DeepMind execs talk about the importance of interdisciplinary research and how AI could help solve tough scientific problems, like predicting shapes of proteins
Greg Williams / WIRED UK : Tweets: @wireduk and @wireduk Tweets: @wireduk : DeepMind's AI has mastered Go and defeated chess grandmasters, but what's its real mission? CEO and co-founder @demishassabis speaks exclusively to WIRED Editor-in-Chief @GregWilliams718 about its ambitious mission to “solve intelligence” https://www.wired.co.uk/... @wireduk : DeepMind has conquered Go, StarCraft and Atari, now it's turning its focus elsewhere. The company's AI is now being trained to unlock the mysteries of protein folding, one of the building blocks of organic life https://www.wired.co.uk/...
Context & Ripple Effects
By 2019, DeepMind had exhausted the benchmark value of games — Go, StarCraft and Atari were demonstrations, not a mission. This WIRED interview with Demis Hassabis is the pivot being argued in public: 'solving intelligence' reframed as interdisciplinary science, with protein folding as the flagship problem.
The argument had already been tested: [[a:936255|AlphaFold won an international protein-structure prediction competition in December 2018, beating 97 entrants]]. What this interview adds is the strategic framing — and the later coverage shows it held, from the Nobel Prize Hassabis calls 'a watershed moment for AI' to Isomorphic Labs and drug programs.
First-order effects
- DeepMind's public identity shifts from game-playing champion to scientific instrument maker, with protein folding replacing Go and StarCraft as the proof-of-capability problem.
- Hassabis gains a platform to argue that interdisciplinary teams — not pure computer science — are what make hard problems like protein structure tractable for AI.
Second-order effects
- Scientific credibility becomes a competitive asset inside Alphabet: later reporting shows executives juggling rapid AI infusion into Google products against long-horizon research like AlphaFold, forcing an explicit trade-off between revenue pressure and mission prestige.
- Biology and drug discovery become adjacent markets for AI labs rather than academic curiosities — the path that leads to Isomorphic Labs and its drug programs.
Third-order effects
- If the pattern holds, AI labs compete on scientific prizes and discoveries as legitimacy markers, not just benchmarks — a trajectory culminating in a Nobel being treated internally as a pursued goal rather than an accident.
- The 'solve intelligence' framing normalizes AI companies setting national-lab-scale scientific agendas, raising questions about who funds and governs discovery-grade compute when it sits inside commercial firms.
The trend: AI labs are converting game-playing dominance into scientific-discovery credibility, using problems like protein folding to justify long-horizon research over near-term product revenue.