Originally posted on LinkedIn on June 30, 2026
“Solve intelligence, then use it to solve everything else.”
That was the ambition behind DeepMind: build AI, then apply it to humanity’s hardest problems. It sounded grandiose at the time. But with AlphaFold and Demis Hassabis’ Nobel Prize in Chemistry, that vision already looks less like science fiction than it once did.
At the same time, AI risks dominate much of the public debate — and for good reasons.
AI can amplify misinformation, deepfakes and online slop. It can help bad actors with surveillance, hacking or biological threats. It may disrupt jobs, increase inequality, exploit creative work, consume enormous amounts of energy, and in the most extreme scenarios even become an existential risk if systems become more capable than we can control.
These risks are real. They deserve serious attention.
But a proper risk-benefit analysis also has to include the upside.
If we build AI systems that are more capable than humans across most cognitive domains, the “solve everything else” part becomes more than a slogan. Intelligence is the most powerful general-purpose tool we know.
Advanced AI could massively accelerate science, medicine, engineering, education, energy and economic productivity. It could help us design better drugs, understand biology, reduce poverty, improve education and perhaps prevent or cure many diseases that currently destroy lives.
To many people, this still sounds like science fiction, or at least something far beyond our lifetime.
But AI and robotics are progressing extremely quickly. And if self-improving AI becomes possible, progress could accelerate even further. That means the positive scenario is not necessarily centuries away. It could plausibly happen within our generation.
Of course, many things could slow this down: technical bottlenecks, economic crises, geopolitical conflict, another pandemic, or failure to manage AI’s risks.
But that is exactly why we need a serious, balanced discussion.
AI may carry unusually large risks. But it may also offer one of the largest upsides in human history. That upside belongs in the risk-benefit analysis too.
Back to AI experiments and musings