AIs gunning for our precious freelancers

[Raphael]

Repeating something I had already posted earlier in this thread, I keep being a bit surprised about how the politics of AI have turned out to be almost exactly the opposite of what many people predicted, with left-wingers being generally sceptical about or downright hostile towards them, and a lot of right-wingers boosting them.

[Travis B.]

Repeating something I had already posted earlier in this thread, I keep being a bit surprised about how the politics of AI have turned out to be almost exactly the opposite of what many people predicted, with left-wingers being generally sceptical about or downright hostile towards them, and a lot of right-wingers boosting them.

The key thing is that what AI is ostensibly for is diametrically opposed to the interests of teh People (and I mean all of the people, not just some select portion of it) within the context of a capitalist society. The idea of a new technology that will simply render large numbers of people unnecessary, however correct or mistaken this view is, is a reason why many left-wingers are opposed to AI. Of course, this need not necessarily be so - within a socialist society AI could very well help ease the burden on the average worker, who will benefit from it - but we don't live in a socialist society, so that is all hypothetical. And sure, people may find new jobs, but this is a pattern that has happened before, and we know who benefits from it (i.e. not us).

[WeepingElf]

I would say that the sooner all the billionaires ship themselves to Mars and die of bad planning and safety standards the better.

I actually have plans for a novel about such a scenario in the back of my mind. Earth has made the transition to global democracy and a sustainable lifestyle, but some rich people are not content with that and try to set up a transhumanist libertarian utopia on Mars. Of course, terraforming fails, the colony attracts other misfits such as terrorists of various stripes, and the result is utter mayhem. A dictator seizes power over the red planet and tries to attack Earth

More: show

but fortunately, that move fails.

[Raphael]

I would say that the sooner all the billionaires ship themselves to Mars and die of bad planning and safety standards the better.

I actually have plans for a novel about such a scenario in the back of my mind. Earth has made the transition to global democracy and a sustainable lifestyle, but some rich people are not content with that and try to set up a transhumanist libertarian utopia on Mars. Of course, terraforming fails, the colony attracts other misfits such as terrorists of various stripes, and the result is utter mayhem. A dictator seizes power over the red planet and tries to attack Earth

Sounds interesting. Potential for a great novel if the execution is good. Any space habitat founded by libertarians might come up with interesting rhetorical contortions when it keeps trying to pay lip service to libertarian principles while having to institute very un-libertarian rules in order to survive.

[WeepingElf]

Thank you!

[linguistcat]

I actually have plans for a novel about such a scenario in the back of my mind. Earth has made the transition to global democracy and a sustainable lifestyle, but some rich people are not content with that and try to set up a transhumanist libertarian utopia on Mars. Of course, terraforming fails, the colony attracts other misfits such as terrorists of various stripes, and the result is utter mayhem. A dictator seizes power over the red planet and tries to attack Earth

More: show

but fortunately, that move fails.

I hope I can read it someday, if it gets finished.

[WeepingElf]

Thank you for your anticipation. I may mention that I have decided a few minutes ago that the Martian Führer isn't even human - it's an AI called the "Archilect". Also, the novel will feature a conlang, "Marban" - a kind of Lojban-based creole, basically Lojban lexicon and English grammar, with some lexical idiosyncrasies, such as words referring to aspects of life on Mars (it isn't elaborated yet). The Martian super-nerds decide to adopt Lojban as their official language, but few if any manage to speak it correctly, so a creole emerges. Alas, I think I'll write at least two other novels before that - one about how Earth becomes the Green-liberal semi-utopia it is in the Mars novel, and one turning around an ancient manuscript which turns out to be written in Old Albic.

[Ares Land]

I dunno about your "trillions" number. Has anyone done the math on that?

The tech industry is on the order of 5 trillion (globally) indeed. I suspect a good amount of this is spent on boring but practical applications though. But there's something around 200-300 billion spent a year on AI globally, which is pretty impressive.

For reasons I cannot quite understand, the tech industry is obsessed with brining about the Singularity and has trillions at its disposal to spend on research.

