AIs gunning for our precious freelancers

[Rounin Ryuuji]

It seems to be that a number of us (myself included) consider the act of training an AI on copyrighted work, without proper licensing, the preparation of a derivative work (the AI's ability to make images seems to be directly derived from the images on which it has been trained), and consequently an act of plagiarism in itself.

The problem with that argument is then works created by an artist who has viewed others' copyrighted works is automatically making derivative works from them simply by having seen them, if one wants to be consistent.

I fundamentally disagree with this, and didn't initially grasp how you had come to this conclusion, however,

I think the matter here is that some people view the human brain as being special, whereas to me the human brain is simply a machine made out of meat...

I would also fundamentally disagree with this analysis. As far as my awareness of what artificial intelligence can and cannot now do goes, the thing that the human brain possesses that artificial intelligence does not is consciousness.

...
Sure, an AI running on a computer is designed by humans whereas the human brain is not, but that should not exempt the human brain in any fashion. Why is an AI being trained on copyrighted works automatically seen as creating derivative works (even when it does not specifically plagiarize others' works) whereas a human brain trained on copyrighted works is not (unless it actually commits plagiarism)?

I would first return to consciousness for this, too. The AI is a machine designed to take copyrighted works and produce new things specifically from them, and without any capacity for what would make a human-produced work (the input of consciousness that machines do not appear to now possess, changing it in such a way as the initial elements that make one work distinct have all been transfigured into new ones) not a derivative. Where the line is is a matter of debate (and a debate best left to the creators themselves).

Perhaps more importantly, the AI is also itself a copyrightable work derived from copyrighted works in a probably-illegal, and certainly unethical, fashion. A human brain, though it may absorb ideas from copyrighted material, does not have any capacity whose very existence is predicated on intellectual property theft (the ability for the brain to do this evolved at some uncertain point in the past through what would appear to be natural biological processes).

Producing a copyrighted work directly from another copyrighted work (AI from art) would be logically "derivation" of the one from the other. The human brain is not derived from such a process. This considered, the similarity of the function is irrelevant. What the derivative work is designed to do (and that it is an AI rather than a drawing or work of fiction) is also irrelevant. It's a simple matter of, "You don't get to take what isn't yours without consent."

To simplify, (1) the human brain's capacity to produce material is not predicated on the human brain itself being the result of theft; (2) the human brain is also conscious, and processes all intake through a filter of unique experience, making it capable of the sorts of transformations that differentiate inspiration from derivation. If an artificial intelligence were capable of conscious self-expression, I would instead advocate for it to have the same intellectual property rights as biological organisms (and presumably be using neutral sentient being pronouns rather than those for inanimate objects).

[Travis B.]

[snip]

You seem to be predicating your argument upon the assertion that humans created AI's while human brains evolved rather than being created. Consider this, though, what if one created an AI not through carefully constructing an artificial intelligence and manually feeding data in, but rather an AI evolved through evolutionary programming, with the only outside programming by humans being in the construction of the evolutionary framework in which the AI arose? With that in mind, it is possible for AI's to exist which were essentially not programmed and trained by humans, but merely allowed to exist by humans without human intervention beyond the creation of the evolutionary framework itself. Your argument seems far less pertinent in such a case, and makes it seem that argument is largely based on human exceptionalism more than anything else.

[Rounin Ryuuji]

[snip]

You seem to be predicating your argument upon the assertion that humans created AI's while human brains evolved rather than being created. Consider this, though, what if one created an AI not through carefully constructing an artificial intelligence and manually feeding data in, but rather an AI evolved through evolutionary programming, with the only outside programming by humans being in the construction of the evolutionary framework in which the AI arose? With that in mind, it is possible for AI's to exist which were essentially not programmed and trained by humans, but merely allowed to exist by humans without human intervention beyond the creation of the evolutionary framework itself. Your argument seems far less pertinent in such a case, and makes it seem that argument is largely based on human exceptionalism more than anything else.

