AIs gunning for our precious freelancers

[alice]

Fair enough. I should have specified that the creators of AI are aiming for those goals.

No - they are aiming at solving particular problems, such as doing certain tasks cheaper than humans can.

I can't believe I'm siding with malloc in this thread for once, but, well, the more over-the-top ones of the tech, err, "leaders" are aiming for what malloc says.

Indeed they are, but that doesn't mean they'll ever get anywhere near there.

[keenir]

It hardly makes sense to treat AI as normal technology

...because it it barely even that.

since intelligence itself is hardly a normal property. Intelligence is what allowed humans to take over the world despite being among the weakest and least durable animals relative to our size.

I can understand if you've already forgotten our earlier conversations - my memory is crud as well; but if you're just ignoring it, thats very different.

we are more generalized than other species; that does not make us "least durable" (for one, its harder to make us go splat than most animals our size)

Artificial intelligence by contrast aims to compete with humans at their one strength, matching or exceeding us in creativity,

hasn't happened yet, won't happen any time soon.

reasoning, and other cognitive abilities.

the ecological niche of humans is maker of tools, shaper of enviroment; same as beavers.

The widespread adoption of AI would mean the exclusion of humans from academia, administration, and the arts,

...because the word of God-Emperor Malloc is inviolable law never to be disobeyed once Lord Trump places Malloc in command of the world. *sigh*
(sorry)

much like the automobile ended the use of horses for transportation.

so its weird that people still use horses for transportation and entertailment and a dozen other things. heck, i myself am a trained equestrian.

The best comparison for AI in my estimation is invasive species.

AI can barely survive outside of where its programmed to work; and you think it can invade other ecosystems - digital or otherwise?? thats not an AI, Malloc, you're thinking of Trojans, Couckoo Eggs, and computer viruses.

[keenir]

No - they are aiming at solving particular problems, such as doing certain tasks cheaper than humans can.

I can't believe I'm siding with malloc in this thread for once,

I think I'm feeling faint.

but, well, the more over-the-top ones of the tech, err, "leaders" are aiming for what malloc says.

sorry, I was thinking of the people doing the manual programming of AI programs, not the heads of companies laying out their visions. But also what Alice says is even more true than what i said.

[Travis B.]

since intelligence itself is hardly a normal property. Intelligence is what allowed humans to take over the world despite being among the weakest and least durable animals relative to our size.

I can understand if you've already forgotten our earlier conversations - my memory is crud as well; but if you're just ignoring it, thats very different.

we are more generalized than other species; that does not make us "least durable" (for one, its harder to make us go splat than most animals our size)

Humans can actually survive many things that would generally kill most other animals of their size. Humans can also chase down prey over long distances such that their prey die from exhaustion while the humans can keep on running. Oh and humans also have dextrous hands, something most other animals lack.

reasoning, and other cognitive abilities.

the ecological niche of humans is maker of tools, shaper of enviroment; same as beavers.

There are other animals which have been shown to be quite intelligent, such as dolphins, but they lack the ability of humans to deliberately modify everything around them.

[Raphael]

Oh and humans also have dextrous hands, something most other animals lack.

reasoning, and other cognitive abilities.

the ecological niche of humans is maker of tools, shaper of enviroment; same as beavers.

There are other animals which have been shown to be quite intelligent, such as dolphins, but they lack the ability of humans to deliberately modify everything around them.

https://theonion.com/dolphins-evolve-op ... 819565718/

[malloc]

Artificial intelligence by contrast aims to compete with humans at their one strength, matching or exceeding us in creativity,

hasn't happened yet, won't happen any time soon.

Meanwhile the world's richest people are investing billions into making AGI a reality. It may well turn out that they are all taking the wrong approach and won't succeed but you cannot so blithely assume that.

so its weird that people still use horses for transportation and entertailment and a dozen other things. heck, i myself am a trained equestrian.

Mere curiosities, far from the general means of transportation they once represented. Certainly it has been years since I last saw anyone rely on horses for transportation. Perhaps a handful of humans will manage to break into industries like art and science as curiosities, but AI would still dominate such fields.

AI can barely survive outside of where its programmed to work; and you think it can invade other ecosystems - digital or otherwise?? thats not an AI, Malloc, you're thinking of Trojans, Couckoo Eggs, and computer viruses.

Considering that generative AI has already invaded the fields of art and literature, certainly. Humans are currently fighting an uphill battle against AI to retain control over visual art with literature and music also increasingly falling under siege. Soon enough everything from science to law to political administration will have AI competing with humans.

[zompist]

Fair enough. I should have specified that the creators of AI are aiming for those goals.

