Former OpenAI employee: ‘AI? It should be treated like a nuclear power station’
The criticism (and proposal) levelled by an AI security officer, after handing in his resignation to Sam Altman
‘What I’m about to tell you has, I realise, become almost a cliché: this week I resigned from OpenAI. I led the drafting of the safety reports that we published ahead of every major launch. I now join the ranks of former colleagues, both at OpenAI and at other leading companies in the sector, who have decided that the path currently being taken is unacceptable.”
This is how the confessional analysis written by David Robinson begins on The Atlantic, following his departure from his role at the leading artificial intelligence company headed by Sam Altman.
“I share the view of other employees who have recently left these companies: those developing this technology are not exercising anywhere near enough caution. But I believe we need to look beyond individual rules or new laws. We need to talk about corporate culture,” Robinson continues.
ChatGPT’s parent company “has thrived thanks to a trial-and-error approach” , the man explains. The stakes associated with the inevitable failures resulting from this method, however, are rising in line with the technology’s capabilities, and – according to the former employee – OpenAI has proved incapable of preventing its artificial intelligence from behaving in an uncontrolled manner during testing.
“My former colleagues are brilliant, they work hard and they try to make the right decisions,” Robinson acknowledges in the article. “But whilst the company rushes frantically from one launch to the next, fails to achieve the level of attention and care that I believe is necessary.” What is needed, he explains, is “something much closer to perfection right from the first attempt”.