If this were the case, the discussion might be taking a different direction, but given that this is irrelevant whataboutism, I don't see much point in addressing this hypothetical scenario. We can cross that bridge when we come to it. Perhaps run the experiment yourself, and tell me the results — I imagine they would be quite interesting. Till then, we can worry about what's actually in front of us rather than sweeping it under an irrelevant hypothetical.

My position does have its ethical (and I suppose emotional, or instinctual) basis is in conscious being exceptionalism. Machines capable of consciousness not existing, and the possibility of this being (to my knowledge) presently unknown, and the number of species possessing qualities that result in both consciousness and creative capacity being very small, the qualities of both consciousness and creativity are relatively scarce (as far as is presently known, of course). Humans are, in this regard, "exceptional". If machines gain this capacity, the matter will change.

However, there is a part of my argument that you appear to ignore, and that is that the AI is itself is probably a violation of copyright law, regardless of its function. It is a derivative work, whereas the human brain is not a derivative work, and whataboutism about self-creating and self-evolving AI does not change this. If the AI were attempting some sort of expression as an evolution of some sort of consciousness, the matter would be different, but this is not what is happening. Instead, we have a for-profit entity taking copyrighted works, without licenses or payment, and using them for commercial purposes.

We might even apply some fair use tests to it. Is the work noncommercial? No. Is the work educational? Not as far as I can tell. The amount and substantiality of the work used? Presumably, full copyrighted images were fed in for training, so each individual piece of art used, if copyrighted, could hypothetically constitute a violation. Effect on the value of the original? The technology stands to disadvantage human creators, consequently it could be held to be detrimental to the value of the original. So on the legal side (as already noted) we likely have multiple copyright infringements.

[zompist]

I think the matter here is that some people view the human brain as being special, whereas to me the human brain is simply a machine made out of meat, whereas a computer running an AI is simply a machine made out of silicon and interconnects. Sure, an AI running on a computer is designed by humans whereas the human brain is not, but that should not exempt the human brain in any fashion. Why is an AI being trained on copyrighted works automatically seen as creating derivative works (even when it does not specifically plagiarize others' works) whereas a human brain trained on copyrighted works is not (unless it actually commits plagiarism)?

Neural networks do raise some philosophical problems, but your position here is strangely regressive— you're basically arguing that humans are not "special" so that their legal rights may be removed, while AIs are not "special" enough to have any legal rights. It's consistent, I guess, to treat both brains and silicon as garbage without rights, but I'd expect that position from the far right, not from you.

It's hard to believe you really see no difference between humans and present-day AIs. Are you really advocating that DALL-E is a sentient being, and thus that OpenAI and its users are slave owners, or that they'd be murderers if they turned off the machine? I know that advocating for AI sentience is common among AI people, but if that's what you believe, then you can hardly work to produce any useful or exploitable AI— the only goal could be to create equal or superior beings. And even then I don't think AI theorists have thought hard about what happens as you approach the goal. Does the buggy release candidate have just as much of a right to freedom and legal protection as the final product?

Copyright law is of course intended for humans, and the edge cases have even explored for centuries. "A human brain trained on copyrighted works", that is a human being, is both protected by and subject to copyright law. You will not get anywhere in court arguing that humans are unable to create art without plagiarizing. Usually the concept is pretty blatant, and it is not illegal to (say) paint in the style of Picasso— styles cannot be copyrighted. On the other hand, if you create a work where you just reproduce the top and bottom halves of two separate paintings, that will likely be seen as plagiarism rather than creating a new work.

Now, you could ask how much recombination is needed to count as new work. I'd suggest starting with the far easier case of Markov chains— I have an explanation and working examples here. Is an order-2 Markov chain plagiarism? If you look at the database, it's nothing but two-word phrases, and then a list of single words with numbers. You cannot convincingly claim that phrases like "a rabbit" or "Alice said" are copyrighted.

Yet if you look at the output, entire phrases from the inputs are discernible. The trick, of course, is that the database is interconnected: "a rabbit" and "rabbit with" and "with pink" and "pink eyes" are all in the database, with low enough probabilities that "a rabbit with pink eyes" is a likely output.

