AIs gunning for our precious freelancers

[alynnidalar]

There is a very big difference between "humans become cogs in the machine of industrialization but are very much still an integral part of the process, including performing almost all of the creative work in the field, which pre-industrialization was honestly still only rarely done by ordinary workers anyway" and "generative AI will replace human workers entirely, especially the creative bits". You can't just take arguments and evidence that apply to one scenario and apply them to the other.

So yes, I think the fact that the overwhelming majority of clothing manufacture (including the creative design work) is not only done by human beings but cannot be done by computers is very much relevant to explain why it's a bad example for you to point to.

(Incidentally I think you also have a very rosy view of the pre-industrial world...)

[Raphael]

(Incidentally I think you also have a very rosy view of the pre-industrial world...)

One of the more annoying things in politics, IMO, are all those people who think of themselves as politically left-wing or progressive or something like that while having a rosy view of the pre-industrial world. Folks, back then, aside from a few utopian religious movements, there was no kind of left-wing or progressive politics whatsoever in most of the places where most people lived. Almost every place that mattered was solidly ultra-reactionary. And to a large extent, the same forces that created the modern industrial world also led to the rise of the first left-wing political movements.

[WeepingElf]

One of the more annoying things in politics, IMO, are all those people who think of themselves as politically left-wing or progressive or something like that while having a rosy view of the pre-industrial world. Folks, back then, aside from a few utopian religious movements, there was no kind of left-wing or progressive politics whatsoever in most of the places where most people lived. Almost every place that mattered was solidly ultra-reactionary. And to a large extent, the same forces that created the modern industrial world also led to the rise of the first left-wing political movements.

Perfectly correct. Indeed, I wouldn't call those romanticizers of the pre-industrial world "progressive". They may pretend to be progressive, but they are in fact as reactionary as Trump, AfD & co., even though in a superficially different way. In a low-tech world, either most people are poor, or all people are. You just can't achieve an affluent society without modern technology. (I met quite a few of those romanticizers in the tabletop role-playing game community, which was one of the reasons why I abandoned that hobby.)

And malloc is moving the goalposts.

[Raphael]

Indeed, I wouldn't call those romanticizers of the pre-industrial world "progressive". They may pretend to be progressive, but they are in fact as reactionary as Trump, AfD & co., even though in a superficially different way.

Not sure I'd go that far. If you subtract all the people who romanticize the pre-industrial world, how many progressives are even left these days?

[WeepingElf]

Indeed, I wouldn't call those romanticizers of the pre-industrial world "progressive". They may pretend to be progressive, but they are in fact as reactionary as Trump, AfD & co., even though in a superficially different way.

Not sure I'd go that far. If you subtract all the people who romanticize the pre-industrial world, how many progressives are even left these days?

Quite some! I don't know where you are looking and see lots of progressives romanticizing the pre-industrial world. Or did you miss the changes e.g. the Green Party has undergone in the last 40 years? The Fundis are long gone, man. Only the far right still believe that the Greens are anti-technology today. Today's progressives embrace shiny new technologies such as wind turbines, photovoltaic cells, state-of-the-art public transit and open-source software. They want to overcome industrialism, sure, but not by romanticizing the pre-industrial world, rather by attaining a new post-industrial world. That's a big difference.

[malloc]

There is a very big difference between "humans become cogs in the machine of industrialization but are very much still an integral part of the process, including performing almost all of the creative work in the field, which pre-industrialization was honestly still only rarely done by ordinary workers anyway" and "generative AI will replace human workers entirely, especially the creative bits". You can't just take arguments and evidence that apply to one scenario and apply them to the other.

So yes, I think the fact that the overwhelming majority of clothing manufacture (including the creative design work) is not only done by human beings but cannot be done by computers is very much relevant to explain why it's a bad example for you to point to.

The real point was that mass production won out over artisanship because it was cheaper. Likewise generative AI will win out over human artists when it becomes cheaper and more efficient to ask a machine for something than to pay an artist to make it. Nobody has explained why employers would forgo the cheaper option of AI for the more expensive option of human workers or why AI couldn't improve in the future. As the recent Hollywood writer strike has shown, plenty of employers are already quite eager to replace humans with machines. Was everyone involved in the strike, whether writer or employer, simply pretending that AI could write scripts or does AI pose a legitimate threat to writers?

