AI 2027

Feels reasonable in the first few paragraphs, then quickly starts reading like science fiction.

Would love to read a perspective examining "what is the slowest reasonable pace of development we could expect." This feels to me like the fastest (unreasonable) trajectory we could expect.

No one knows what will happen. But these thought experiments can be useful as a critical thinking practice.

The slowest is a sudden and permanent plateau, where all attempts at progress turn out to result in serious downsides that make them unworkable.

Like an exponentially growing compute requirement for negligible performance gains, on the scale of the energy consumption of small countries? Because that is where we are, right now.

The forecasts under "Research" are distributions, so you can compare the 10th percentile vs 90th percentile.

Their research is consistent with a similar story unfolding over 8-10 years instead of 2.

Even if this were true, it's not quite the end of the story is it? The hype itself creates lots of compute and to some extent the power needed to feed that compute, even if approximately zero of the hype pans out. So an interesting question becomes.. what happens with all the excess? Sure it probably gets gobbled up in crypto ponzi schemes, but I guess we can try to be optimistic. IDK, maybe we get to solve cancer and climate change anyway, not with fancy new AGI, but merely with some new ability to cheaply crunch numbers for boring old school ODEs.

If you described today's AI capabilities to someone from 3 years ago, that would also sound like science fiction. Extrapolate.

> Feels reasonable in the first few paragraphs, then quickly starts reading like science fiction.

That's kind of unavoidably what accelerating progress feels like.