What about an order-3 generator? Now we have phrases like "a rabbit with", and the output text will be closer to the original— often duplicating entire sentences. What about order-5, or order-12? By that point the database is arguably citing the original text, and the program will reproduce the text precisely.

The point being: there is a progression here from not violating copyright to certainly violating it, but because humans don't usually create this way, there is not a clear legal line. No judge has ever said "order-n is all right, order-(n+1) is not." Still, there are obviously copyright issues, which is why my page uses public domain or my own texts.

I don't think you can handwave the ethical and legal issues aside by asserting that humans and AI do the same thing. That's a distraction, especially if you are using the argument to remove the real rights of real human beings, and give AI companies freedom to act without ethical concerns.

[malloc]

The title of this thread sounds really dismissive to me. I feel like people don't recognize the existential threat this development poses. If everything, even art, falls under the purview of automation, then practically everyone is unemployable. Billions of people, from menial factory workers to renowned authors, will starve to death because machines have taken their jobs and left few new ones to fill their place. Only the independently wealthy and their courtiers will survive.

[Rounin Ryuuji]

Now that's quite a bit of unnecessary catastrophising. There is no certainty the future will (or even conceivably could) have that form. I don't particularly like the thread title either, but I don't read quite so much spiteful spirit into it, especially as the person doing the splitting would appear to be on the side of the people whose lives stand to be made worse by this.

[Travis B.]

I think the matter here is that some people view the human brain as being special, whereas to me the human brain is simply a machine made out of meat, whereas a computer running an AI is simply a machine made out of silicon and interconnects. Sure, an AI running on a computer is designed by humans whereas the human brain is not, but that should not exempt the human brain in any fashion. Why is an AI being trained on copyrighted works automatically seen as creating derivative works (even when it does not specifically plagiarize others' works) whereas a human brain trained on copyrighted works is not (unless it actually commits plagiarism)?

Neural networks do raise some philosophical problems, but your position here is strangely regressive— you're basically arguing that humans are not "special" so that their legal rights may be removed, while AIs are not "special" enough to have any legal rights. It's consistent, I guess, to treat both brains and silicon as garbage without rights, but I'd expect that position from the far right, not from you.

It's hard to believe you really see no difference between humans and present-day AIs. Are you really advocating that DALL-E is a sentient being, and thus that OpenAI and its users are slave owners, or that they'd be murderers if they turned off the machine? I know that advocating for AI sentience is common among AI people, but if that's what you believe, then you can hardly work to produce any useful or exploitable AI— the only goal could be to create equal or superior beings. And even then I don't think AI theorists have thought hard about what happens as you approach the goal. Does the buggy release candidate have just as much of a right to freedom and legal protection as the final product?

Copyright law is of course intended for humans, and the edge cases have even explored for centuries. "A human brain trained on copyrighted works", that is a human being, is both protected by and subject to copyright law. You will not get anywhere in court arguing that humans are unable to create art without plagiarizing. Usually the concept is pretty blatant, and it is not illegal to (say) paint in the style of Picasso— styles cannot be copyrighted. On the other hand, if you create a work where you just reproduce the top and bottom halves of two separate paintings, that will likely be seen as plagiarism rather than creating a new work.

Now, you could ask how much recombination is needed to count as new work. I'd suggest starting with the far easier case of Markov chains— I have an explanation and working examples here. Is an order-2 Markov chain plagiarism? If you look at the database, it's nothing but two-word phrases, and then a list of single words with numbers. You cannot convincingly claim that phrases like "a rabbit" or "Alice said" are copyrighted.

Yet if you look at the output, entire phrases from the inputs are discernible. The trick, of course, is that the database is interconnected: "a rabbit" and "rabbit with" and "with pink" and "pink eyes" are all in the database, with low enough probabilities that "a rabbit with pink eyes" is a likely output.

What about an order-3 generator? Now we have phrases like "a rabbit with", and the output text will be closer to the original— often duplicating entire sentences. What about order-5, or order-12? By that point the database is arguably citing the original text, and the program will reproduce the text precisely.

The point being: there is a progression here from not violating copyright to certainly violating it, but because humans don't usually create this way, there is not a clear legal line. No judge has ever said "order-n is all right, order-(n+1) is not." Still, there are obviously copyright issues, which is why my page uses public domain or my own texts.

