@megan_gafford 🏆
Feels like doing this kind of drawing creates a different, deeper category of memory.
Way beyond my skill level as an artist, so I'm just guessing..
@Matt_Hougan The bust phase will probably be short-lived (6-15mos). Not some multi-year thing
AI is too powerful, improvements are coming too fast and adoption/revenue will accelerate and eventually justify all the infrastructure and optimism
@Matt_Hougan This feels spot on
Revenue ramp goes more slowly than anticipated. Causes brief period of overcapacity + drop in mood (and valuations) but model capabilities keep improving and then adoption (and revenue) go parabolic and the party resumes.
That's my base case
@GarrisonLovely Some country will go all in on AGI as a replacement for human labor. And then other countries might be forced to do this in order to maintain sovereignty (even if people don't want it)
Overall the base incentives create real risk of AGI even if it's net negative for the majority
@GarrisonLovely 3. Governments, for national security, have Manhattan-type Projects aiming for AGI and the world has no idea. It'd be surprising if there are zero nation-state level efforts to do this
Once the recipe for AGI exists, it's hard to keep it locked down. Look at nukes.
@awiltschko Great insight. Feels spot on
Given this, what is your timeline for AGI (let's define it as at least expert level equivalence in 95% of domains)?
@deredleritt3r@aidan_mclau Also, distillation is a real concern. Ant says they solved it, but they need all frontier models to solve it
It's also an issue that every company with the capital can touch the (public) edge of the frontier. A lot of competition right now
@deredleritt3r@aidan_mclau What sectors beyond software does the revenue come from? I did my own digging on this and the math is tough imo (if they don't hit AGI)
Adoption by enterprise needs to happen. Very fast and at massive scale. Beyond coding.
I don't know if businesses will move that quickly
@DKThomp@dylan522p bcs my gut reaction is that this is the kind of wild, hyperbullish talk you get near the peak of the bubble
and I say this as someone who started in crypto at the beginning of 2013 and have ridden a few hype cycles
@DKThomp "extraordinary claims require extraordinary evidence" - Sagan
This is so wildly optimistic that it starts to undermine credibility in the absence of a very strongly supported path to that kind of revenue
I like @dylan522p but damn.. I hope follows up and explains the 'how'
@TheStalwart Two open questions:
1. Can they achieve coding/math-level capabilities in other fields quickly enough to keep the party going
2. Can they turn those capabilities into enough revenue quickly enough to keep the party going
@jgarzik@jon_stokes Yes, but for human nature
we too often don't do the work until we're forced to
We *could* use sota models to make core digital infrastructure bulletproof now, but that would require getting very serious, very soon
more likely we'll be forced to act after something goes wrong