I noticed that the stakes aren't even in the bet (Holden can win 5K, Bryan can win 1K) - should we interpret that as Holden thinks there's a >1/6th chance (but <1/2 chance) that his side turns out True, or was there a non-odds-related factor that went into the bet sizing?
What if the quantity of goods produced has increased as holden predicts, and it is counted in official GDP figures, but AI has driven real prices so low that official GDP hasn't increased that much; would your "conceptual" GDP measure count this as a win for holden? (basically I'm wondering about the case where much of production is counted officially but just not adequately, due to extreme low prices, similar to many currently free or near-free internet services)
It is based on real output, not nominal. If your scenario resulted in mass deflation, then the huge amount of *real* production would indeed count as a win for Holder, so long as humans were interested in purchasing/consuming the stuff.
But if, e.g. it resulted in a 100,000% increase in zucchini production, but the world had no use for the increase after the first 5,000% increase, then the extra would not count as a win for Holder.
Consumer surplus like the free benefits consumers obtain from use of the free Internet is a different issue. I’m very curious about that point myself.
I agree that if prices were uniformly lowered (and the goods-basket remained fixed) then this would be properly handled by inflation-correction, but if for instance manufacturing prices were lowered by automation, but prices for (luxury) human-provided services remained high, it's less clear that inflation-adjustment would address this. Also, I don't understand why you see the case of cheap manufactured goods (from automation) as entirely different from the internet case, given that in both cases extremely low (or zero) real prices mask large impacts on consumer surplus.
All fun and good to see people put their money behind a view. But, who knows what AI will mean either in the short or long term; just way too soon to have enough info to really have good view of it. Hopefully, it will be of benefit to increased human flourishing since a ton of capital investment is being made that investors hope will provide them with returns commensurate with the risks related to the massive capital investment now under way and to continue over coming years.
I would be tempted to take Holden's side of the bet: an above average growth in productivity and thus GWP growth seems like it might have more than 20% probability. But I must agree with Bryan that the timelines are perhaps a bit too short for this to happen already.
Overall, I'm very happy to see that this bet doesn't have even odds, like many previous bets where Bryan's side was almost ridiculously likely to with more than 50% probability.
“Overall, I'm very happy to see that this bet doesn't have even odds, like many previous bets…”
Agreed.
At 50-50 Holder’s bet is dumb.
I’d personally put his chances of success at more like 10% - and a big fraction of that is the risk of war downside that might not even have much to do with AI - so I lean Bryan on this one. But the point is it’s not a slam dunk in either direction at these implied odds.
Incredibly funny if it ends up like your EU bet and AI only reaches 3.99x
I would definitely take your (Caplan's) side in this bet!
I noticed that the stakes aren't even in the bet (Holden can win 5K, Bryan can win 1K) - should we interpret that as Holden thinks there's a >1/6th chance (but <1/2 chance) that his side turns out True, or was there a non-odds-related factor that went into the bet sizing?
I see why you would be proud of his confidence in you. Why are you humbled?
Manifold market for this bet: https://manifold.markets/GaryMiguel/bryan-caplan-wins-feastorfamine-ai
What if the quantity of goods produced has increased as holden predicts, and it is counted in official GDP figures, but AI has driven real prices so low that official GDP hasn't increased that much; would your "conceptual" GDP measure count this as a win for holden? (basically I'm wondering about the case where much of production is counted officially but just not adequately, due to extreme low prices, similar to many currently free or near-free internet services)
It is based on real output, not nominal. If your scenario resulted in mass deflation, then the huge amount of *real* production would indeed count as a win for Holder, so long as humans were interested in purchasing/consuming the stuff.
But if, e.g. it resulted in a 100,000% increase in zucchini production, but the world had no use for the increase after the first 5,000% increase, then the extra would not count as a win for Holder.
Consumer surplus like the free benefits consumers obtain from use of the free Internet is a different issue. I’m very curious about that point myself.
I agree that if prices were uniformly lowered (and the goods-basket remained fixed) then this would be properly handled by inflation-correction, but if for instance manufacturing prices were lowered by automation, but prices for (luxury) human-provided services remained high, it's less clear that inflation-adjustment would address this. Also, I don't understand why you see the case of cheap manufactured goods (from automation) as entirely different from the internet case, given that in both cases extremely low (or zero) real prices mask large impacts on consumer surplus.
A great bet.
All fun and good to see people put their money behind a view. But, who knows what AI will mean either in the short or long term; just way too soon to have enough info to really have good view of it. Hopefully, it will be of benefit to increased human flourishing since a ton of capital investment is being made that investors hope will provide them with returns commensurate with the risks related to the massive capital investment now under way and to continue over coming years.
Relevant: https://www.metaculus.com/questions/4694/gwp-in-2047-in-trillions-usd/
I would be tempted to take Holden's side of the bet: an above average growth in productivity and thus GWP growth seems like it might have more than 20% probability. But I must agree with Bryan that the timelines are perhaps a bit too short for this to happen already.
Overall, I'm very happy to see that this bet doesn't have even odds, like many previous bets where Bryan's side was almost ridiculously likely to with more than 50% probability.
“Overall, I'm very happy to see that this bet doesn't have even odds, like many previous bets…”
Agreed.
At 50-50 Holder’s bet is dumb.
I’d personally put his chances of success at more like 10% - and a big fraction of that is the risk of war downside that might not even have much to do with AI - so I lean Bryan on this one. But the point is it’s not a slam dunk in either direction at these implied odds.
I think which side is good here depends a lot on the time frame. The longer you have, the more weird things can happen.