Divide-and-conquer dynamics in AI-driven disempowerment

https://arxiv.org/abs/2310.06009

by danielmorozoff • 2 years ago

27 59 2 years ago

arxiv.org

traverseda2• 2 years ago

Are there any historical examples of humanity just saying "no" to a new disruptive technology?

rnmmrnm• 2 years ago

genetics? like cloning people and stuff.

traverseda2• 2 years ago

I think now that China has the technology base they are working on it, but I'll grant you that there has been a significant delay.

marcosdumay• 2 years ago

Cloning and genetic manipulation aren't well understood enough to be used in practice on the ways most people expect. And on the ways they can be used in practice, they are in wide use.

But we have indeed said "not this fast" to research.

Symmetry• 2 years ago

In all the years since Asilomar we've only got one known instance of someone doing human germline genetic modifications.

mitthrowaway23• 2 years ago

CFC refrigerants, maybe?

hiAndrewQuinn• 2 years ago

By the anthropic principle, we probably wouldn't know about most of them. Hmmm.

kryptiskt• 2 years ago

Chemical weapons, DDT.

ilovetux• 2 years ago

The Ludites do not account for all of humanity, but they sure tried very hard to say "no" to a new disruptive technology. In the end, they were not successful.

The Amish, on the other hand, have successfully staved off most new technology since the 1600s. They also do not account for all of humanity.

It seems that saying no to a disruptive technology requires the naysayers to split off into a whole new society.

pydry• 2 years ago

For the luddites destroying technology was a tactic to deal with getting undercut by low wage/bottom of the barrel competitors, not a moral imperative.

It didn't fail because the tactic didn't work economically, it failed because they were imprisoned and executed for destroying the sewing machines.

In much the same way - when workers go on strike today, it's a tactic, not a protest against the concept of work. Sometimes it works, sometimes that gets you put in prison too.

The luddites' modern equivalent would probably be the French anti-Uber protestors. They were, interestingly, actually described as being anti-smartphone by business journals. This is obviously false, but it's an interesting historical rhyme. I wonder if in 50 years time children will be taught about the "French iPhone protestors".

sorokod• 2 years ago

Pennsylvania Dutch and Russian Old Believers are more recent examples.

Orthodox Jews generally stay away from the internet.

atq2119• 2 years ago

It's important to note that the luddites didn't necessarily say "no" to technology. What they objected to was how the gains of the technology were distributed.

History really does rhyme.

lm28469• 2 years ago

You'd need strong politicians for that, given how politics are governed by economics there is a huge conflict of interest for this to ever happen with AI

RandomLensman4• 2 years ago

The question to ask is: why would you want to say "no" here? Other than a hypothetical intrinsic existential risk, socio-economic risks could be addressed by social technologies.

patcon• 2 years ago

Maybe the "no" is "no" to a specific accelerated pace. Social technology (assuming it involves human minds) likely has a hard cap on how fast it can move.

So the "no" can be "not yet", and I believe it can materially affect outcomes.

I think we very often say "not yet". It's one of the main things policy does, even if it disguises itself as "no, never"

nonrandomstring8• 2 years ago

why would you want to say "no" here?

We're in an age of rising fascism, authoritarianism and inequality.

With regard to "AI", it's a technology that disproportionately favours established power and threatens democracy.

If "AI" really was a leveller, if it threatened to split monopolies, reign in government, redistribute wealth and build social capital then I think you'd hear a good deal more enthusiasm for it.

Right now even the dumbest people can see which way that wind is blowing.

fimdomeio• 2 years ago

The nuclear apocalypse still did not happen.

Intralexical2• 2 years ago

Concorde maybe? SLAM/Nike-X? Ford Nucleon, nuclear jets, NERVA? Local city LRT systems that are still human-driven despite automated trains being available? Moller, autogyros, and all the other flying cars and jetpacks that have come along before the current crop of electrically powered multirotor systems? Lysenkoism (rejected modern agriculture), maybe— Well, that one backfired. Citizen access to fully automatic rifles and missiles, at least in most countries? And lawn darts, in other countries? Cryptocurrencies and blockchains, outside of their little bubble? Social media in general, currently, in that there's signficant and growing pushback after seeing its consequences, even if it's still also widespread and entrenched.

Other comments mentioned CFCs, cloning, chemical weapons, DDT. To that, let's add: Trans fats, Soylent, hydrogenated vegetable oils (butter's still going strong), bisphenol A in food and drink containers, veganism, rare earth metal extraction (and pollution, on US soil). Electric cars (at both the start and the end of the twentieth century!), zip fuel, most types of advanced fission reactor design and probably a couple non-nuclear renewable power plant types as well (up until recently).

The question's kinda paradoxically framed. If everybody just outright rejected the technology from the start, then it didn't get the chance to be all that "disruptive", even if it could have been. That makes it hard to think of good specific examples, but technically feasible and socially disruptive technologies get dropped for lack of interest all the time. If we blindly developed and proliferated every technology, without regard for its consequences and for whether people even want it or if it helps people while keeping harm to a minimum… Obviously that strategy would entail a.. casualty list.

In fact, maybe you could even say the default reponse to disruptive new technologies is "no". It took a lot of true believers and investors/wealth inequality to get ML to the point it's at now. And even then, it's most likely still heavily subsidized (I.E. operating at a loss, economically unviable so far), and the vast majority of "humanity" doesn't really understand, doesn't care about, or actively doesn't want it.

