AIs gunning for our precious freelancers

[malloc]

I think you have to refine what you actually want to know.

* How the Incatena developed? Start with my web page, or ask in the Almea forum.
* How conworlders/SF authors envision a better future? Read more; Iain Banks is the classic start. Or start a thread in Conlangery.
* What is likely to happen in this century? A buncha stuff; start a new thread here if you really want to know. (But ask, don't preach your own dystopia.)
* What should be done about computronium? The facile answer is "burn it with fire." Less flippantly: this is conworlding, and your nightmare projections are no better than anyone else's-- nor are they even the most important possible apocalypse.

I already said what I wanted to know, namely how organic beings can reliably defeat computronium ones given their greater efficiency and productivity. The only plausible answer I can imagine is that the Incatena neutralizes computronium before its users have the chance to capitalize on its inherent advantages while also tightly regulating the materials used in its construction much like nuclear materials today.

[zompist]

I already said what I wanted to know, namely how organic beings can reliably defeat computronium ones given their greater efficiency and productivity. The only plausible answer I can imagine is that the Incatena neutralizes computronium before its users have the chance to capitalize on its inherent advantages while also tightly regulating the materials used in its construction much like nuclear materials today.

OK. First, you're aware computronium doesn't exist? It's a speculative fiction. Your question is basically asking how the US government should store its kryptonite, Green Lanterns, and Rings of Power.

Second, you're aware that it's matter? It's a hypothetical substance that provides the densest possible substrate for computation. What happens if you fire a rifle into it? A machine gun? A nuke? A relativistic-speed projectile? It's optimized for computation, not defense.

Third, the supposed "greater efficiency" is your supposition. You want to scare yourself by imagining a process with no obstacles that can stop it. Why not worry about fruit flies instead? If nothing stops them, fruit flies will multiply till they take every bit of matter on the Earth. Somehow this doesn't happen. Unstoppable processes do not exist.

You're making a shitload of other assumptions too, such as that laissez-faire capitalism persists, that techbros stay the same forever, that humans and machines are separate, that the rule of law is impossible, that organic life is inferior, that computers are "efficient", that people want to be turned into computronium, that singularitarians can survive a climate catastrophe, etc., etc. I don't have time to list out everything for you and refute it.

[WeepingElf]

It makes sense to me that natural human languages should present roughly equal difficulties to human beings, as an atypically easy one would have scope to add more difficulty. However, should I be unsurprised that LLMs don't have a different difficulty ranking to humans?

They already do: it's based on the size of the training data. Humans don't need that much raw data; they evidently have tricks up their crania that LLMs don't have.

People tend to underestimate the amount of "training data" children derive their native language from. They don't just use what the adults talk to them, but everything the adults (and older children) utter in their presence. And that over about six years - that's quite big a corpus.

[malloc]

OK. First, you're aware computronium doesn't exist? It's a speculative fiction. Your question is basically asking how the US government should store its kryptonite, Green Lanterns, and Rings of Power.

Sure but there are multiple trillion-dollar corporations and numerous venture capitalists trying to veritable Rings of Power right now. They've already made the palantír and they won't stop there. More seriously, though, many fantastically wealthy people and organizations are obsessed with bringing about the Singularity and throwing vast resources at AI research. We can stop them but we have to start now rather than waiting until the AIs are already fooming.

[WeepingElf]

OK. First, you're aware computronium doesn't exist? It's a speculative fiction. Your question is basically asking how the US government should store its kryptonite, Green Lanterns, and Rings of Power.

Sure but there are multiple trillion-dollar corporations and numerous venture capitalists trying to veritable Rings of Power right now. They've already made the palantír and they won't stop there. More seriously, though, many fantastically wealthy people and organizations are obsessed with bringing about the Singularity and throwing vast resources at AI research. We can stop them but we have to start now rather than waiting until the AIs are already fooming.

We all concur with you that the techbros' love affair with AI is worrying, and something has to be done about it. But as for now, it is the techbros we should be worried about rather than the AIs themselves. AI may be a powerful tool, but it is a tool, not a superpowered entity, and can be used for harm - but it is the people who use it for harm who are to blame.

[keenir]

. More seriously, though, many fantastically wealthy people and organizations are obsessed with bringing about the Singularity and throwing vast resources at AI research.

can evil people make something that is more good than they are? or does their evil cause them to fail?

We can stop them but we have to start now rather than waiting until the AIs are already.

why bother? you're right - humans are too evil to be worth saving. tis better to just create SkyNet and get it over with. theres nothing good about any humans except the few in conlang forums.

[rotting bones]

So is part of the challenge that there are in fact several seemingly very different Bengali languages for LLMs to get to grips with, and they're not well-labelled, so the LLM doesn't immediately know which it is dealing with?