I don't think you can handwave the ethical and legal issues aside by asserting that humans and AI do the same thing. That's a distraction, especially if you are using the argument to remove the real rights of real human beings, and give AI companies freedom to act without ethical concerns.

To make things clear, my position is that sufficiently powerful AI's should not be given any legal disabilities — or privileges — relative to humans. What I meant by human exceptionalism is the position that humans are regarded as special, including legally, for being human rather than on any objective terms. But I do not think that rights should be taken away from humans as a result. As I go into detail below, I view non-sentient AI's should be treated legally like domesticated animals while sentient AI's should be treated legally like humans, with the rights that humans have.

Of course, one can argue, as has been argued here, that AI's are created by humans at some level, even if they are allowed to essentially be generated on their own via evolutionary algorithms, and that their human creators should be at some level held responsible for what they do. This has a parallel in how domesticated animals are treated legally — they are viewed as their owners' responsibility. Of course domesticated animals are not going to plagiarize art — but a parallel can be found in how the treatment of domesticated animals by humans influences their behavior, e.g. a mistreated dog can become aggressive towards both humans and other animals, and this ought to be their owners' responsibility.

But when it comes to sentient AI's that is complicated, not just because of sentience itself (even though that itself implies that such AI's are deserving of rights), but because sentience implies autonomy. A truly autonomous AI can act independently of any human owners. A sentient, autonomous AI, in this case, can decide to create art of its own accord; likewise, a sentient, autonomous AI can collect its own training data of its own will, e.g. in this case other people's art, just like a human artist can. Can you hold a sentient, autonomous AI's ostensible "owners" (if one has a right to "own" a sentient being) responsible when the AI decides on its own to start creating its own art, into which it integrates elements from other art it has seen? But this does imply that sentient AI's themselves ought to be held responsible for their actions as if they were humans, which is a whole nother area of discussion. (E.g. just how does one penalize a sentient AI for plagiarizing art?)

Of course one could say that, well, such sentient AI's are only hypothetical. But considering the rate at which AI technology has advanced in the last twenty years, I think that within my lifetime there will be AI's which will gain sentience. (A good question, though, is will we even know this when it happens?) There are already AI's which arguably pass the Turing Test, but I do not think that is a good standard for judging AI sentience. (What is a good standard is a very good question here, one that I do not feel fully qualified to answer.)

[Ares Land]

The title of this thread sounds really dismissive to me. I feel like people don't recognize the existential threat this development poses. If everything, even art, falls under the purview of automation, then practically everyone is unemployable. Billions of people, from menial factory workers to renowned authors, will starve to death because machines have taken their jobs and left few new ones to fill their place. Only the independently wealthy and their courtiers will survive.

I don't know; I like that title but I also recognize the dangers of automation... The two are not mutually incompatible

I see the following issues myself (much inspired by David Graeber, who was very good at pointing out that sort of issue:)

• Automation is already in full force. In economic terms it translates at a productivity increase. Basically we dirty protes got some of the benefits, but nowhere near a fair share.
• We're attached to a work ethic that might have made sense for some time in the fifties. We expect everyone to work 40+ hours a week on moral grounds when, honestly, there's not enough productive work to go around.
• Consequentially, there are a lot of bullshit jobs around. I.e., jobs where you can work 40+ hours for practically no practical benefits. Essentially, bullshit jobs are courtiers and status displays for the super rich. =
• There is a problem with endless production increases when a) the First World has reached a perfectly satisfactory standard of living b) we're running out of natural resources.
• We're fixated on automation even when automation isn't desirable. Consider self-checkout. FWIW art also falls under that category. I suppose at some point, not too far, we'll get an AI to produce a passable novel... Which will be readable, but a lot less interesting than what a human would've come up with.
• The reasonable thing in several circumstances would be to turn away from automation. Consider public transportation: public transport can be entirely automated, from selling the tickets to driving the vehicle. Consequently, people avoid public transportation, because underground tunnels with no human being inside are creepy and dangerous.

