AIs gunning for our precious freelancers

[keenir]

Yeah, that sentiment has always perplexed me. It never made sense to me unless one assumes implicitly that human intelligence requires supernatural mechanisms that computers cannot use. If one accepts a purely materialist worldview as most leftists claim, then one must logically accept the possibility of artificial intelligence.

given the rarity of human-level intelligence in anything other than humans, and the absence of anything more intelligent than humans, then one must either accept that AI will not equal - much less surpass - human intelligence, or that AI will not occur...

Err, what? How so? Are you arguing that things that are rare can't exist? Are you arguing that what is currently rare must stay rare? Are you arguing that what currently doesn't exist can't exist in the future?

No to all of those. I was attempting to argue against Malloc's claim that, just because one uses a materialist worldview, that doesn't de facto result in mankind being crushed under the feet of SkyNet - because Malloc seemed to be arguing that, without a need for "a soul" or other spiritual element absent from machines, anything can develop intelligence - including machines...possibly that machines must develop overwhelming intelligence.

So I pointed out the rarity of intelligence on Earth,

[rotting bones]

the much more currently urgent question of whether any of the stuff that currently gets pushed on the public under the label "AI" is anywhere near being that AI. On this, you are wrong, and Zitron is right, because the current forms of AI are pathologically incapable of telling truth from falsehoods, and that incapability is built into their very structure, which makes them mostly useless for anything else than playing around.

It's the same with humans. If fact, humans are more inclined to believe lies than truths, so AI is marginally better in that regard.

As jcb pointed out a few posts ago here, if you ask them for book recommendations, they'll make up books that don't exist. As other people pointed out in the past, they once sent tourists asking for restaurant recommendations to a food bank for homeless people. Shortly ago, I saw a post by someone who had asked a new AI chatbot that's supposed to answer any questions people in California might have about fires and fire safety what the phone number of the fire department is, and couldn't get an answer.

You have to use the right tools or at least frame the prompt correctly. For example, if you use the Consensus GPT to search for papers, every paper will be correctly referenced. Even if you insist on misusing a vanilla chatbot for the evaluation because you hate AI, at least turn on the web search feature and ask ChatGPT to "find" the thing you are looking for. An LLM is not intended to be used as a memory bank.

I'm telling you, there is scientific evidence that AI can do things other methods cannot. It can solve protein folding. It can be used to interpret neural signals from damaged nerves to let disabled people control a robotic arm. Statistical inference can be combined with deductive inference to achieve levels of performance that neither system is capable of in isolation.

Yeah, AIs are clearly ready to replace human beings.

I was going to say that CEOs should hire humans to summarize their emails, but hearing leftists talk about AI, I see why AI is clearly the superior alternative.

Pretty sure the mainstream leftist commentator line today is that AI can't do any real task right, fails on everything and is also conveniently evil. Here in America, the first inclination of leftists is to cancel any project that uses AI in any capacity whatsoever. I speak from personal experience.

I don't see what's supposed to be wrong with that line.

All of the things I've been saying for years now. More counter-evidence against the claim that AI is always wrong:

For context, Google just used AI to find a bunch of new math proofs: https://www.youtube.com/watch?v=sGCmu7YKgPA

When I asked ChatGPT to summarize a Marx quote, the result was accurate, but I felt like it left out too many of the subtler points.

...

It's always like this. Most people go around saying batshit crazy things, while intellectuals like Ares Land keep up an endless stream of gaslighting: "I don't know this, I've never heard of this, ..."

[Ares Land]

It's always like this. Most people go around saying batshit crazy things, while intellectuals like Ares Land keep up an endless stream of gaslighting: "I don't know this, I've never heard of this, ..."

First, calm down. Second, what one or several people say on a bulletin board doesn't automatically represent the opinion of 'leftists.'

Third, well yeah, plenty of people are suspicious of AI. It's not the end of the world. Nobody is cancelling AI projects anyway. If AI turns out useless, the bubble will pop sooner or later. If AI turns out useful, well, the bubble may still pop but the technology will stay on.
AI companies are adequately funded; they're not exactly Galileo and they're not really in danger of being burned at the stake. They're perhaps not terribly popular, but by ignoring the worst abuses they did not exactly try to win hearts and minds.