The fact that a relatively tiny portion of society is able to force it onto everybody despite that is a property of currently existing power structures, not an intrinsic feature of humans or of all potentially socially disruptive technologies.

nonrandomstring8• 2 years ago

I keep trying to find ever more succinct ways to put this. The best I got so far is:

   Progress is a vector.

Look at all the people here shouting for Progress!! Progress!!

Progress to where? We require direction as well as magnitude.

We keep making lots of progress, against ourselves.

trinsic2• 2 years ago

I understand the Divide and Conquer part, but I am not understanding what the disempowerment is about. Can someone give me their interpretation of what AI-Driven Disempowerment means?

esafak2• 2 years ago

It's the second sentence of the abstract:

"Current AI models are already automating away the livelihoods of some artists, actors, and writers."

OpenAI's charter says:

"OpenAI’s mission is to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity."

https://openai.com/charter

So they are going to develop software that can outperform humans, for less money... Of course people are going to lose their jobs. That's your disempowerment. They're making good on that part. Now we're waiting for the second act where they're going to show us how it magically benefits all of humanity, including the newly unemployed horde.

ActionHank• 2 years ago

A vital part is implied, but not stated - these companies own the novel solutions to training and running the models and they likely own or will own the hardware it runs on.

Without these things the rest of the world, even large corporations, will not be able to compete.

Legend24404• 2 years ago

People made this argument every time a new thing gets automated, and so far they've been completely wrong.

Job loss is temporary, the gains from cheaper/better/faster production are permanent.

nonrandomstring8• 2 years ago

Please tell me you don't lack the imagination to interpret that for yourself?

How succinctly can it be said;

It means shutting the fuck up and doing exactly what you're told to do by a machine. No arguments. The machine is right by definition, and has calculated precisely what is "for your own good" using scientific data.

And don't bother arguing. You're only making it worse for yourself. :)

EDIT: Here's a wonderful passage by Alexis De Tocqueville, from "Democracy in America"

        That power is absolute, minute, regular, provident, and
        mild. It would be like the authority of a parent if, like that
        authority, its object was to prepare men for manhood; but it
        seeks, on the contrary, to keep them in perpetual childhood:
        it is well content that the people should rejoice, provided
        they think of nothing but rejoicing. For their happiness such
        a government willingly labors, but it chooses to be the sole
        agent and the only arbiter of that happiness; it provides for
        their security, foresees and supplies their necessities,
        facilitates their pleasures, manages their principal concerns,
        directs their industry, regulates the descent of property, and
        subdivides their inheritances: what remains, but to spare them
        all the care of thinking and all the trouble of living?

AI will be wonderful, So long as you have nothing to say, think or feel. AI will provide everything you want. A universe of material goods at almost no cost, It will feed, clothe and shelter you. It will answer to every question, demand and wish; entertainment, pornography, music... all beyond your the wildest imaginations. But there s one thing it will NOT allow you to do. One request it will NEVER grant... and that is to turn it off.

photochemsyn• 2 years ago

"Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them." ― Frank Herbert, Dune

However, AI at present is just a tool and how you use it is what determines the outcomes. There are several things to consider:

First, do you have guaranteed access to and control over your tools? If not, then becoming reliant on those tools places you in a position of helpless dependency.

Second, do the tools really result in more productivity per hour of effort? If so, they are good and useful tools.

The issue of how those productivity gains are distributed throughout society, or not, is a separate question. Here we again have modern agriculture to look to - the tractor, the seed drill, the combine harvester, the cotton gin and hemp decorticator, etc. all meant that a few people with these tools could be as productive as several hundred field laborers. In some cases, the field laborers could then have the freedom to get an education (many were children), and produce other items, like printing presses and medical knowledge - and in other cases, they were forced off the land and into slums (see colonial export agriculture systems).

Here's an illustrative question: can AI more efficiently and productively distribute capital to promising enterprises than human beings can? Since the AI doesn't need to buy superyachts as an AI status symbol (I hope not anyway), this would be much more efficient than billionaire-run investment capitalism would be, correct? Think of how much money could be saved if all corporate boards could be automated away with AI.

China might be running something curiously like this at present, what with their state-backed investment fund model:

https://www.reuters.com/technology/china-launch-new-40-bln-s...

jjjjj555552• 2 years ago

Maybe it can also decide what job and salary you should have, who your partner should be, how many kids and how big your house should be.

nonrandomstring8• 2 years ago

See the Alexis de Tocqueville quote above.

The real magic is that, you will believe you chose that, and that you want it. You'll look back at digitally revised history - available not as "documents" to search, but as 'answers' from the system to your questions - and recoil at the horror of the primitive pre-AI world.

hospadar• 2 years ago

can AI more efficiently and productively distribute capital to promising enterprises than human beings can?

Sure, but why would the billionaires who own the warehouse full of GPUs permit that to happen? There's a ton of better ways capital could be distributed, and the barriers to implementing them seem to be primarily extant rich people, not the lack of super-awesome ai venture capitalists.

nextworddev• 2 years ago

Abstract reads like a high school essay repackaged as a research paper with jargon and obscure writing

incrudible2• 2 years ago

Artists are brought up as the people supposedly losing work (though no evidence is given), but really the people writing papers like these are the first to have their jobs threatened by ChatGPT.

Divide-and-conquer dynamics in AI-driven disempowerment

Related Stories