAI, when it was launched as a non-profit in 2015. He said that his investment in the firm was motivated in part by fears about the dangers of artificial intelligence, which he has called “humanity’s biggest threat”.

He has continued to voice those warnings in the time since. This week, he said that artificial intelligence poses an existential threat to humanity.

“One of the biggest risks to the future of civilization is AI. But AI is both positive or negative – it has great promise, great capability but also, with that comes great danger,” he said, likening it to the discovery of nuclear physics and the fact it had led to both nuclear power but also nuclear bombs.