Rather off topic, but I suspect the main reason clothing manufacturing hasn't been more thoroughly automated is that most clothes are made by children in sweatshops getting 5¢ an hour. Corporations have little incentive to research automation when they can just exploit such cheap labor. There really is nothing magical about Homo sapiens that makes them uniquely able to cut sheets of cloth to size and push threaded needles through them. If malnourished toddlers earning almost nothing can assemble t-shirts, industrial robots certainly could as well.

(Incidentally I think you also have a very rosy view of the pre-industrial world...)

I said little about the pre-industrial world other than that all the products were made by hand rather than machine. That much is undeniable. There were no computerized lathes or printers or anything of that sort obviously. Wood and stone were worked by hand. The merits of artisanship versus mass production, let alone the Middle Ages versus now, are beside the point.

For what it's worth, I agree that the Left does tend toward undue nostalgia for the pre-modern world. Quite often I see people claiming that medieval peasants worked far less than modern people (while ignoring the absence of amenities like indoor plumbing and antibiotics). Others go even further and elevate hunter-gatherer societies as the height of human existence. That said, I feel like they're reacting against the prevailing social climate of valorizing technological innovation and assuming the superiority of Western modernity to the indigenous societies it destroyed. There is value in looking critically at institutions we take for granted such as technology and considering the possible downsides. Furthermore, some technologies like fossil fuels have been genuinely quite harmful and pose significant threats to humanity. Given the looming catastrophe of global warming, you can hardly blame some people for having doubts about industrialization.

[Ketsuban]

Rather off topic, but I suspect the main reason clothing manufacturing hasn't been more thoroughly automated is that most clothes are made by children in sweatshops getting 5¢ an hour. Corporations have little incentive to research automation when they can just exploit such cheap labor.

A machine that would take raw cloth as input and output T-shirts without human operation would operate on pennies of electricity a day. Exploitable labour is not the reason people haven't been completely automated out of clothing manufacture, it's a response to the reality that you can't avoid hiring people to do this job.

(Also, seriously, if you actually care about the treatment of workers in clothing manufacture consider making your own clothing. A huge downside of modern industrially-produced fitted clothing is the amount of cloth that's wasted; industrialisation has made it so cheap we don't think about it, but even home sewing patterns like the ones my mother gets for her hobby are designed the same way and that seems wrong to me. Traditional patterns like the ones Zomp shows in the Planet Construction Kit were designed to avoid waste and to some extent minimise labour.)

[Linguoboy]

The real point was that mass production won out over artisanship because it was cheaper. Likewise generative AI will win out over human artists when it becomes cheaper and more efficient to ask a machine for something than to pay an artist to make it. Nobody has explained why employers would forgo the cheaper option of AI for the more expensive option of human workers

Yes they have, you just haven't listened.

or why AI couldn't improve in the future.

No one has said AI couldn't improve; what we dispute is that it could improve to the degree necessary to produce your nightmare scenarios. The burden of proof is on your to demonstrate their plausibility.

Some of the Dadaists claimed[*] that photography would make traditional art obsolete because why bother drawing something when a machine can produce a perfect image of it? How has that prediction panned out?

[*] Whether they were trolling or not is left as an exercise for the student.

[malloc]

No one has said AI couldn't improve; what we dispute is that it could improve to the degree necessary to produce your nightmare scenarios. The burden of proof is on your to demonstrate their plausibility.

The very existence of humans with their respective level of intelligence demonstrates their plausibility. If humans could achieve our current level of intelligence with only matter and the laws of physics, AI could do the same. You must demonstrate that cognition and intelligence require something that physical substrates like neurons or silicon cannot provide. Absent that, we must assume that AI can match any cognitive abilities that humans have or even exceed them. If you have some bombshell revelations regarding the metaphysics of cognition, by all means present them.

Some of the Dadaists claimed[*] that photography would make traditional art obsolete because why bother drawing something when a machine can produce a perfect image of it? How has that prediction panned out?