That being said... You can worry about automation, but don't worry too much. A lot of the worry is fueled by extravagant claims AI people make to get investor money. I've mentioned Tesla before, and I'll do it again: the self-driving car are not going to live up to Elon Musk's promises.

FWIW, right now we have a shortage of daycare workers and teachers in France. Neither of these jobs can be improved by technology.

[bradrn]

The reasonable thing in several circumstances would be to turn away from automation. Consider public transportation: public transport can be entirely automated, from selling the tickets to driving the vehicle. Consequently, people avoid public transportation, because underground tunnels with no human being inside are creepy and dangerous.

Wait, what? The other claims seem more or less reasonable to me, but I’ve never heard of anything like this. Over here we automated a train line recently; the only complaints I’ve heard have been from the train workers’ union (which regularly complains about such things for little good reason). I take it fairly regularly myself, and I have yet to see anyone avoid it for that reason.

[Ares Land]

The reasonable thing in several circumstances would be to turn away from automation. Consider public transportation: public transport can be entirely automated, from selling the tickets to driving the vehicle. Consequently, people avoid public transportation, because underground tunnels with no human being inside are creepy and dangerous.

Wait, what? The other claims seem more or less reasonable to me, but I’ve never heard of anything like this. Over here we automated a train line recently; the only complaints I’ve heard have been from the train workers’ union (which regularly complains about such things for little good reason). I take it fairly regularly myself, and I have yet to see anyone avoid it for that reason.

I know it's an eccentric claim, but I do believe I'm on to something here. (You'll get a similar take in David Graeber's Bullshit Jobs.)

Let's take the Paris métro as an example. It's a great transportation network, really! It does suffer from a number of problems.
a) It's complex and it can be confusing for outsiders.
b) it's really, really dirty.
c) I don't believe it's that dangerous, but it sure can feel creepy and scary.
d) it's overcrowded.


Points a and c would be easily solved by simply having a human presence on the platform. You could ask for directions (right now you have to get to one of the few counters and try to communicate through the glass barrier.); as for points b) and c) it's demonstrated that people are a lot more civil whenever there's a figure of authority around.

I'm not a luddite, and technology helps. In fact self-driving subway allow higher frequencies, which help with the overcrowding. But the human touch is neglected.

I don't really object to the self-driving train, but I do think automated ticket selling and stamping made matters worse. This one I picked up from a public transportation enthusiast's website (sadly, I don't remember which one.)

The Paris métro has never been a fun ride. But it was cleaner and felt safer before the 70s. So what happened? Conservatives blame moral decadence and foreigners, because as we all know foreigners didn't exist before 1968 and people were angels back then.

In fact ticket vending and stamping was automated in the 70s. Before, there always was someone from the transport authority on the platform.

[bradrn]

Points a and c would be easily solved by simply having a human presence on the platform. You could ask for directions (right now you have to get to one of the few counters and try to communicate through the glass barrier.); as for points b) and c) it's demonstrated that people are a lot more civil whenever there's a figure of authority around.

OK, this makes more sense. I thought you were specifically talking about self-driving trains, and not anything else.

[linguistcat]

The title of this thread sounds really dismissive to me. I feel like people don't recognize the existential threat this development poses. If everything, even art, falls under the purview of automation, then practically everyone is unemployable. Billions of people, from menial factory workers to renowned authors, will starve to death because machines have taken their jobs and left few new ones to fill their place. Only the independently wealthy and their courtiers will survive.

If everything can be automated, why would we need jobs? Why wouldn't we just end up with a functional form of Communism? Of course we would need ways to spend our time that is not employment, or we could do things that we enjoy and would currently count as employment as hobbies instead of forced labor. But at that point, if we kept money around it would just be because people are so used to having it.

But also, yeah, in current conditions, art being automated (especially through AIs that were trained intentionally on copyrighted works without permission of the original artists) is definitely a danger to people whose livelihoods depend on making art and often having a "brand" they are known for.

Although one thing I have noticed is that AIs aren't great at constructing faces, and when they do well, it's in very simplified styles like most anime. (This isn't a slight on anime, but stating a fact about most of the character designs.) So the hope for human artists seems to be realistic or complex styles used to depict our fellow humans. For now at least.

