Hacker News Reader

Comment by Sevii | Hacker News Reader

It's not going to be half of the population. We are approaching a world where 1% can contribute useful work and the other 99% cannot contribute anything beyond entertainment. Then the question is why should the 1% accept a governance structure where they get only 1% of the power for 100% of the work?

View on HN

I'm not certain of any specific percentage, but here's a thought experiment:

Imagine there was an AI which was normal compared to a human mind — as in, you can reasonably assign it an IQ score the same way you would a biological human, and the score wouldn't be biased by it doing the maths and memory puzzles a million times faster even as it fails a simple word game or is unable to either count or subitize a small number of objects in an image.

IQ scores are normally only valid in the "+/- 2 standard deviations" range, which by definition is IQ 70 to 130.

(IQ tests are just a direct mapping onto standard deviations, not at all a linear scale, so while I'm very impressed with ChatGPT, I don't know if it would always outperform an IQ 70 human or not, even though it can definitely sometimes outperform IQ 130 human).

Assume this AI's electricity bill is equal to Example Everyperson's wages.

If the AI has an IQ of 70, then 2.3% of the population are permanently uneconomical through no fault of their own; 85, then 15.9%; 100, then 50%; 115, then 84.9%, 130, then 97.7%.

What is the likely IQ of the first human-like AI whose electricity cost matches an average salary? And how fast is it likely to go up?

~~~~

From the other direction: I don't know for sure how much ChatGPT costs to run, though I think it's in the order of cents per response. What I do have is an upper bound for Stable Diffusion of $0.0001 per image, less than half of the $0.00022 it takes to keep a human (on the $1.90 per day UN abject poverty threshold) alive for 10 seconds to type in the prompt[0].

I don't know what the future of AI will be, but I suspect that it will be so cheap that that won't need to be a superior intelligence to all humans in all domains, to cause literally all humans in all domains to be uneconomical.

[0] https://kitsunesoftware.wordpress.com/2022/10/09/an-end-to-c...

Those people with IQs below 70 are already uneconomical. I don't know where the line is for "we can find productive work for this idiot", but it's definitely higher than "cannot tie/Velcro shoes"

If military standards are anything to go by, it's roughly IQ 83 by law. Standard requirements are closer to 91, those this is a rough equivalence.

The military found that people with lower scores were not even up to correctly sweeping the floors consistently.

The actual test is not an IQ test but the Armed Services Vocational Aptitude Battery, which has rough equivalents to IQ scoring.

This is a pretty shrewd comment. It won't matter if an AI is more capable than a human, if it costs more to run that task than it would to hire a human. There are still people working in mines in poor countries because they're cheaper than the machines used in rich countries.

Which 1% of the population are you counting as doing useful work? From how I see it - those doing useful work right now are compensated far far less than those doing well paid but useless work

indeed, there is a lot of useful work that people could do that is currently not paid well. for example education, science, healthcare.... imagine what we can achieve, when all those who are not needed to sustain our life study science and research or medicine. parents need not work but can focus their entire energy on raising their children.

or a different model, if 1% need to work 40 years to sustain us, then 10% can do the same while working only 4 years, or everyone only works one year. (ok, it doesn't scale like that because it takes more time to get experience for certain tasks, but even then, with such an abundance of available labor, every worker can spend half of their work time to teach others about their work)

all it takes is an understanding that the best thing to do with those not needed for work is to educate them to do something else that is useful for us.

thank you, this is exactly what I was trying to say but worded so much nicer

True. Also, reciprocally, why would the 99% accept a governance structure where they get 0% of the governance?

If the 99% need the work of the 1% but the 1% do not need them, then they have no leverage and must accept what is offered.

Well, not that I'm advocating for it, but if 99% feel oppressed by 1%, you should expect a revolution.

The scenario here is one where 99% of people's jobs have been automated out, so I'm kind of assuming that there are tons of robots and the 1% are somehow necessary to control the robots. And then there can't really be a revolution because the 1% will just tell the killbots to kill the rebels.

Don't forget that violence is always a possible outcome.

> Then the question is why should the 1% accept a governance structure where they get only 1% of the power for 100% of the work?

A job doesn't entitle you to power.

If they think they are underpaid, they can ask for more money.

If the wages of that sector go up a lot, watch as half the world applies for those jobs.

I think the implication of the parent comment is a world in which only the smartest 1% are able to do things that can't be automated by an AI.

It's not specifically what was written, but it's what I see between the lines.

That's a weird definition of "useful work" though. If they're not farming or manufacturing or distributing or maintaining things then why should they get significant power?

And if 99% of the population is too dumb to do it, then it probably doesn't need to be done.

>And if 99% of the population is too dumb to do it, then it probably doesn't need to be done.

This is very hard to believe. That's basically conceding that we will never cure cancer/colonize mars/solve productive nuclear fusion etc.

Damn, you really don't have a high opinion of the bulk of the world's population. You really think it takes special geniuses to do those things, not just experts? 125 IQ need not apply?

IQ 125+ is a bit less than the top 5%.

The country with the highest percentage of PhDs is Slovenia, at 3.8% of 25-64 year-olds.

I know academic performance doesn't perfectly equate to IQ, especially at this level because right now smart people have many other options, but on the other hand that's the highest percentage and the OECD average is just short of 1.1%.

OECD PhD percentages: https://doi.org/10.1787/888933978645

China I'm not sure about, but seems to be about 500k/year, so if I estimate population of 1.5e9 and 80 year life expectancy, that's a steady-state level of 3.75% of the entire population.

So the minimum for those things being above IQ 125 seems at least plausible. May not be correct, but it is plausible.

Farming, manufacturing, distributing, and maintaining things are all directly getting automated. The fist stage of this automation, when the automation wasn't computerised, was named "the Industrial Revolution", but automation didn't stop there, and computerisation is continuing to reduce human involvement in each sector.

As for "why should they get significant power?" — well, 'should' is a question of values. Do you believe in democracy and self-determination? Or is it fine to treat them all like animals, by which I mean spay, neuter, euthanise without asking their opinions, and/or selectively breed to be decorative and fit into handbags? Because the worst horrors that Charlie Stross and Ian Banks and Harlan Ellison have imagined are all as easy as Fully Automated Luxury Space Communism if you have an AI which can automate almost everything, regardless of whether anything remains purely in the domain of a tiny elite.

I'm going to support democracy. That's why I think an unemployable 99% should have, in a reasonable sense, 99% of the power.

I think you might have misread me. When I said "significant power" I meant that in a proportional sense. The 1% getting much more than 1%. I wasn't arguing against the 99% getting 99%.

Ah, thanks; that explains the apparent discrepancy between what I thought you were saying here vs. your other comments.

As for why the 1% may have much more than 1% of the power in practice… that's kinda where I was going with the Hyper-Dystopia vs. Fully Automated Luxury Space Communism example: whoever controls the AI gets to play as Gods to a much larger degree than mere billionaires get to act like real life is a Civilization/KSP/SimCity game, and the default for who controls the AI is whoever has skills such that they can't be replaced with an AI. Or the current rich, if everyone can be displaced by AI.

Why should the 99% accept a governance structure where they get only 1% of the power/wealth? I think the core issue is wealth inequality, and although I initially liked UBI, it doesn't really solve that issue. If we don't solve it, I wouldn't be surprised to see a situation where we end up with another failed attempt at communism/revolution and everyone loses.

Reply on news.ycombinator.com