Thanks Dan, I was about to tell you that I meant that the null hypothesis can't be rejected,[1] but then I asked Opus to confirm. Opus says that, while all of the metrics I posted here do indeed show a nonsignificant difference, age-adjusted post-chatGPT YC companies are worth 1.5x more on a log scale, and that this is significant at p = 0.001 (which survives corrections for multiple comparisons).
I've changed the title to say that they aren't growing "much" faster, which hopefully conveys the accurate takeaway.
I tried to say this informally in the article: "YC companies are arguably growing a bit faster than pre-2023 companies (including AfterQuery, YC’s fastest-ever unicorn), but the difference isn’t large relative to the underlying variance"
why, if AI truly makes most developers 10x more productive and transforms the unit economics of customers, are the valuations of a cohort of prestigious "AI-enabled" startups round about their non-AI enabled 2019 peers?
Is this true? In my previous post I noted that the YC data was out of step with other data sets like Stripe's, but I haven't found a great dataset to answer this conclusively.
I also worry about the converse: because various AI safety people are genuinely well-intentioned, some people assume that their work must be good for the world.
Sadly, good people do indeed sometimes do bad things, and we are well to remember that fact.
I am reminded of the trend where people would post a university's "everyone belongs here" inclusion statement next to the rejection letter they got from that university.
More seriously: I often talk to people who overestimate the extent to which AI safety "experts" have things under control, and also those with imposter syndrome. I worry that the alternatives you list don't really work well for this audience, and I'm curious if you have other suggestions (or disagree with this claim)?
I would be interested in a version of this contest attempting to answer something like "If you believe Vasco that donations to SWP are better than torture, should you also believe that AI alignment is better than misalignment?"
I think that is what Richard Chappell is referring to when he mentions "radical skeptics" here. But I would be interested in a version of his post which defends the claim that EAs are reasonably justified in making the trade-offs we currently make (even if a radical skeptic would not be convinced of this defense).
Thanks for starting this group!
Thanks Dan, I was about to tell you that I meant that the null hypothesis can't be rejected,[1] but then I asked Opus to confirm. Opus says that, while all of the metrics I posted here do indeed show a nonsignificant difference, age-adjusted post-chatGPT YC companies are worth 1.5x more on a log scale, and that this is significant at p = 0.001 (which survives corrections for multiple comparisons).
I've changed the title to say that they aren't growing "much" faster, which hopefully conveys the accurate takeaway.
I tried to say this informally in the article: "YC companies are arguably growing a bit faster than pre-2023 companies (including AfterQuery, YC’s fastest-ever unicorn), but the difference isn’t large relative to the underlying variance"
Thanks for writing this Geoffrey, it seems important.
I think this is true, with some exceptions (e.g. MIRI)
Is this true? In my previous post I noted that the YC data was out of step with other data sets like Stripe's, but I haven't found a great dataset to answer this conclusively.
Congrats on the new role, Caroline!
I also worry about the converse: because various AI safety people are genuinely well-intentioned, some people assume that their work must be good for the world.
Sadly, good people do indeed sometimes do bad things, and we are well to remember that fact.
I am reminded of the trend where people would post a university's "everyone belongs here" inclusion statement next to the rejection letter they got from that university.
More seriously: I often talk to people who overestimate the extent to which AI safety "experts" have things under control, and also those with imposter syndrome. I worry that the alternatives you list don't really work well for this audience, and I'm curious if you have other suggestions (or disagree with this claim)?
Thanks Vasco, this is helpful.
I would be interested in a version of this contest attempting to answer something like "If you believe Vasco that donations to SWP are better than torture, should you also believe that AI alignment is better than misalignment?"
I think that is what Richard Chappell is referring to when he mentions "radical skeptics" here. But I would be interested in a version of his post which defends the claim that EAs are reasonably justified in making the trade-offs we currently make (even if a radical skeptic would not be convinced of this defense).
DiGiovanni states:
A single counterexample of a pair of actions where one is c-preferred over the other would suffice to disprove his claim.
It seems noteworthy that none of the solutions attempted this.