Working on the Next Big Thing is kind of a status symbol. As a rockstar developper, of course you're an expert on the Next Big Thing. As a Big Important Tech Company of course you're market leader on the Next Big Thing.
Add to this the fear of missing out. As a developer, you're afraid to be passed over from promotion, or have trouble changing jobs... As a company, of course, you're afraid of being overtaken by competitors.
We should not overlook the fact that the Next Big Thing is often genuinely cool too! GenAI is intellectually fascinating.

I think the real underlying reason, what ultimately fuels this whole cycle is very simple though. Some people just have way too much money on their hands. Above a certain level, there's just too much money around.

Interestingly, you can analyse this from a left-wing or a right-wing perspective, and both analyses are probably right.
The left-wing side: remember these huge profits? the 1%? Well, they have to put all that money somewhere. It doesn't all go into tech, of course, but a high-risk, high-growth industry is a great addition to a huge investment portfolio.
The right-wing side: to a politician, the tech industry looks pretty great, it sounds modern and all, plus it's always been a very healthy sector, so of course a lot of public money goes into it. (That's certainly true in the EU; don't know how it holds in the US. Plus AI being somewhat scary, the canny politician/high-ranking civil servant can also make grave statements and promise regulations, which further promotes and inflates the bubble.) Then there's the money supply, which is generally growing, and the extra cash must be injected somewhere in the economy.

[Raphael]

I dunno about your "trillions" number. Has anyone done the math on that?

The tech industry is on the order of 5 trillion (globally) indeed. I suspect a good amount of this is spent on boring but practical applications though. But there's something around 200-300 billion spent a year on AI globally, which is pretty impressive.

Thank you!

Working on the Next Big Thing is kind of a status symbol. As a rockstar developper, of course you're an expert on the Next Big Thing. As a Big Important Tech Company of course you're market leader on the Next Big Thing.
Add to this the fear of missing out. As a developer, you're afraid to be passed over from promotion, or have trouble changing jobs... As a company, of course, you're afraid of being overtaken by competitors.

This reminds me, I've thought for a while that Apple originally introduced Siri because they wanted to present themselves as "the company that invents the future", or something like that, and by the time Siri got developed, devices that people can talk to had been a stereotypical part of "The Future" for so long.

[malloc]

With all that money and enthusiasm being pushed toward AI, none can doubt that it will only get more powerful in the coming decades. Human writers and artists are already struggling to compete, regardless of the many denials I read here. When the technology becomes even more powerful and cheaper, they will be driven out of their fields entirely. Remember that technology keeps improving but humans have changed very little since they first evolved some quarter million years ago. Computers keep getting new abilities and more power while we must make the best of whatever finite endowments evolution has given us. Every year the machines get closer to our abilities, forcing us to work even harder for less just to retain jobs that once belonged to our species by default. What happens when they match or even exceed our abilities?

[Ares Land]

With all that money and enthusiasm being pushed toward AI, none can doubt that it will only get more powerful in the coming decades.

I've worked in tech for twenty years now and I can attest, from first hand experience, that the sector can spend extraordinary amounts of money, energy and material resources into bullshit. Okay, applications that turn out not to have that much of an impact on real life. Okay, bullshit's still the best word.

What really troubles me is the huge sums diverted towards bullshit (apologies for the language, but bullshit is the appropriate word), while the useful part of the economy is chronically underfunded.

Every year the machines get closer to our abilities, forcing us to work even harder for less just to retain jobs that once belonged to our species by default. What happens when they match or even exceed our abilities?

The trouble is... they don't really. Around here, we have a shortcare of doctors, nurses, daycare workers, craftspeople, teachers, and a host of assorted professions that are definitely not being replaced by machines. Never heard of AI doing any of that. (Well, okay, supposedly AI can handle diagnoses; but the story about the cancer-diagnosing AI is ten years old at this point and I never heard about any real-life practical applications; it turns out being a doctor is more complex than looking at chest X-Rays, who could've guessed that?)

This relates to the 'bullshit economy' problem I outlined above in a fairly obvious way: the daily rate for a GenAI consultant is about equivalent to the monthly wage of a daycare worker.