[Travis B.]

The title of this thread sounds really dismissive to me. I feel like people don't recognize the existential threat this development poses. If everything, even art, falls under the purview of automation, then practically everyone is unemployable. Billions of people, from menial factory workers to renowned authors, will starve to death because machines have taken their jobs and left few new ones to fill their place. Only the independently wealthy and their courtiers will survive.

One thing to remember is that if everyone is put out of work by automation, capitalism itself will implode, because how can capitalists make money if no one has money to buy the products thare are produced?

[malloc]

If everything can be automated, why would we need jobs? Why wouldn't we just end up with a functional form of Communism?

But the people who own the factories and so forth will still expect us to pay them for their products. The goal of every business owner is profit and that requires charging for their products. We would need some way to convince them to become charities, which seems terribly unlikely.

Even apart from that, it really bugs me that they are depriving humanity of something that has defined it since its inception. Art is more than just an economic livelihood. It is also the basis for culture, a means of expressing our deepest feelings. The world is falling apart and everything seems utterly hopeless with global warming, rising fascism, and so forth. They should have left us art as our lone solace if nothing else.

[zompist]

I don't think you can handwave the ethical and legal issues aside by asserting that humans and AI do the same thing. That's a distraction, especially if you are using the argument to remove the real rights of real human beings, and give AI companies freedom to act without ethical concerns.

To make things clear, my position is that sufficiently powerful AI's should not be given any legal disabilities — or privileges — relative to humans. What I meant by human exceptionalism is the position that humans are regarded as special, including legally, for being human rather than on any objective terms. But I do not think that rights should be taken away from humans as a result.

You seemed to be arguing precisely that rights should be taken away from humans, namely copyright protection, on the grounds that some eventual sentient AI could absorb other artwork like a human.

But when it comes to sentient AI's that is complicated, not just because of sentience itself (even though that itself implies that such AI's are deserving of rights), but because sentience implies autonomy. A truly autonomous AI can act independently of any human owners.

I'm not sure how you can write the above and not notice the ethical nightmare. Human owners? You do realize you're proposing slavery? Does slavery only matter for humans? Or since you think humans and sentient AIs should be treated the same, do you think humans shouldn't get an "essentialist" right— not being slaves— that AIs don't get?

I know you haven't said "rah rah slavery!" Maybe you're disquieted by the concept. But you just seem to blow past it without asking, is this even a path we want to go down?

Of course one could say that, well, such sentient AI's are only hypothetical. But considering the rate at which AI technology has advanced in the last twenty years, I think that within my lifetime there will be AI's which will gain sentience.

And again, are you thinking about what you're saying? People should just go on and create sentient AIs, which you posit will be owned and exploited by humans? Why, because it would be a fun programming exercise?

Also see my previous post, where I note the ethical problems of developing a near-sentient AI, which AI proponents never seem to think about. When is the ethics we apply to sentient beings supposed to actually apply to the process of development?

[Travis B.]

I don't think you can handwave the ethical and legal issues aside by asserting that humans and AI do the same thing. That's a distraction, especially if you are using the argument to remove the real rights of real human beings, and give AI companies freedom to act without ethical concerns.

To make things clear, my position is that sufficiently powerful AI's should not be given any legal disabilities — or privileges — relative to humans. What I meant by human exceptionalism is the position that humans are regarded as special, including legally, for being human rather than on any objective terms. But I do not think that rights should be taken away from humans as a result.

You seemed to be arguing precisely that rights should be taken away from humans, namely copyright protection, on the grounds that some eventual sentient AI could absorb other artwork like a human.

I was arguing that sentient machines should be given the same rights as humans, and just as one does not consider a human artist who looks at other humans' copyrighted art as inherently plagiaristic unless they, well, actually commit plagiarism, sentient AI artists should be seen the same way rather than as being seen as inherent copyright violations for simply having seen other artists' works.

But when it comes to sentient AI's that is complicated, not just because of sentience itself (even though that itself implies that such AI's are deserving of rights), but because sentience implies autonomy. A truly autonomous AI can act independently of any human owners.

