Humans in the Loop

Demis Hassabis, CEO of Google DeepMind, paints an inspiring vision of AI as a catalyst for human flourishing. In his interview with Steven Levy, Editor at Large at Wired in June 2025 Hassabis laid out some of his views on the world of work, and the future of humanity in a post AGI world.

His forecasts, ranging from a productivity boom to solving planetary-scale problems, feels a tad like a TED Talk version of a sci-fi utopia: hopeful, sleek, and, let’s be honest, a touch speculative. To give him some flowers, Hassabis doesn’t completely ignore the risks. He admits the scale of the change will likely dwarf past technological shifts. But while his optimism is admirable, there’s something almost too frictionless in his narrative, especially when tougher questions are glossed over or deferred. It could have been the edit, but geopolitical challenges and disruption coming were brushed over as things we need to work through. True. But they are as important to the future as water is to flowers.

The idea that AI will supercharge productivity and make people “a little bit superhuman” is tantalising. Tools like AlphaFold have already demonstrated AI’s potential to accelerate scientific discovery. I do think saying we’re entering a “golden era” of creativity sounds more like a venture pitch than an analysis grounded in the messy, often political, realities of the world of work. It is arguably reductive to an extent. It assumes equitable access to tools, supportive governance, and benign corporate incentives, all of which are far from guaranteed. In fact, I can guarantee it won’t be that way. Unless there is money in it based on the current direction.

Take jobs. Hassabis follows a familiar Tech CEO script: new technologies displace old roles but create better ones. Perhaps. I actually think this is modestly true. That said, history shows transitions are uneven. The internet did create new work, but it also gutted industries and deepened inequalities. And this time, the speed and breadth of AI’s capabilities might leave entire sectors without a clear on-ramp to “reinvented” work. Hassabis concedes this could be different. Then quickly sidesteps the question of what happens if it is. I do intend digging into this.

On the human front, Hassabis offers a compelling distinction: doctors might be augmented and partially replaced; nurses won’t, because of their empathy. It’s neat, digestible, and again reductive. Empathy isn’t exclusive to nursing, nor is diagnosis free of human nuance. Suggesting a clean line between ‘machine tasks’ and ‘human roles’ underestimates how often care, trust, and social context are entangled in decision-making. And more importantly what people want. A cuddle goes a long way, right?

It also skirts the messier challenge: what happens to societal cohesion when large swaths of work are automated and only some get to “10X” their output by becoming AI-native? No jobs. No economy. No society. At least as we know it today.

we’ll have these incredible tools that supercharge our productivity, make us really useful for creative tools, and actually almost make us a little bit superhuman – Demis Hassabis

Hassabis’s advice to graduates is just and to the point: get fluent in the tools, learn to prompt, and understand systems. He says get fluent in the systems but that assumes all can be experts in using the new tech, or access it. The gap between experts in LLMs and those working precariously or outside the digital economy may widen. In framing success as a question of individual savvy, his outlook understates the societal scaffolding of policy, education, inequality, nationalism and more that determines who gets to thrive.

His long-term vision of “radical abundance” and galactic exploration is seductive sci-fi. If I was advising him I’d have recommended he stays out of the Star Trek vision just yet. He predicts AGI fully enabled in 5 years. That’s what people care about. His utopia felt a deeply ironic act in celebrating the power of AGI to solve scarcity while offering few specifics on how those gains will be distributed. Abundance isn’t really a technical problem, it’s a political one. And the leap from technical capability to moral progress remains aspirational, not assured. It’s not for Big Tech to solve this. But maybe it is not for Big Tech to exacerbate it?

Ultimately, I look at Hassabis’s vision is best read as a provocation: this is what could be possible. But if we want the future he describes to work for more than just the most “productive” and privileged, we’ll need deeper answers than tech optimism alone can provide.

It is wrong to blame Hassabis for the holes I poke above. It is not is not his problem to solve alone. The reality is these tech organisations will keep inventing and keep seeking to profit. And they are driven by the geopolitical imperatives, see my points on Hamburg Declaration versus the America AI Plan for how these can influence things. Google knows second is last. However idealistic the Tech Bros may think they are, share price first and all that. Which means there’s a continuation of the foundational societal exposures happening in real-time. It needs a movement, fast.

I thought how to summarise my thoughts on the whole piece but then I decided this could be a good way to end.

Reading the comments on YouTube below the line was enlightening too, I loved this one from JLee4361: Steven Levy said something that struck me 14:20 and it’s a brilliant point. Demis isn’t trying to be a salesman here, but I think his quest to create this thing that will magically solve our problems shows a fundamental misunderstanding in how we currently operate as humans. Steven’s point is that we live in a world where we can have peace NOW.

James Stewart