Mustafa Suleyman, co-founder of DeepMind, has quit Google after seven years in which the pioneering British operation made breakthroughs in artificial intelligence but struggled to ensure its technology would be used ethically by its giant US owner.
Suleyman was one of three founders of DeepMind, which was set up in the UK in 2010 and was bought by Google in 2014 for $500m. On Thursday, he left the company to become a partner at Silicon Valley venture capital firm Greylock, after a sometimes controversial tenure that included complaints from staff of an aggressive management style.
The DeepMind division sought to operate at arm’s length from its owner, maintaining a research ethos that insulated it from commercial pressures and attempting to come up with a governance structure for its powerful AI.
DeepMind’s first attempt to create an independent AI ethics board to oversee its work was slow to take off, and the company later made an unsuccessful attempt to carve out a legal structure that would give it more independence inside Google.
Speaking in a recorded interview with Greylock partner Reid Hoffman that was released on Thursday, Suleyman said: “We made a lot of mistakes in the way that we attempted to set up the board, and I’m not sure that we can say it was definitely successful, but I do believe that radical experimentation is essential here.”
“We’ve experimented with different oversight boards, with different ethical charters, with types of research,” he added.
But he said of the tech industry’s broader efforts in this area: “I definitely feel that we haven’t really come close to cracking this nut of how we make technology platforms, software, and of course AI feel like it’s happening with people, and where people have significant influence in shaping how it arrives in their world and doesn’t just happen to people.”
Google itself has struggled with how to introduce checks and balances to its development of AI. Its plan for an independent AI ethics board foundered over staff protests at the make-up of the body, and the company’s two heads of AI ethics left the company amid recriminations over its handling of their work.
Suleyman lost some of his responsibilities after complaints from DeepMind workers about his management style, and he moved to take a position at Google’s headquarters in California.
Addressing the staff complaints, he said: “I really screwed up. I was very demanding and pretty relentless.” He added that he had set “pretty unreasonable expectations” that had led to “a very rough environment for some people. I remain very sorry about the impact that caused people and the hurt that people felt there.”
DeepMind, which is led by co-founder Demis Hassabis, cemented its reputation as one of the leaders in deep learning in 2016 when one of its systems beat the world champion at the board game Go.
Its 2020 research using AI to predict how proteins fold into different shapes was heralded as a breakthrough that could lead to advances in fields such as drug discovery.