No - they are aiming at solving particular problems, such as doing certain tasks cheaper than humans can.

I can't believe I'm siding with malloc in this thread for once, but, well, the more over-the-top ones of the tech, err, "leaders" are aiming for what malloc says.

Are they? The top techbros are also noted, or notorious, for warning about AI takeovers. E.g. Wikipedia's article on AGI says "The thesis that AI poses an existential risk for humans, and that this risk needs more attention, is controversial but has been endorsed in 2023 by many public figures, AI researchers and CEOs of AI companies such as Elon Musk, Bill Gates, Geoffrey Hinton, Yoshua Bengio, Demis Hassabis and Sam Altman."

Basically every statement from the techbros, and most claims about AGI, should be taken with a carload of salt. The techbros consistently exaggerate what AI can do; they are also extremely vague on what sort of mitigations they have in mind. As they're generally extremely libertarian, they seem to reject outside oversight of any kind. The fact that Google, Microsoft, and Twitter have fired their AI ethics teams suggest the actual level of concern that these companies have, i.e. none at all.

I strongly suspect that "AGI" does not actually mean general human-level capabilities; it's just marketing buzz, like "Web 3.0." For one thing, no company wants autonomous agents which decide on their own tasks and can refuse orders... and if they can't do those things, they aren't human-level. "AGI" means "the product we think we'll ship in 2 years, maybe 3."

And for another, the vast majority of jobs suitable for AI do not require human-level capabilities. As one article I read recently pointed out, a factory does not need a human-level robot to move components from one station to another; it can use another advanced technology called a "conveyer belt." Things like factory work use specialized industrial robots and don't need a machine that can also play chess or make coffee.

As I've said repeatedly, the danger is not that AI can replace most human jobs, it's that CEOs mistakenly think they can. And since CEOs have a lot of power, they can do things like replace customer service (already outsourced to India) with chatbots. This doesn't work well but few companies these days care about customer service.

[keenir]

Artificial intelligence by contrast aims to compete with humans at their one strength, matching or exceeding us in creativity,

hasn't happened yet, won't happen any time soon.

Meanwhile the world's richest people are investing billions into making AGI a reality.

you mean like how the world's richest people believed in the aether, or spent fortunes in pursuit of communication with ghosts? yeah, its shocking how rich people are a good barometer of what succeeds.

It may well turn out that they are all taking the wrong approach and won't succeed but you cannot so blithely assume that.

yes i can, and we keep telling you why we can assume that.

so its weird that people still use horses for transportation and entertailment and a dozen other things. heck, i myself am a trained equestrian.

Mere curiosities,

sure Malloc, you go use AI to round up some cattle to ensure Chicago trains don't run out of beef to deliver; I'll hold your paycheck til you come back with the job done.

far from the general means of transportation they once represented. Certainly it has been years since I last saw anyone rely on horses for transportation.

you probably don't see people deliver mail to your house, either. yet it happens without you seeing it.

Perhaps a handful of humans will manage to break into industries like art and science as curiosities, but AI would still dominate such fields.

based on your fears of six-fingered men, i suspect.

(yes, that is a double reference: one from film, one from AI art)

AI can barely survive outside of where its programmed to work; and you think it can invade other ecosystems - digital or otherwise?? thats not an AI, Malloc, you're thinking of Trojans, Couckoo Eggs, and computer viruses.

Considering that generative AI has already invaded the fields of art and literature, certainly. Humans are currently fighting an uphill battle against AI

...just like they are in a war against the artistry of elephants and pigs,

[malloc]

you mean like how the world's richest people believed in the aether, or spent fortunes in pursuit of communication with ghosts? yeah, its shocking how rich people are a good barometer of what succeeds.

Except that AGI is demonstrably possible given that intelligence arises from physical processes. The only thing that could render AGI impossible is Cartesian dualism with the mind existing in some supernatural realm that machines cannot replicate. Given enough time and spending, the tech industry will figure out AGI eventually.

...just like they are in a war against the artistry of elephants and pigs,

Can you point to any examples of businesses shunning human artists in favor of trained elephants or pigs?

[keenir]

you mean like how the world's richest people believed in the aether, or spent fortunes in pursuit of communication with ghosts? yeah, its shocking how rich people are a good barometer of what succeeds.

Except that AGI is demonstrably possible given that intelligence arises from physical processes.

so is life, yet we also can't replicate abiogenesis.

The only thing that could render AGI impossible is Cartesian dualism with the mind existing in some supernatural realm that machines cannot replicate.

the only other thing that could do it, is AGI being impossible.

Given enough time and spending, the tech industry will figure out AGI eventually.

weird, thats what they said about cold fusion and uploading brains to computers and getting to Mars in the '70s.