They failed to consider the possibility of stylized and abstract art. Furthermore, the photography of their day could not even render color, which greatly limited its ability to replace even realistic artists. Today of course, photography can replicate real world scenes better than any artist, so practically nobody bothers trying. Generative AI can handle stylized and abstract art as easily as realistic images, however, allowing it to replace any artists in any genre and even copy their specific styles.

[Travis B.]

Some of the Dadaists claimed[*] that photography would make traditional art obsolete because why bother drawing something when a machine can produce a perfect image of it? How has that prediction panned out?

They failed to consider the possibility of stylized and abstract art. Furthermore, the photography of their day could not even render color, which greatly limited its ability to replace even realistic artists. Today of course, photography can replicate real world scenes better than any artist, so practically nobody bothers trying. Generative AI can handle stylized and abstract art as easily as realistic images, however, allowing it to replace any artists in any genre and even copy their specific styles.

There is the other possibility, that they were trolling.

[zompist]

The very existence of humans with their respective level of intelligence demonstrates their plausibility. If humans could achieve our current level of intelligence with only matter and the laws of physics, AI could do the same. You must demonstrate that cognition and intelligence require something that physical substrates like neurons or silicon cannot provide. Absent that, we must assume that AI can match any cognitive abilities that humans have or even exceed them. If you have some bombshell revelations regarding the metaphysics of cognition, by all means present them.

Bombshell revelation: cognition is really, really hard. AI researchers have been trying for 70 years, and they're not even close.

Interstellar travel is "plausible" too: we can see gases traveling millions of light years from stellar explosions. Doesn't mean Elmo Musk can create it tomorrow.

There are a lot of things to worry about with LLMs, but a) they are not AI; b) there is already evidence that the alarmism is overdone; c) it is far from costless; d) those opposing the over-use of LLMs are not powerless. Also, alarmism about general AI actually plays into the hands of the techbros. They hope to deflect attention from LLMs, and create a regulatory environment they can control, by getting people worked up over nonexistent general AI.

Your concerns are, for once, not baseless: LLMs can do real harm. But you don't educate yourself on the actual costs or technology, and you don't seem to understand the difference between LLMs and human cognition.

[malloc]

Bombshell revelation: cognition is really, really hard. AI researchers have been trying for 70 years, and they're not even close.

You yourself said that with the development of LLMs, they were basically halfway to AI with human level intelligence. Now you claim they aren't even close. Fundamentally the human brain is not that complicated, just an agglomeration of fleshy transistor analogs lacking the reliability of artificial transistors. The lingering influence of religion has us convinced that the human mind involves something beyond crude matter, but I see no evidence for such things.

There are a lot of things to worry about with LLMs, but a) they are not AI;

If you went 70 years back in time and told the first AI researchers about LLMs, would they agree with that assessment? Even putting aside drawing "art" in any style and writing screenplays, computers these days can match or even exceed humans at pretty much any intellectual feat. They can perform any mathematical calculation as easily as we breath. They effortlessly beat even the greatest chess masters. They can translate texts between hundreds of languages instantaneously. Quite honestly, I think that if you listed everything modern computers can do to those original AI researchers, they would conclude that we already have artificial intelligence.

b) there is already evidence that the alarmism is overdone;

Then you believe the Hollywood writers were wrong to strike because they incorrectly assumed LLMs could replace them? Numerous writers and artists are terrified of this technology destroying their livelihoods. Are all those people really just alarmists?

c) it is far from costless;

Neither are human writers and artists. But the price of LLMs will presumably decrease like any other technology, whereas the price of human labor remains constrained by our inherent need for survival goods. Not to mention most people will not tolerate ever shrinking wages to compete with tireless machines.

Also, alarmism about general AI actually plays into the hands of the techbros. They hope to deflect attention from LLMs, and create a regulatory environment they can control, by getting people worked up over nonexistent general AI.

Not sure what you mean by that. I have certainly never claimed that we shouldn't worry about LLMs because more powerful forms of AI are even more dangerous. Quite the opposite, I have emphasized the need to address the threat to our culture and economy that LLMs pose. Nonetheless technology develops by building on simpler predecessors. LLMs are one step in the development of even more powerful AI so it makes sense to consider such future developments in my critique.