[malloc]

The trouble is... they don't really. Around here, we have a shortcare of doctors, nurses, daycare workers, craftspeople, teachers, and a host of assorted professions that are definitely not being replaced by machines. Never heard of AI doing any of that. (Well, okay, supposedly AI can handle diagnoses; but the story about the cancer-diagnosing AI is ten years old at this point and I never heard about any real-life practical applications; it turns out being a doctor is more complex than looking at chest X-Rays, who could've guessed that?)

People once said the same thing about art and literature. The first time I raised the problem of automation many years ago, everyone assured me that we'd always have creative jobs if nothing else. Then everyone was proven dead wrong. Replacing physicians with AI seems quite simple in principle. Computers have already made considerable advances in learning how to diagnose as you said. Knowing what to prescribe for a given condition seems equally simple, essentially a matter of looking up the appropriate medication and treatment. Robotic surgery has also made considerable headway. We might still need someone to coördinate the process physically but that would be an unskilled job hardly worth calling "doctor".

Regarding teachers, if AI progresses to the point of taking over all intellectual work, then humans would hardly need education. It would hardly make economic or practical sense to teach humans anything when machines do all our thinking for us. Even assuming we still need education for whatever reason, automating that seems trivial. Plenty of people teach themselves quite a lot from books after all.

[Linguoboy]

With all that money and enthusiasm being pushed toward AI, none can doubt that it will only get more powerful in the coming decades.

Seems like a safe bet.

Human writers and artists are already struggling to compete, regardless of the many denials I read here.

Are they though? Where's your non-anecdotal evidence? The WGA just ended their strike having gained basically everything they were asking for, including limits on studios' use of AI. That's only one data point, but it's a hugely important one.

When the technology becomes even more powerful and cheaper, they will be driven out of their fields entirely.

Here's where your argument goes completely off the rails. Mass-produced pottery has been around for centuries. Are there no more artisanal potters left? What about glass-blowers, carpenters, weavers, tailors, leatherworks, papermakers, etc. etc.? Sure, relatively few folks go into these fields compared to, say, insurance or retail. But how many people want to spend the years required to master these skills? How many did it in the past when there wasn't any competition from industrial-scale production of these goods?

[Travis B.]

Personally, as a professional computer programmer, I feel completely safe from AI at this point. What AI can do at this point requires much handholding, such that to do non-trivial tasks you are better off just paying a programmer to do it than paying an AI-herd to coax an AI to do the same with much more effort and cost. For instance, I program heavily in Forth (not at my day job but on my own). I have yet to see ChatGPT generate a truly correct factorial function in Forth. 90% correct, yes, but something always needs to be fixed before it will work. It should be noted that a factorial function is truly trivial, and any half-decent human programmer should be able to write one without a problem. So if ChatGPT can't do a factorial function right, what does that say about its doing more advanced programming?

[Ares Land]

Personally, as a professional computer programmer, I feel completely safe from AI at this point.

Yes, programming is safe. In addition to the limitations you just mentioned... How much of programming is spent actually writing code, as opposed to figuring out what exactly is needed, how best to interface with other systems, what would be the most efficient approach/infrastructure, figuring out weird bugs, fixing unexplainable production incidents?

A joke doing the rounds is that managers are learning to write exceedingly precise requirements to ChatGPT; something we IT folks have been waiting for for decades

[Raphael]

Well, back in February, I wrote

Back in the late 1990s, in school, they had an event - not sure what they'd call it in the USA - where students' parents talked about their jobs, and what they did in those jobs, and how to get those jobs. I went to a talk by a father who worked in IT. He told us that wanting to become a programmer was a bad idea, because computers would soon program themselves.

Seems now that that statement was a bit premature.

[Travis B.]

Personally, as a professional computer programmer, I feel completely safe from AI at this point.

Yes, programming is safe. In addition to the limitations you just mentioned... How much of programming is spent actually writing code, as opposed to figuring out what exactly is needed, how best to interface with other systems, what would be the most efficient approach/infrastructure, figuring out weird bugs, fixing unexplainable production incidents?

A joke doing the rounds is that managers are learning to write exceedingly precise requirements to ChatGPT; something we IT folks have been waiting for for decades

Precisely. For what I do at my day job, actual programming is a small portion of what I do, and even what I do outside my job, much more of what I do is testing my code than actually writing it.