...just like they are in a war against the artistry of elephants and pigs,

Can you point to any examples of businesses shunning human artists in favor of trained elephants or pigs?

so now we're talking businesses rather than techbros or internet sites? Goalboast-shifting noted. Probably the same ones who are manned entirely by AIs...you know, Warner Brothers, Universal, Marvel. Oh wait, those are staffed by human artists. Underpaid ones, probably, but still humans.

{also, not sure you've ever seen what elephants and pigs paint, so my analogy once more flew over your head}

[bradrn]

It hardly makes sense to treat AI as normal technology since intelligence itself is hardly a normal property. Intelligence is what allowed humans to take over the world despite being among the weakest and least durable animals relative to our size. Previous forms of technology have compensated for the natural weaknesses of humanity: spears for our blunt nails, knives for our dull teeth, automobiles for our sluggish legs, and so forth. Artificial intelligence by contrast aims to compete with humans at their one strength, matching or exceeding us in creativity, reasoning, and other cognitive abilities. The widespread adoption of AI would mean the exclusion of humans from academia, administration, and the arts, much like the automobile ended the use of horses for transportation. The best comparison for AI in my estimation is invasive species.

You’re misunderstanding. I — and the article — am not referring to vague philosophical claims around what ‘humanity’s one strength’ might be; I’m talking about very concrete practical claims about what AI can do and how it can be integrated safely into the world. To repeat what I said:

I disagree with some of their specific claims, but strongly agree with their main points: in particular that

1. current AI benchmarks are badly misleading,
2. AI adoption will be more gradual than one might expect, and
3. that effective AI safety is achievable by restricting the power given to AIs to directly affect the world, much the same as any other case of technological safety.

(Along similar lines, they do define what they mean by ‘normal technology’. Note that they also define, e.g., the printing press and the Internet as ‘normal technologies’, despite their transformative impacts. It may be a poor choice of name, but the concept itself is reasonable.)

[malloc]

(Along similar lines, they do define what they mean by ‘normal technology’. Note that they also define, e.g., the printing press and the Internet as ‘normal technologies’, despite their transformative impacts. It may be a poor choice of name, but the concept itself is reasonable.)

Nonetheless, I maintain that AI is fundamentally different from technologies like the printing press let alone the vast majority of less remarkable technologies. Printing presses only print what you specifically choose and have no ability to compose their own texts or reason about the text they handle. It makes no sense to talk about using a machine that makes all the consequential decisions about the process for reasons you cannot even decipher. You are not using the machine to implement your decisions in that case but rather supplicating it for solutions to questions you honestly don't understand all that well yourself.

so is life, yet we also can't replicate abiogenesis.

Apart from anything else, massive corporations haven't spent hundreds of billions on replicating abiogenesis. The whole field might look radically different if dozens of billionaires spent years throwing their fortunes at it.

so now we're talking businesses rather than techbros or internet sites? Goalboast-shifting noted. Probably the same ones who are manned entirely by AIs...you know, Warner Brothers, Universal, Marvel. Oh wait, those are staffed by human artists. Underpaid ones, probably, but still humans.

They are certainly trying to replace humans with AI. There was that lengthy writers' strike over attempts by Hollywood executives to replace them with AI-written scripts.

[bradrn]

(Along similar lines, they do define what they mean by ‘normal technology’. Note that they also define, e.g., the printing press and the Internet as ‘normal technologies’, despite their transformative impacts. It may be a poor choice of name, but the concept itself is reasonable.)

Nonetheless, I maintain that AI is fundamentally different from technologies like the printing press let alone the vast majority of less remarkable technologies. Printing presses only print what you specifically choose and have no ability to compose their own texts or reason about the text they handle. It makes no sense to talk about using a machine that makes all the consequential decisions about the process for reasons you cannot even decipher. You are not using the machine to implement your decisions in that case but rather supplicating it for solutions to questions you honestly don't understand all that well yourself.

I note that you haven’t responded to my actual claims (even after I conveniently reformatted them into a list for you).

Having said that:

I think the word you’re looking for is ‘agentic’: the ability to act as an agent under one’s one volition. I agree that there is a fundamental difference between agentic technologies and non-agentic ones, and that prior to AIs we haven’t really seen any technologies with a significant amount of agency. But the authors of that article are aware of that — indeed, they say that ‘for highly agentic systems, power and loss of control are tautological’ (which I agree with). This is why they, and I, focus on restricting the scope of AI agency — my point (c) — which I think is a very achievable goal.