Yes. The media I've seen in Unicode include song lyrics consisting of a few paragraphs each, and titles of web pages and YouTube videos. As for users typing in Unicode, I still remember only journalists doing that. Also:

* Identifying the language embedded in graphics and using OCR on each one seems expensive at the training step. I'm not sure how far OCR technology has progressed on Bengali.

* LLMs are much more dependent on context than humans. Like I keep telling Eddy, humans evolved their abilities over millions of years. There don't seem to be as many discussions of Bengali grammar online as for other languages with a comparable number or even fewer speakers. Bengali's currency seems to have gone down in my lifetime. Maybe ChatGPT's performance will improve if our native grammar nazis continue to thrive, finally figure out how to use the internet and adopt Unicode en masse.

Bengali was a much more formalized language than either Hindi or English when I was growing up. I've known people to transition away from it as an act of rebellion. (Not that it's necessarily formal. Bangla Rock has been a thing for decades now.)

[malloc]

We all concur with you that the techbros' love affair with AI is worrying, and something has to be done about it. But as for now, it is the techbros we should be worried about rather than the AIs themselves. AI may be a powerful tool, but it is a tool, not a superpowered entity, and can be used for harm - but it is the people who use it for harm who are to blame.

Sure but my point is that we really ought to address this issue now when it's still manageable rather than waiting until AI becomes too powerful. Many people here believe that AI development will stall soon, preventing AI from overtaking humans or replacing them in the economy for the foreseeable future. However nothing guarantees that outcome and we must consider the equally likely possibility that AI development continues its current trajectory until it leaves humanity in the dust. Everyone here once assured me that automation would only target bad jobs that nobody really likes while leaving good jobs like artist and scientist alone and we can all see how that prediction turned out. Why should I accept predictions from the same people that AI will soon hit a wall?

[Travis B.]

You're missing the point, malloc -- the fundamental problem here is the techbros' use of AI, not AI itself. AI is not going to put people out of work by itself by any means -- it's the techbros deciding that they can replace humans with AI that is the problem. In the end, the problem is people not AI.

[Ketsuban]

Sure but there are multiple trillion-dollar corporations and numerous venture capitalists trying to veritable Rings of Power right now.

I thought I was a catastrophiser, but Jesus Christ would you stop shortcutting straight from "people with more money than sense are throwing money at a stupid idea" to "they're going to achieve the stupid idea oh god we're all going to die"?

They've already made the palantír and they won't stop there.

They made a company with a name taken from a book they didn't actually read. They have not in fact created a MacGuffin from a fantasy novel.

More seriously, though, many fantastically wealthy people and organizations are obsessed with bringing about the Singularity and throwing vast resources at AI research. We can stop them but we have to start now rather than waiting until the AIs are already fooming.

Did you perhaps look at the cited uses of the word whose Wiktionary definition you linked? One says "some advocates of this […] scenario" (and before you start, advocate here means "proponent", not "cheerleader") and points out they're talking nonsense with no evidentiary basis, and the other straight-up says "there are no signs of [it]". This is not a projected course for the development of AI, this is a word bandied about by idiots who think they're clever because they read one more science fiction book than the median person.

[alice]

They've already made the palantír and they won't stop there.

They made a company with a name taken from a book they didn't actually read. They have not in fact created a MacGuffin from a fantasy novel.

And in any case there were seven palantíri, not just one. What substance do you think this palantír is made of? Computronium?

[Raphael]

Again, part of the problem with malloc is that he thinks in clichés from popular fiction.

malloc, do you remember how much money Facebook used to pump into the Metaverse, even renaming itself after it, and telling us all that it would be The Future Of Everything?

[malloc]

I thought I was a catastrophiser, but Jesus Christ would you stop shortcutting straight from "people with more money than sense are throwing money at a stupid idea" to "they're going to achieve the stupid idea oh god we're all going to die"?

But nobody has explained what makes it unworkable like cold fusion or perpetual energy. They gesture at our limited understanding of human cognition while ignoring that cognition derives entirely from natural processes amenable to scientific study. Given enough time and money, we could create machines that replicate and even exceed human cognition just as we created machines that fly better than any bird.

They made a company with a name taken from a book they didn't actually read. They have not in fact created a MacGuffin from a fantasy novel.

Well yeah, that was intended as a joke.

You're missing the point, malloc -- the fundamental problem here is the techbros' use of AI, not AI itself. AI is not going to put people out of work by itself by any means -- it's the techbros deciding that they can replace humans with AI that is the problem. In the end, the problem is people not AI.

Sure but one could make the same point about numerous other dangerous technologies like guns or nuclear weapons. Even the most powerful nuke does nothing until someone drops it after all. Nonetheless we also recognize the importance of regulating dangerous technologies and even working toward their abolition if necessary.

