Efficiency gains are often cited as the reason to bring more AI into the workplace, though concerns over resulting job losses are never far behind. It might surprise you, then, to learn that the head of Google DeepMind, Demis Hassabis, actually thinks job cuts caused by AI are not the way forward.

Today’s wealth of machine learning tools means that anyone can vibe code—including this dog. Still, Hassabis thinks it’s premature to call time on software development as a profession. “I have no idea why people are going around talking with certainty about that,” Hassabis told Wired this week.

In that same conversation, he later went on to criticise companies looking to replace engineers with AI, saying, “I think it’s a lack of imagination—and a lack of understanding of what’s really going to happen.”

So, what future does Hassabis envision instead? Rather than efficiency gains at the cost of livelihoods, Hassabis instead asks why not keep the efficiency and the engineers; he says, “From my point of view, from DeepMind and Google’s point of view, if engineers are becoming three or four times more productive, then we just [want to] do three or four times more stuff.”

To me, that sounds a lot like ‘doing more stuff with the same amount of resources’ and not necessarily something ultimately sustainable, but then I’m not a software engineer. “I have a million ideas, from lab drug discovery to game design,” Hassabis continues, “I’d love to have some free engineers to go and do those kinds of things.”

You’ll recall that many moons ago, Hassabis worked in the games industry, spending a number of years at Lionhead Studios before moving on to found Elixir Studios in 1998. Hassabis designed and directed 2003’s Republic: The Revolution, and also worked on 2004’s Evil Genius, though the studio closed in 2005.

Hassabis also reflected on the subject of games and how today’s AI has yet to create a killer title without human help. He said, “I think there’s something missing.”

Last year, Google DeepMind’s own Genie 3 could generate an interactive world that ran at 720p, 24fps, and only remembered what you did for one minute. The model has come a ways since then, with the public preview version now simulating physics and generating the path in front of your player avatar in real time.

That said, The Verge noted in their hands-on that Genie 3’s overall output remains “much worse than an actual handcrafted video game or interactive experience.” Not to bang on about the nebulous concept of ‘human creativity’, but I think it’s clear that ‘something missing’ may never be in the grasp of a generative AI model’s ability.