One way of viewing takeoff speed disagreements within the safety community is: most people agree that growth will eventually be explosively fast, and so the question is "how big an impact does AI have prior to explosive growth?" We could quantify this by looking at the economic impact of AI systems prior to the point when AI is powerful enough to double output each year.

(We could also quantify it via growth dynamics, but I want to try to get some kind of evidence further in advance which requires looking at AI in particular---on both views, AI only has a large impact on total output fairly close to the singularity.)

The "slow takeoff" view is that general AI systems will grow to tens of trillions of dollars a year of revenue in the years prior to explosive growth, and so by the time they have automated the entire economy it will look like a natural extension of the prior trend.

(Cost might be a more robust measure than revenue, especially if AI output is primarily reinvested by large tech companies, and especially on a slow takeoff view with a relatively competitive market for compute driven primarily by investors. Revenue itself is very sensitive to unimportant accounting questions, like what transactions occur within a firm vs between firms.)

The fast takeoff view is that the pre-takeoff impact of general AI will be... smaller. I don't know exactly how small, but let's say somewhere between $10 million and $10 trillion, spanning 6 orders of magnitude. (This reflects a low end that's like "ten people in a basement" and a high end that's just a bit shy of the slow takeoff view.)

It seems like growth in AI has already been large enough to provide big updates in this discussion. I'd guess total revenue from general and giant deep learning systems^[1] will probably be around $1B in 2023 (and perhaps much higher if there is a lot of stuff I don't know about). It also looks on track to grow to $10 billion over the next 2-3 years if not faster. It seems easy to see how the current technology could do $10-100 billion per year of revenue with zero regulatory changes or further technology progress.

If the "fast takeoff" prediction about pre-takeoff AGI impact was uniformly distributed between $10 million and $10 trillion, that would mean we've already run through 1/3 of the probability mass and so should have made a factor of 3/2 update for slow takeoff. By the time we get to $100 billion/year of impact we will have run through 2/3 of the probability mass.

My sense is that fast takeoff proponents are not owning the predictions of their view and now acknowledging this fact as evidence. Eliezer in particular has frustrated me on this point. These observations are consistent with fast takeoff (which as far as I can tell practically just says "anything goes" until right at the end), but that they are basically implied by the slow takeoff model.

(One might object to this perspective by saying that there is no analogous form of evidence that could ever support fast takeoff. But I think that's just how it is---if Alice thinks that the world will end at t=1 and Bob thinks the world will end at a uniformly random time between t=0 and t=1, then Bob's view will just inevitably become less likely as time passes without the world ending.)

1. ^{^}

   It's unclear if you should treat "general and giant deep learning systems" as the relevant category, though I think it's probably right if you want to evaluate the views of actual slow takeoff proponents in the safety community. And more broadly, within that community I expect a lot of agreement that these systems are general intelligences that are the most relevant comparison point for transformative AI.