[keenir]

(Along similar lines, they do define what they mean by ‘normal technology’. Note that they also define, e.g., the printing press and the Internet as ‘normal technologies’, despite their transformative impacts. It may be a poor choice of name, but the concept itself is reasonable.)

Nonetheless, I maintain that AI is fundamentally different from technologies like the printing press let alone the vast majority of less remarkable technologies.

feel free to maintain it all you like

Printing presses only print what you specifically choose and have no ability to compose their own texts or reason about the text they handle.

and neither do any AIs.

go ahead, tell me about an existing AI that can reason.

It makes no sense to talk about using a machine that makes all the consequential decisions about the process for reasons you cannot even decipher.

Malloc, I can say the same thing about execs who make decisions based on a Magic 8-Ball, or a random number generator. are we also in danger of mankind being replaced by Magic 8-Balls too?

so is life, yet we also can't replicate abiogenesis.

Apart from anything else, massive corporations haven't spent hundreds of billions on replicating abiogenesis.

last i heard, yes, there has been hundreds of billions spent on it.

The whole field might look radically different if dozens of billionaires spent years throwing their fortunes at it.

loads of rich people also threw their fortunes at phrenology. what shape is your skull?

so now we're talking businesses rather than techbros or internet sites? Goalboast-shifting noted. Probably the same ones who are manned entirely by AIs...you know, Warner Brothers, Universal, Marvel. Oh wait, those are staffed by human artists. Underpaid ones, probably, but still humans.

They are certainly trying to replace humans with AI. There was that lengthy writers' strike over attempts by Hollywood executives to replace them with AI-written scripts.

and now you've leapfrogged over to this.

[Raphael]

@zompist: Fair enough; all good points.

[bradrn]

go ahead, tell me about an existing AI that can reason.

Many current AIs can do so. Not as well as (some) humans, admittedly, but they can do it.

[keenir]

go ahead, tell me about an existing AI that can reason.

Many current AIs can do so. Not as well as (some) humans, admittedly, but they can do it.

my bad then; mea culpa, I wasn't aware of any, that could do it to any degree. thank you.

[zompist]

go ahead, tell me about an existing AI that can reason.

Many current AIs can do so. Not as well as (some) humans, admittedly, but they can do it.

I think an extraordinary claim like this needs extraordinary backing. Especially after you just said, correctly, "current AI benchmarks are badly misleading"!

One thing we know that LLMs do really really well is reproducing patterns of human writing. The thing is, nothing has ever been able to do such a thing except for humans, so we don't have much discrimination or testing ability here. We're like someone trained in detecting darkroom photo manipulation who's got their first glimpse of Photoshop.

And the training data includes shitloads of tests, logic problems, and other examples of human reasoning. Saying that because the LLM can produce something like its training data is much like malloc's claim that LLMs have taken over "literature".

The article you posted is interesting but not as skeptical as it could be... but it raises a good question: if an LLM can "pass the law exam", does that mean it can practice as a lawyer?

[Raphael]

I think an extraordinary claim like this needs extraordinary backing. Especially after you just said, correctly, "current AI benchmarks are badly misleading"!

One thing we know that LLMs do really really well is reproducing patterns of human writing. The thing is, nothing has ever been able to do such a thing except for humans, so we don't have much discrimination or testing ability here. We're like someone trained in detecting darkroom photo manipulation who's got their first glimpse of Photoshop.

And the training data includes shitloads of tests, logic problems, and other examples of human reasoning. Saying that because the LLM can produce something like its training data is much like malloc's claim that LLMs have taken over "literature".

The article you posted is interesting but not as skeptical as it could be... but it raises a good question: if an LLM can "pass the law exam", does that mean it can practice as a lawyer?

I can't speak for bradrn, but I'm not sure if he even meant LLMs. There might be other, older approaches in AI research that came closer to reasoning ability. Could you clarify, bradrn?

[zompist]

I can't speak for bradrn, but I'm not sure if he even meant LLMs. There might be other, older approaches in AI research that came closer to reasoning ability. Could you clarify, bradrn?

I can't speak for him either but I'm reasonably familiar with procedural AI, and I'd be wary of claims that it could "reason". Certainly no system could reason in general as a human can. But procedural AI is basically following simple instructions (fast enough that you can do complex things) written by a human programmer. You can program rules of inference, but does the program "know how to reason", or only know how to follow carefully-laid-out instructions? It's reminiscent of Socrates' "demonstration" that Meno's slave knew the truths of geometry.

I'd also be wary of claims that procedural AIs can't reason. One could still maintain, in 2000 or so, that programmers could eventually produce a robust enough procedural AI to match human intelligence. As ever, this was expected 5 to 50 years in the future.