Hacker News Reader

Comment by ben_w | Hacker News Reader

> Something often touted in the psychometrics literature is that the number of jobs for folks with IQ under 85 are quite rare in the US where once they were common. That's ~15% of the population who are close to un-employable.

Although I was completely unaware of this literature, I have spent the last few years essentially making the same point with regard to automation: you don't need a super-duper ASI to cause severe employment problems, all you need is something with an IQ of 85 that costs at most minimum wage to run (modulo robotics keeping up and being part of that price), and then 16% of the population are permanently and irreversibly no longer economically employable.

View on HN

There are many things a robot with "IQ" 85 cannot do that a human with "the same" IQ can: childcare / elder care, social work, basically anything that requires garnering trust and a lot of social interaction, but not necessarily complex skills. There will probably be more jobs like this in the future. Even now, people pay extra for handmade items when the technology to mass-produce the same thing for cheaper has been around for decades, so even if a robot can do your job, that doesn't necessarily make your skills worthless.

It’s safe to talk about EQ in addition to IQ regarding some jobs then. It’s more and more going to be about EQ jobs where AI feels lacking now, emphasis on the now.

Automation in theory SHOULD, in the process of destroying jobs, make many things cheaper so people are able to afford things that were previously out of reach, creating new jobs in those sectors.

Last I checked, no grocery store got price drops thanks to self-checkout.

Due to stickiness prices almost never drop, but they can grow slower than inflation, which in the limit has the same effect.

In addition to the other comment, the cashier is almost none of the price.

Assuming £10/hour and a cashier that can scan an average of 20 items per minute (which is probably 20 items in 20 seconds and then waiting 40 more for the customer to dig out their wallet), that's 20*60 items per hour = 1200 items per hour = 120 items/£1.

I assume it's mainly useful for other reasons?

This is why I put emphasis on the word "should".

> There are many things a robot with "IQ" 85 cannot do that a human with "the same" IQ can:

Only if you don't take IQ tests seriously — I think I've heard them dismissed as "only good at measuring your score on an IQ test".

If you do take them seriously, then by definition an AI which genuinely scores 85 (and isn't, say, just programmed to be good at that specific set of puzzles), then it must be able to do what a human of that IQ can do.

In this case I mean the latter, what with it being a thought experiment to demonstrate why we get problems well before having to ask if the AI can take over.

Practical issues like "a native speaker with IQ 85 is almost certainly more fluent than an IQ 160 polyglot on their 53rd, but for an AI they're all equal", or "we can easily hard-code in an algorithm for winning chess and another for doing calculus" likewise don't really matter.

> childcare / elder care, social work, basically anything that requires garnering trust and a lot of social interaction, but not necessarily complex skills

One of the problems right now with ChatGPT is that people trust it waaaay too much, and it's not really even trying.

> that doesn't necessarily make your skills worthless.

worthless != longer economically employable

The arts have significant social value, and I suspect may be to humans what tails are to peacocks; but at the same time, painters and novelists and musicians and actors and playwrights are infamous in their collective struggles to earn enough to get by on those careers alone.

> then it must be able to do what a human of that IQ can do.

Has anyone ever said this about any test? Are there any tests that are so general that anyone who does better on it is better at every single task than someone who does worse? Is such a test even possible?

That's pretty much exactly the claims that the strongest proponents of IQ tests are making.

I lack the skills to determine that, so I am limited to trusting the people who tell me the tests are "pretty good but not perfect".

Source? Who is claiming the having a high IQ makes you better at for example athletics?

To pre-empt you expanding that to things like running speed, I considered it too obvious to bother mentioning that I'm taking about mental tasks.

For that specific example, athletics, there does appear to be a correlation. It may be any or all of A->B, B->A, and C->(A, B), but "healthy body, healthy mind" is a meme that appears to have a similar age as the concept of writing.

Finally, as I have been specifying IQ 85 as the "huh interesting" threshold rather than "high" IQ, I'll ignore your goalpost shift and say that your citation is me right now saying: I reckon IQ improvements in at least this zone (say +/- 0.5σ) is causal for better outcomes with regard to [avoiding injury, pushing performance limits, maintaining training regimes, learning techniques], all of which improve overall performance.

Reply on news.ycombinator.com