I'm not sure how you can write the above and not notice the ethical nightmare. Human owners? You do realize you're proposing slavery? Does slavery only matter for humans? Or since you think humans and sentient AIs should be treated the same, do you think humans shouldn't get an "essentialist" right— not being slaves— that AIs don't get?

I know you haven't said "rah rah slavery!" Maybe you're disquieted by the concept. But you just seem to blow past it without asking, is this even a path we want to go down?

You seem to have missed where I wrote:

Can you hold a sentient, autonomous AI's ostensible "owners" (if one has a right to "own" a sentient being) responsible when the AI decides on its own to start creating its own art, into which it integrates elements from other art it has seen?

I don't believe a human has a right to own a sentient AI in the first place for the very reason that it would be slavery.

Of course one could say that, well, such sentient AI's are only hypothetical. But considering the rate at which AI technology has advanced in the last twenty years, I think that within my lifetime there will be AI's which will gain sentience.

And again, are you thinking about what you're saying? People should just go on and create sentient AIs, which you posit will be owned and exploited by humans? Why, because it would be a fun programming exercise?

While humans may believe they "own" sentient AI's, I do not believe they have a right to do so. About people just going off and creating sentient AI's, I suspect that the first sentient AI's may be created unintentionally, that AI designers may create systems which achieve sentience without that being their intent (especially as a sentient AI is likely to be disadvantageous to AI creators, because an AI with a will of its own is likely to not act as its creators wish, and may very well disobey them).

Also see my previous post, where I note the ethical problems of developing a near-sentient AI, which AI proponents never seem to think about. When is the ethics we apply to sentient beings supposed to actually apply to the process of development?

There are definite ethical issues with near-sentient AI, because it allows the advantages to its "owners" of sentient AI, without as much capability for said AI to act truly autonomously and have a will of its own. It may very well enable a sort of slavery where the slaves have no capacity of disobeying or resisting.

[Ares Land]

Even apart from that, it really bugs me that they are depriving humanity of something that has defined it since its inception. Art is more than just an economic livelihood. It is also the basis for culture, a means of expressing our deepest feelings. The world is falling apart and everything seems utterly hopeless with global warming, rising fascism, and so forth. They should have left us art as our lone solace if nothing else.

I don't think AI is doing art. It's using what's already in the data sets it's been trained, but that's all.

If Munch or Picasso didn't exist... DALL-E would never come up with anything like Guernica or The Scream ; there'd be nothing like them in its data set.

On another note: we're nowhere near automating everything.
Some idiot at work pointed out, some time ago, that I should automate my job. I work in IT; everything I can automate or script I have automated or scripted long ago. My job is essentially checking out weird bugs and explaining complex stuff to people. None of it could be automated in the foreseeable future.
I mentioned a shortage of daycare workers earlier...we have nothing like an AI that could take care of a small child.

(As for sentient AI: we sometimes hear AI insiders claiming their AI is self-aware. They may be delusional, but it's more likely it's all publicity stunts.)

[linguistcat]

But the people who own the factories and so forth will still expect us to pay them for their products. The goal of every business owner is profit and that requires charging for their products. We would need some way to convince them to become charities, which seems terribly unlikely.

Do you think that if there are products being produced but no money to buy them, people won't just resort to stealing? Or taking over the automated factories for themselves if the owner shuts them down? We're going to have a UBI long before all this goes down, or the 1%ers are putting their own heads in the guillotines.

But if everything is automated maybe Earth will just become a world of competing "paperclip machines" and even the 1%ers will have to run in fear so they don't become leather handbags no one will ever use. If you're going to catastrophize, get more existential horror and less teen dystopia.

[hwhatting]

Topical.

[malloc]

What is the official reason for creating these AIs in the first place? With inventions like antibiotics or air conditioning, one can point to obvious practical benefits and new functionality. Yet these AIs are intruding on a niche that was already adequately filled. Given the profusion of far right extremists in software engineering like Mencius Moldbug, it seems impossible for me to interpret this as anything but an attack on human workers.