Genius may stay human.
But AI could become the lens that helps us *see sideways* more frequently (more bombastically!)

Will it ever happen? Quote tweet or reply. Would love your take.
(Full 20-min read coming soon — ping if you'd like it early.)

- Autonomous literature reviewers (ditto)
- Discovery scaffolds that suggest paths humans wouldn’t consider (let's see)

New cognition, symbiotic genius.
---
Maybe the first AI Nobel isn’t for thinking *like* a genius.
Maybe it’s about enabling genius in *entirely new ways*.

What kind of scientific discovery do you think AI might make first?
Medicine? Materials? Theoretical physics? Something stranger?
---
Maybe AI doesn’t replace genius —
But becomes the *substrate* for it.
Imagine:
- Hypothesis generators (was using this morning)

@DarioAmodei (Anthropic) suggests we stop anthropomorphizing genius.
> What if AI’s path to discovery isn’t like ours?
> What if it’s combinatorial, emergent, not wierd — but real?

darioamodei.com/machines-of...

but even autocomplete can surprise us.
> “Every once in a while it completes a thought better than I would have.”

Still — completion is not creation.
Creativity means framing the *right* question, not just answering one.

www.youtube.com/watch?v=bZQ...

Scaling alone doesn’t equal originality.

thomwolf.io/blog/scient...

Have you seen models do something *surprisingly original*? Drop links — I’d love to see edge cases.
---
@karpathy has said we treat LLMs like “autocomplete on steroids,”

Can scaled systems like AI cross that line?
---
Thomas Wolf @Thom_Wolf (Co-founder Hugging Face) questioned the “AI will win a Nobel” claim.
Rightly so.
> “You’ve gotta give these models something *new* to do — or they’ll just sit there wasting GPUs.”

'Innovation' — @naval maintains — '...is art. It can't be scaled.'
Genius is inefficient, nonlinear, often irreverent.
It doesn’t come from committees or consensus.
It comes from intuition, boldness, luck — and usually *outside the frame*.

x.com/naval/statu...