Will AGI be achieved before AIs are able to smell?

Why we chose it

This is Moravec's paradox with money on it: the things that impress us — reasoning, essays, code — turned out to be the easy half of intelligence, while a capability every dog has remains unsolved hardware. The 67% for AGI-first is the crowd betting that curve continues: models keep leaping while machine olfaction stays a lab problem, because smell needs chemistry — physical molecules hitting physical sensors — and you cannot scale a nose with more GPUs.

The case for the other side is quietly good, though: digital-scent startups have already demonstrated molecule-to-description prediction and even "teleporting" a scent from one lab to another, a nose being, in the end, a finite sensor problem. Whereas AGI has the harder handicap of nobody agreeing what it is. That's the real joke the market runs on: one side of the race has a definition and no progress metric; the other has sensors, benchmarks — and no hype.