[Travis B.]

You're missing the point, malloc -- the fundamental problem here is the techbros' use of AI, not AI itself. AI is not going to put people out of work by itself by any means -- it's the techbros deciding that they can replace humans with AI that is the problem. In the end, the problem is people not AI.

Sure but one could make the same point about numerous other dangerous technologies like guns or nuclear weapons. Even the most powerful nuke does nothing until someone drops it after all. Nonetheless we also recognize the importance of regulating dangerous technologies and even working toward their abolition if necessary.

The difference is that guns and nuclear weapons by their very nature are designed to kill and destroy, while AI is only a negative force when it is specifically used as such by the capitalists. AI by itself is neutral -- take what they use AI for at my job, which is enhancing MR images to provide better quality results than a radiologist could provide alone -- this is a positive use of AI.

[malloc]

The difference is that guns and nuclear weapons by their very nature are designed to kill and destroy, while AI is only a negative force when it is specifically used as such by the capitalists. AI by itself is neutral -- take what they use AI for at my job, which is enhancing MR images to provide better quality results than a radiologist could provide alone -- this is a positive use of AI.

Except that the people who spent thousands of dollars studying radiology now find themselves unemployable and saddled with debt. An entire field of employment, both well-paying and prestigious, evaporates with nothing to replace it. Artificial intelligence by its very nature replaces humans in the economy and other institutions. The logical conclusion of AI is humans expelled from the arts and sciences, academia and administration, reduced to starvation from mass unemployment or passive dependence on whatever allowance our AI overlords deign to provide.

[keenir]

We all concur with you that the techbros' love affair with AI is worrying, and something has to be done about it. But as for now, it is the techbros we should be worried about rather than the AIs themselves. AI may be a powerful tool, but it is a tool, not a superpowered entity, and can be used for harm - but it is the people who use it for harm who are to blame.

Sure but my point is that we really ought to address this issue now when it's still manageable rather than waiting until AI becomes too powerful.

and yet you don\t want to do this with Trump&friends.

[keenir]

The difference is that guns and nuclear weapons by their very nature are designed to kill and destroy, while AI is only a negative force when it is specifically used as such by the capitalists. AI by itself is neutral -- take what they use AI for at my job, which is enhancing MR images to provide better quality results than a radiologist could provide alone -- this is a positive use of AI.

Except that the people who spent thousands of dollars studying radiology now find themselves unemployable and saddled with debt.

for fock sake, Malloc, read the bloody line you're pretending to reply to.

[Lērisama]

AI by itself is neutral -- take what they use AI for at my job, which is enhancing MR images to provide better quality results than a radiologist could provide alone -- this is a positive use of AI.

Except that the people who spent thousands of dollars studying radiology now find themselves unemployable and saddled with debt.

(Emphasis in Travis's post mine)

Where here does it say the AI is doing what you claim it is? A human is still involved in the decision process; a human is still doing the decision they have been qualified to make: the AI is assisting rather than replacing. Lives are being saved and no humans are losing their jobs. What is so bad about this?

Edit: beaten in pointing this out by keenir

---

Incidentally, I remember watching a Horizon¹ episode from the 80s about microchips, which confidentiality stated that they were “why our children will grow up without jobs.” On the plus side, malloc is not alone the in assuming new technology X will lead to our inevitable unemployability and subjection to the machine, but on the other hand, maybe it's time to see the pattern for what it is now?

¹ A long running BBC science programme, of which many, many back episodes are² made available to watch
² Or were, a few years ago, when I watched this one

[malloc]

Where here does it say the AI is doing what you claim it is? A human is still involved in the decision process; a human is still doing the decision they have been qualified to make: the AI is assisting rather than replacing. Lives are being saved and no humans are losing their jobs. What is so bad about this?

Sure but many AI advocates are working on models that make human workers completely superfluous as opposed to merely assisting them with drudgery. You can point to AI models that politely limit themselves to enhancing images and letting humans make the final call. Meanwhile many in the tech industry look forward to AI outdoing humans at everything and leaving us with no place in the economy.

[keenir]

Edit: beaten in pointing this out by keenir

I may have posted first, but you posted better.

Where here does it say the AI is doing what you claim it is? A human is still involved in the decision process; a human is still doing the decision they have been qualified to make: the AI is assisting rather than replacing. Lives are being saved and no humans are losing their jobs. What is so bad about this?

Sure but many AI advocates are working on models that make human workers completely superfluous as opposed to merely assisting them with drudgery.

really? who are doing that?

You can point to AI models that politely limit themselves to enhancing images and letting humans make the final call.

politeness has naught to do with anything. they are unable to do more than they are programmed for.

Meanwhile many in the tech industry look forward to AI outdoing humans at everything and leaving us with no place in the economy.

Elon Musk =/= many