[zompist]

Bombshell revelation: cognition is really, really hard. AI researchers have been trying for 70 years, and they're not even close.

You yourself said that with the development of LLMs, they were basically halfway to AI with human level intelligence. Now you claim they aren't even close.

Halfway is impressive on the timescale of AI: if they keep it up, they might reach AI by 2100.

Or maybe not. Like a lot of hard problems in science, intelligence is fractal. You solve one problem, and another comes up. If just throwing money at the problem worked, we'd have had it fifty years ago.

Fundamentally the human brain is not that complicated,

Do you know a single thing about the brain? Tell you what, build an AI out of Legos if it's so simple. Legos are "crude matter", right?

There are a lot of things to worry about with LLMs, but a) they are not AI;

If you went 70 years back in time and told the first AI researchers about LLMs, would they agree with that assessment?

Who cares? AI researchers and observers have a piss-poor record in predicting how close we are to intelligence. Do you recall that sf writers have thought it was within the range of possibility for 200 years?

b) there is already evidence that the alarmism is overdone;

Then you believe the Hollywood writers were wrong to strike because they incorrectly assumed LLMs could replace them? Numerous writers and artists are terrified of this technology destroying their livelihoods. Are all those people really just alarmists?

You're making up bullshit I never said. Is that supposed to fool me?

The Hollywood writers were concerned about LLMs for rational reasons, and they won. This is not in any way a vindication of your fears.

c) it is far from costless;

Neither are human writers and artists. But the price of LLMs will presumably decrease like any other technology,

I already posted something about the costs of LLMs; go find it, or actually do some research for once.

Also, alarmism about general AI actually plays into the hands of the techbros. They hope to deflect attention from LLMs, and create a regulatory environment they can control, by getting people worked up over nonexistent general AI.

Not sure what you mean by that. I have certainly never claimed that we shouldn't worry about LLMs because more powerful forms of AI are even more dangerous. Quite the opposite, I have emphasized the need to address the threat to our culture and economy that LLMs pose. Nonetheless technology develops by building on simpler predecessors. LLMs are one step in the development of even more powerful AI so it makes sense to consider such future developments in my critique.

Not really, it lessens your critique and plays into techbro mindgames. You're diverted into arguing about nonexistent AIs instead of talking about the real problems of LLMs.

If you don't know what I'm talking about, read this. What the hell do you think is going on? A bunch of AI researchers are bringing up the exact general-AI alarmism that you are. Why is that? Do they want AI research, the stuff they do and are making money on, to be stopped?

Most likely it's a feint: they want "regulation" that will vaguely address the nonexistent threat but allow them to keep making money with their current and planned projects. So they reassure people like you ("general AI would destroy humanity!!1!") precisely in order to let them exploit LLMs to the max, and perhaps make it harder for others to compete with them.

[elgis]

To me it seems a bit pointless to argue whether or not LLM are AI. Depending on who you ask, even simple search algorithms are AI. This textbook, first published in 1995, has four chapters on search algorithms. I think part of the reason for the ambiguity (and the reason that researchers keep changing the criteria for intelligence) is that we don't really understand intelligence.

[zompist]

To me it seems a bit pointless to argue whether or not LLM are AI. Depending on who you ask, even simple search algorithms are AI. This textbook, first published in 1995, has four chapters on search algorithms. I think part of the reason for the ambiguity (and the reason that researchers keep changing the criteria for intelligence) is that we don't really understand intelligence.

If we don't understand intelligence (and I agree with you on that), then we don't have AI.

There's a simple test for anyone who's tempted to call LLMs intelligent: can we arrest the owners of ChatGPT for holding slaves?

[bradrn]

There's a simple test for anyone who's tempted to call LLMs intelligent: can we arrest the owners of ChatGPT for holding slaves?

This is a bit disingenuous: creatures beside humans are intelligent. I don’t say that dog-owners should be arrested for holding slaves, even though dogs are very intelligent.

Now, if by ‘intelligence’ you mean ‘human-level intelligence’, that’s a bit different… though even then it wouldn’t be comparable, since LLMs are not human.

[malloc]

Do you know a single thing about the brain? Tell you what, build an AI out of Legos if it's so simple. Legos are "crude matter", right?

