The chief executive of Arm has said he worries about humans losing control of AI and that overrides are needed to shut down systems in an emergency.
When asked what keeps him up at night in an interview with Bloomberg, Rene Haas said: “The thing I worry about most is humans losing capability [over the machines].
“You need some override, some backdoor, some way that the system can be shut down.”
Cambridge-based Arm is a key supplier in the booming AI market. Arm’s processor architecture designs are licensed by many of the world’s chip firms, including Nvidia, whose graphics processing units are widely used to power AI applications.
Haas believes AI will “find its way into everything that we do, and every aspect of how we work, live, play. It’s going to change everything over the next five to 10 years.”
The Arm CEO’s comments come amid growing political concern over AI safety. The UK held an international summit in November in which 28 nations agreed to a shared understanding of AI risks.
The European Union last week agreed to a landmark provisional deal to govern the use of AI.
Arm listed in New York with a $54.5bn valuation in September. The blockbuster IPO came after failed lobbying attempts from the UK government for a dual listing in London.
Shares in Arm are up by nearly 11% since its debut on the Nasdaq. Arm owner SoftBank, the Japanese tech conglomerate, had previously attempted to sell the chip designer to Nvidia before the deal was ultimately scrapped due to regulatory hurdles.
Last month Arm posted record revenues of $806m amid the AI boom. However, Arm’s growth has not made the company immune to tech layoffs. In the UK, the firm has shed jobs equivalent to around 40% of the roles it created as part of a 2016 pledge by SoftBank to the British government.
Haas cited access to talent in the UK as a concern.
“We were born here, we intend to stay here,” he said. “Please make it very easy for us to attract world-class talent and attract engineers to come and work for Arm.”
Last month, Microsoft unveiled a new in-house chip based on Arm architecture to run general-purpose compute workloads in the cloud.