(removed duplicate)

Remember when people promised “an iPad and google and third worlders can learn anything!”

Those same people now think AI will make every student two standard deviations better or transform roles that marginally if at all leverage intelligence as is 🤔

Healthcare, education, and construction don’t exactly move quickly. Arguably have negative productivity growth over the last few decades.

Perhaps writing more words or code is possible—but this is a small slice of gdp.

Over decades: transformative enough to replace broken industries (in other countries and free regions); short-term wouldn’t expect the cartels to change much.

Robots on the rise
Economy's demise?
AI's the culprit, experts surmise

If everyone dies then GDP per capita will increase significantly.

@MartinRandall this comment is not the most helpful

@ErickBall Oh, what's your theory of change for that claim?

@MartinRandall if everyone dies then we won't have statistics on GDP, and also we'll be dead. Tbh I googled "theory of change" and it all sounds like nonsense. The realistic scenario we're betting on makes the implicit assumption (reasonable imo) that human civilization will be continuous through 2028.

@ErickBall I think an AI that kills everyone in five years will be able to simulate Scott Alexander in sufficient fidelity to determine the correct resolution for this market. Killing everyone will produce a visible break in GDP, GDP per capita, unemployment, and productivity. I agree that it's unknown whether markets will resolve correctly in this scenario.

It doesn't have to be human extinction to impact GDP per capita. For example:

/JonathanRay/will-ai-kill-20-of-the-human-popula (8% at time of linking)

The impact of mass human death on GDP per capita is apparently more nuanced than I thought, from this 2020 paper laying out the impact of the Black Death.

https://www2.gwu.edu/~iiep/assets/docs/papers/2020WP/JedwabIIEP2020-14.pdf

8% is low, but in the context of a market at 28%, significant.

@MartinRandall If everyone dies, the environmentalists and socialists will finally be satisfied, because we will acheive the Kyoto Protocol targets and the Gini Coefficient will be 0.

@AlQuinn we're 10 years too late to meet the Kyoto Protocol targets, they were set in 1997 for 2008-2012.

Also I think genocidal AI may end up with high carbon emissions.

@MartinRandall don't be so negative; perfection is the enemy of good

@MartinRandall you made me curious. Turns out the Kyoto protocol was actually met on the aggregate, though most countries did not meet their individual targets. https://www.iedm.org/56519-have-the-kyoto-protocol-targets-been-respected/

@PatrickDelaney That is fascinating. Thanks.

15%

@tailcalled why so low? 5 years for AI to have a noticeable effect on (sluggish) gdp growth (or any other metric). When are you expecting the release of gpt5? When are you expecting rhe first wide use of free agents/assistants?

I'm at >50% (and mostly this low because of epistemic humility)

@ElliotDavies 😅 I don't really have any calculations or anything, I just make bets based on gut feelings.

I love that these get included in Scott's writing now. <3