That hardly makes sense. Legos have none of the properties of neurons or transistors that make them useful for calculation. Neurons are quite crude and unreliable compared to transistors but they have similar abilities to conditionally conduct and block impulses. Much as it pains me to say it, though, brains really suck for the most part. Ask anyone who suffers from mental health issues or look at the prevalence of conspiracy theories.

You're making up bullshit I never said. Is that supposed to fool me?

The Hollywood writers were concerned about LLMs for rational reasons, and they won. This is not in any way a vindication of your fears.

Perhaps it was unclear but the first sentence was intended as a question. It sounded like you were dismissing concerns about LLMs as alarmism and I was asking if you considered the striking writers alarmist. It sounds like we both agree with the Hollywood writers that LLMs would take over their field, however, unless they forced employers to refrain from using them.

I already posted something about the costs of LLMs; go find it, or actually do some research for once.

Again, the current cost hardly matters in the long run. Fifty years ago only a select few institutions could afford computers at all and now everyone has one or several. The same with televisions and automobiles and so forth. Technology always starts out expensive and declines in price. You must explain why LLMs, uniquely among technologies, would buck this trend or else why its cost will never fall below decent wages for artists and writers.

Not really, it lessens your critique and plays into techbro mindgames. You're diverted into arguing about nonexistent AIs instead of talking about the real problems of LLMs.

If you don't know what I'm talking about, read this. What the hell do you think is going on? A bunch of AI researchers are bringing up the exact general-AI alarmism that you are. Why is that? Do they want AI research, the stuff they do and are making money on, to be stopped?

Most likely it's a feint: they want "regulation" that will vaguely address the nonexistent threat but allow them to keep making money with their current and planned projects. So they reassure people like you ("general AI would destroy humanity!!1!") precisely in order to let them exploit LLMs to the max, and perhaps make it harder for others to compete with them.

Fair enough, but it's hard not to think about the future of this technology. I am only in my thirties, still young enough that I could plausibly live to see the late 21st century. You seem to have the impression that I am attacking AGI with the intent of letting LLMs off the hook. Quite the contrary, I have been the staunchest critic of LLMs in this thread and certainly haven't been arguing that we should ignore them. The fact that successfully resisting LLMs would make it harder for more dangerous forms of AI to emerge is certainly a desirable bonus.

[zompist]

There's a simple test for anyone who's tempted to call LLMs intelligent: can we arrest the owners of ChatGPT for holding slaves?

This is a bit disingenuous: creatures beside humans are intelligent. I don’t say that dog-owners should be arrested for holding slaves, even though dogs are very intelligent.

There are animal cruelty laws too. Do those apply? When do any ethical considerations whatsoever apply to AI bros and their toys?

[zompist]

Do you know a single thing about the brain? Tell you what, build an AI out of Legos if it's so simple. Legos are "crude matter", right?

That hardly makes sense. Legos have none of the properties of neurons or transistors that make them useful for calculation. Neurons are quite crude and unreliable compared to transistors but they have similar abilities to conditionally conduct and block impulses. Much as it pains me to say it, though, brains really suck for the most part. Ask anyone who suffers from mental health issues or look at the prevalence of conspiracy theories.

This is a rather sad and strange move. If humans are so worthless, why are you worried that they could be replaced? By your value system it would be an improvement.

Brains are the most amazing three-pound machine we know of. I'm sorry you worship electronic devices instead. Without denying that computers are pretty neat toys, they are nowhere near as perfect, scary, or costless as you think they are. (And seriously, if you aren't aware of how prone LLMs are to misinformation, you are incredibly inattentive.)

[bradrn]

There's a simple test for anyone who's tempted to call LLMs intelligent: can we arrest the owners of ChatGPT for holding slaves?

This is a bit disingenuous: creatures beside humans are intelligent. I don’t say that dog-owners should be arrested for holding slaves, even though dogs are very intelligent.

There are animal cruelty laws too. Do those apply? When do any ethical considerations whatsoever apply to AI bros and their toys?

Well, I could argue that LLMs can’t feel pain or suffer, as far as we’re aware… but to address your broader argument: yes, I think ethical considerations will apply at some point, but I’m not going to hazard a guess as to when. We know too little to say anything about that.