Demis Hassabis says AI is taking coding and maths first not because they are easy, but because you can check the answer: Asked at Davos whether he still expected a 50% chance of human-level AI by the end of the decade, he said his timeline had not moved: "Yeah,... look, I think I'm still on the same kind of timeline. I think there has been remarkable progress." "But I think some areas of kind of engineering work, coding, or so you could say mathematics, are a little bit easier to see how they would be automated, partly because they're verifiable what the output is." "Some areas of natural science are much harder to do than that. You won't necessarily know if the chemical compound you've built or this prediction about physics is correct." "It may be, you may have to test it experimentally and that will all take longer." Hassabis runs Google DeepMind, which builds in both the coding and the science domains he is ranking here. - Demis Hassabis (Demis Hassabis), CEO of Google DeepMind (Google DeepMind), at the World Economic Forum in Davos, January 2026.show more

code and maths have cheap feedback loops. nature doesn’t. once AI can generate hypotheses faster than we can experimentally test them, labs, robotics, measurement and the physical world become the constraint. AI can make thinking cheap. reality still has latenc

The verifiability point is huge—when outputs can be checked, adoption can move faster. The harder question is how teams redesign work around that leverage without losing judgment.

verification makes automation low‑hanging fruit, science will lag until we trust experimental outputs

Se crearán puestos de científicos para probar los experimentos de la IA

How old is this video

only possible with wetware.

编码先落地，部分原因是输出结果是可验证的，科学得等实验闭环

Unchecked human intelligence dies in unchecked arrogance.

@demishassabis in clinical medicine you simply have to ask/verify the patient to check the answer. a nurse under ai guidance could do it as well as a doctor. on the republican principle of "do harm now save more later" ai should arrive in clinical practice immediately. no excuse

Worth adding: even in verifiable domains this isn't a done deal.

I agree with Demis. At least 10 years. But it is always moving. Next year it will 10 years. Etc. There is still much to discover. I wonder if university should consider pushing students into new areas. There are many.

"This Is embarassing"

This is 10 freaking months ago

This is the actual bottleneck for everything AI touches next, not compute, not data. Reality doesn't grade your homework the way a compiler does. Coding and math are the easy 20%. The other 80% still has to wait in line behind a lab bench.

Makes sense. In my SaaS work, anything with clean pass/fail gets automated fast. Anything needing human feel still eats my weekends.