AI should be treated as a new species with values that may be alien to humanity, a former Google DeepMind safety researcher has warned.
Josh Engels, who has left the company’s safety team to join the non-profit organisation METR, said rapidly advancing systems could pose “very large risks to humanity” if people failed to retain control over their development.
“I think in some ways it’s better to think of it not as a technology but as a second species that we are almost giving birth to,” he told the BBC.
Mr Engels said humans could soon share the planet with entities “much, much smarter than us”, adding: “I’m worried that if we don’t have the ability to ensure they have our best interests in mind, there’s no guarantee humans will keep control of the future.”
He said an artificial species could disempower humanity in several ways, not necessarily because it would develop hatred towards people.
“It’s not that these systems will hate humanity,” he said. “It’s just that they have alien motivations that don’t have humans front and centre.”
Mr Engels added that there was currently no reliable way to shape the systems being developed because their values and motivations were not understood.
Warnings over AI development
His warning follows comments from researcher Jacob Coxon, who left leading AI company Anthropic and accused it and rival OpenAI of “gambling with our lives”.
Mr Coxon told the BBC: “If we don’t slow down at the current rate of progress, there is a strong chance that we could all die in the immediate future.”
Other figures in the industry said last week that there was a 10 per cent chance of AI killing all humans within the next decade. The bosses of leading AI firms subsequently called for the pace of development to slow.
The warnings come as the King prepares to host a summit of key AI figures at Dumfries House in Ayrshire this week. He is expected to urge them to use the technology “for the good of humanity”.
A royal source said the King hoped the discussions would improve understanding of the responsibilities involved in developing AI, while ensuring technological progress remained focused on human benefit.
MPs and peers are also due to call for the Government to introduce a new AI Bill addressing threats to human rights.
The Joint Committee on Human Rights has proposed a regulatory system imposing greater obligations on higher-risk systems. It also wants some uses of AI banned outright, including profiling intended to predict whether an individual is likely to commit a crime.
The committee said the provision of highly powerful systems, including Artificial General Intelligence and Artificial Superintelligence, should be prohibited where they risk causing widespread and serious harm or could evade effective human control.
