AI used well can restore middle class jobs [pdf] • Hacker News Board

Re: AI used well can restore middle class jobs [pdf]

#141

Post by pydry » Wed, Apr 03, 2024, 1:00 PM UTC

bumby wrote: post #138

pydry wrote: post #136

Earlier quoted context omitted.

>software has no economic value until it is put into use This is exactly my point. This ^ is categorically false. >If you could quantify fundamental research's value to the economy It seems like you're committing the https://en.wikipedia.org/wiki/McNamara_fallacy

Please explain how shelfware provides economic value. Imagine I make some wizbang software capable of increasing GDP by 50%, but it never gets used. What value did that bring to the economy? I think you are misunderstanding the fallacy you linked. I've already said multiple times there are multiple dimensions to value. The economy is specifically about goods and services sold, so yes, we should measure production. It…

>You seem to be saying "See that idle factory over there!? Look at everything it's contributing to the economy!" and I'm saying "It's not contributing anything because it's idle."

I'm saying that factory has value - sometimes quite a lot of economic value.

The specific example you chose is actually interesting because in the west we largely do view idled factories as valueless and have shifted to a just-in-time manufacturing philosophy whereas Russia has taken the opposite approach of building in a lot of slack into their industrial capacity so that they can scale up quickly.

The practical upshot of this is that our differing conceptions of value are a large part of why we are no losing the largest land war in Europe since WW2. We simply don't have the capacity to scale production to necessary levels.

This is an aside, though.

>Since this discussion concerns wages, my claim was that economic value addition is the most highly correlated with pay.

Yes, this is still false. It correlates with leverage, of which your ability to give the impression of economic value addition is just one part.

Re: AI used well can restore middle class jobs [pdf]

#142

Post by pksebben » Wed, Apr 03, 2024, 2:01 PM UTC

2devnull wrote: post #140

pksebben wrote: post #117

Earlier quoted context omitted.

Not sure I follow you. Do you mean that the existence of AI will undermine trust that applicants know what they're doing? It hasn't really been around long enough for an empirical analysis of effects. It's also incredibly hard to measure effects with the systems themselves being in such flux. For what it's worth, I think categorical statements of any stripe are a little premature. Even the EA folks speak in probabili…

“AI will undermine trust that applicants know what they're doing?” Yes. Is that not obvious? It’s already a major problem in educational settings. I’ve seen video of people using it for job interviews. Have you been handed any code yet that was written partially or wholly by gpt? If not, you will soon, and then you’ll get it.

I see. I suppose I'm sort of partially blind to that given that I've been really lucky to mostly have interviews where the code portions were more "build this whole thing" rather than "whiteboard this function".

Re: AI used well can restore middle class jobs [pdf]

#143

Post by bumby » Wed, Apr 03, 2024, 3:07 PM UTC

pydry wrote: post #141

bumby wrote: post #138

Earlier quoted context omitted.

Please explain how shelfware provides economic value. Imagine I make some wizbang software capable of increasing GDP by 50%, but it never gets used. What value did that bring to the economy? I think you are misunderstanding the fallacy you linked. I've already said multiple times there are multiple dimensions to value. The economy is specifically about goods and services sold, so yes, we should measure production. It…

>You seem to be saying "See that idle factory over there!? Look at everything it's contributing to the economy!" and I'm saying "It's not contributing anything because it's idle." I'm saying that factory has value - sometimes quite a lot of economic value. The specific example you chose is actually interesting because in the west we largely do view idled factories as valueless and have shifted to a just-in-time manuf…

>building in a lot of slack into their industrial capacity

This is exactly what I was cautioning against. Slack in a system is not economically productive. It is risk mitigation. You're conflating economically productive value with risk-reduction value. For the same reason, insurance is considered a liability, not an asset. That doesn't mean adding extra capacity isn't warranted from a risk perspective, but it doesn't mean you can add that excess capacity to the plus side of your ledger. It really seems like we're talking past each other because you aren't able to parse the different types of value I'm talking about.

>It correlates with leverage

I've already conceded this point. I used economic leverage because it is the most generalizable method to get leverage. Sure, you could instead blackmail the boss or date the boss's daughter to get leverage. But I don't think those hold as a general rule for a good strategy so we're mainly left with providing differential economic value. It's like quibbling that offense/defense doesn't matter in football, only points matter. In the same way, harping on leverage is both true and not particularly distinguishing in its helpfulness. Or to tie this into the OP, adding value to society does not give you much leverage in a capitalist system. Adding economic value does. And I think people focus on the wrong value system and get frustrated when the expected results don't follow.