David Manheim posts LessWrong analysis arguing the agricultural and industrial revolutions worsened health and mortality for living generations despite long-term gains, suggesting similar near-term costs even if AI proves a normal technology

@davidmanheim @BenSasse The IT revolution was mostly positive even while it was happening, but yes, agreed re Industrial Revolution.

@alexolegimas @BenSasse I agree that it's possible we manage the economic transition better - but we're pretty clearly demonstrating that our political capacity for mitigating the externalities is at best lethargic. The industrial revolution's downside was a century; AGI's might be a decade, or a year.

Not only is this largely consistent with our views, it is a big part of the reason why we wrote AI as Normal Technology in the first place! Past technological revolutions led to a lot of misery in the short term, and maybe we can do better this time if we more proactively address these "normal" harms.

@petz_a @alexolegimas @BenSasse ...because the rich and the ruling class have always been rational and careful about avoiding angering the masses?

@davidmanheim @BenSasse I think we're actually much better at managing fast transitions that slower ones. See COVID for example, you can't say that fiscal response was lethargic. Compare that with de-industrialization which was much much slower.

@davidmanheim @BenSasse The good news is that we literally did not know how the economy works at all during the Industrial Revolution; hopefully 200 years have taught us something.

@alexolegimas @davidmanheim @BenSasse Also ppl know that baricades and pitchforks work… and so do politicians… so hopefully it gets avoided…🤷‍♀️

@alexolegimas @davidmanheim @BenSasse People got good low skill jobs during the Industrial Revolution same with the IT relocation.

In the AI revolution, the top 0.01% of talent is getting paid ungodly amounts to automate the work of millions.

New jobs like data annotation is being farmed out to Africa for cheap

Well in footnote 18 you're just saying it's unlikely, not why that is. And the "Stochastic Parrots" paper is just one POV,

I think what you have to grapple with is that we already have amazing AI technology, we've had it for years and yet the impacts have not been nearly as transformative as the average person probably would have guessed if you had described this technology to them ten years ago. So why is that? The answer to this question might tell you something about the future development.

@_everythingism @alexolegimas @BenSasse Footnote 13 and 18 deals with this possibility quite exhaustively, in my view.

@_everythingism @alexolegimas @BenSasse 1. My response to the stochastic parrots paper was to point out that it doesn't mean what you think, or it was never true. And at least one of the authors of the original paper admitted as much, if you read the linked post.

2. You are confusing ML and AGI, just stop.

@_everythingism @alexolegimas @BenSasse You never mentioned which paper?

@davidmanheim @alexolegimas @BenSasse I never mentioned that paper so it's unclear why you think it's a response to the idea that AI could have a more muted impact than some are predicting.

@random_walker Thanks!

I still think that undersells the fact that the industrial revolution was net-negative for the vast majority of those living in industrializing countries for around a century - and also think it's very likely that AI will not be a normal technology in the coming years.

@davidmanheim @petz_a @alexolegimas @BenSasse look cake has gotten demonstrably cheaper

@alexolegimas @BenSasse I don't think that the problems we're going to have with AI will be solved with a fiscal response.

And if you think public health responded reasonably to COVID, I was working on it; I promise, it was a clusterfuck, it lasted for years, we still disagree about what went wrong.

Posted on my Substack, and on Lesswrong. https://www.lesswrong.com/posts/4MCuvdsZFEBAaGCsb/if-ai-is-normal-technology-history-is-not-reassuring Also, pinging some of the people whose work or ideas are discussed or mentioned: @tobyordoxford, @deanwball, @random_walker, @sayashk, @SBenzell

@alexolegimas @davidmanheim @BenSasse Yes, hopefully we do better. 4 in 10 children in the richest country in the world were living in absolute poverty after 150 years of the Industrial Revolution. Stunted and diseased.

Peter Laslett, The World We Have Lost: Further Explored, Routledge 2005, ch 11.

@davidmanheim @alexolegimas @BenSasse Huh? You're the one who brought it up not me.

I'm saying debunking that one paper doesn't rebut the idea that AI could have a more moderate impact. There are many reasons someone could think that.