[malloc]

Here's where your argument goes completely off the rails. Mass-produced pottery has been around for centuries. Are there no more artisanal potters left? What about glass-blowers, carpenters, weavers, tailors, leatherworks, papermakers, etc. etc.? Sure, relatively few folks go into these fields compared to, say, insurance or retail. But how many people want to spend the years required to master these skills? How many did it in the past when there wasn't any competition from industrial-scale production of these goods?

Realistically though, how much of our clothing and pottery comes from traditional artisans? We are looking at a future where humans write only 1% of novels and screenplays, compose 1% of songs, and so forth. Functionally art will no longer represent a human endeavor and humans will no longer create their own culture. Even aside from the jobs lost, the radical alienation of humans from their own culture alarms me.

Are they though? Where's your non-anecdotal evidence? The WGA just ended their strike having gained basically everything they were asking for, including limits on studios' use of AI. That's only one data point, but it's a hugely important one.

But the fact that studios have the option of replacing writers with AI merely proves how powerful it has gotten. Imagine explaining to someone a decade ago that studios would seriously consider replacing writers with computers and only months of struggle prevented them.

Personally, as a professional computer programmer, I feel completely safe from AI at this point. What AI can do at this point requires much handholding, such that to do non-trivial tasks you are better off just paying a programmer to do it than paying an AI-herd to coax an AI to do the same with much more effort and cost. For instance, I program heavily in Forth (not at my day job but on my own). I have yet to see ChatGPT generate a truly correct factorial function in Forth. 90% correct, yes, but something always needs to be fixed before it will work. It should be noted that a factorial function is truly trivial, and any half-decent human programmer should be able to write one without a problem. So if ChatGPT can't do a factorial function right, what does that say about its doing more advanced programming?

Aren't you worried about AI improving to the point of taking over programming, though. Software is the natural environment of AI after all and cold inerrant rationality is its natural mode of thought. If it can master the squishy and messy world of art, then the purely rational world of software engineering seems pretty trivial to master as well.

[Travis B.]

Personally, as a professional computer programmer, I feel completely safe from AI at this point. What AI can do at this point requires much handholding, such that to do non-trivial tasks you are better off just paying a programmer to do it than paying an AI-herd to coax an AI to do the same with much more effort and cost. For instance, I program heavily in Forth (not at my day job but on my own). I have yet to see ChatGPT generate a truly correct factorial function in Forth. 90% correct, yes, but something always needs to be fixed before it will work. It should be noted that a factorial function is truly trivial, and any half-decent human programmer should be able to write one without a problem. So if ChatGPT can't do a factorial function right, what does that say about its doing more advanced programming?

Aren't you worried about AI improving to the point of taking over programming, though. Software is the natural environment of AI after all and cold inerrant rationality is its natural mode of thought. If it can master the squishy and messy world of art, then the purely rational world of software engineering seems pretty trivial to master as well.

There is much more to computer programming than simply being able to spit out code. There is design and architecture. There is documentation. There is integration. There is testing, verification, and validation. There is responding to bug reports and comments from the user. And all of these tie in closely to one another. This applies to both in the professional world and outside it (even I have to respond to comments from others in my own work, as I have actual users at this point). Even if AI's in the future generate better code than they do today, they won't be able to do all this.

[Linguoboy]

We are looking at a future where humans write only 1% of novels and screenplays, compose 1% of songs, and so forth. Functionally art will no longer represent a human endeavor and humans will no longer create their own culture.

You've offered literally no evidence for this. It's nothing more than a blanket assertion of your worst fears and, thus, not worthy of further engagement.

Are they though? Where's your non-anecdotal evidence? The WGA just ended their strike having gained basically everything they were asking for, including limits on studios' use of AI. That's only one data point, but it's a hugely important one.

But the fact that studios have the option of replacing writers with AI merely proves how powerful it has gotten. Imagine explaining to someone a decade ago that studios would seriously consider replacing writers with computers and only months of struggle prevented them.

What it doesn't prove, however, is that consumers would actually prefer AI-written media. The studio execs think it would be passable enough that it's worth gambling on and alienating their creatives in the process, but these same studio execs have a pretty dismal record when it comes to demonstrating any understanding of the appeal of human creativity. They can--and do--make bad decisions with the goal of maximising profits all the time.