FWIW they did improve the reliability and the amount of hallucinations is largely overstated. People on, say, Bluesky will periodically give examples of hallucinations but I can't reproduce them with ChatGPT.

The way people use LLMs is a bit weird. Often it seems to substitute for a search engine -- Google would have given the answer just as well. Some coders swear by ChatGPT and why not, but it doesn't seem to save much effort, as compared to actually writing the code. You can get it to write cover letters (which I guess serves recruiters right for requiring them in the first place )
As an IT guy I resent deeply the urge of non-technical executives to demand AI in projects where it's not the right tool for the job.

Other uses of machine learning are more interesting. I do like PlantNet. I've been incidentally involved in a machine learning project (long story short, digitizing manuscripts) and that was both fun and useful.

[Raphael]

the much more currently urgent question of whether any of the stuff that currently gets pushed on the public under the label "AI" is anywhere near being that AI. On this, you are wrong, and Zitron is right, because the current forms of AI are pathologically incapable of telling truth from falsehoods, and that incapability is built into their very structure, which makes them mostly useless for anything else than playing around.

It's the same with humans. If fact, humans are more inclined to believe lies than truths, so AI is marginally better in that regard.

Yeah, a lot of human beings believe a lot of lies. Nevertheless, if you're in the concrete situation where you need a specific piece of information, and you need it to be accurate, then you can't rely on AI, because it gets stuff wrong so often. This means that, in that situation, if you insist on using AI, afterwards you have to check independently whether what the AI told you is true. Which, in turn, means it might well be simpler and easier and less cumbersome to just do the research yourself without AI.

That might be part of why business leaders and their friends in politics love AI so much: they rarely ever face any bad consequences for making bad decisions. Therefore, they don't have to care how good or bad their decisions are; and therefore, they don't have to care how reliable or unreliable the information on which their decisions are based is.

You have to use the right tools or at least frame the prompt correctly. For example, if you use the Consensus GPT to search for papers, every paper will be correctly referenced. Even if you insist on misusing a vanilla chatbot for the evaluation because you hate AI, at least turn on the web search feature and ask ChatGPT to "find" the thing you are looking for.

So AI works fine as long as you jump through enough hoops in exactly the right way when using it. Problem is, once the effort required to jump through enough hoops in exactly the right way becomes too big, it's less effort to simply do the thing in question without AI.

An LLM is not intended to be used as a memory bank.

It isn't? I had the impression that that's one of the things as which its promoters promote it.

I'm telling you, there is scientific evidence that AI can do things other methods cannot. It can solve protein folding. It can be used to interpret neural signals from damaged nerves to let disabled people control a robotic arm. Statistical inference can be combined with deductive inference to achieve levels of performance that neither system is capable of in isolation.

Fair enough. Those are highly specialized and technical applications, though. There are a lot of specialized types of lab equipment that can do things nothing else can do. That doesn't mean there's a viable market for putting those specialized types of lab equipment into every business and every home in the world.

Pretty sure the mainstream leftist commentator line today is that AI can't do any real task right, fails on everything and is also conveniently evil. Here in America, the first inclination of leftists is to cancel any project that uses AI in any capacity whatsoever. I speak from personal experience.

I don't see what's supposed to be wrong with that line.

All of the things I've been saying for years now.

Most of them have been refuted by people here. Remember when you tried to explain away the case where an AI had turned a low-res photo of Barack Obama into a high-res image of a white man?

[rotting bones]

Second, what one or several people say on a bulletin board doesn't automatically represent the opinion of 'leftists.'

I have consumed the content of all the largest online leftist commentators, and what I say is largely true. I've waded through too much trash content to calm down.

Third, well yeah, plenty of people are suspicious of AI. It's not the end of the world. Nobody is cancelling AI projects anyway. If AI turns out useless, the bubble will pop sooner or later. If AI turns out useful, well, the bubble may still pop but the technology will stay on.
AI companies are adequately funded; they're not exactly Galileo and they're not really in danger of being burned at the stake. They're perhaps not terribly popular, but by ignoring the worst abuses they did not exactly try to win hearts and minds.

That's even worse. Not only are leftists canceling the wrong things, they're too incompetent to cancel the right things; namely, all for-profit use of LLMs. I don't know what the open source alternatives to ChatGPT are, if any.

The way people use LLMs is a bit weird. Often it seems to substitute for a search engine -- Google would have given the answer just as well. Some coders swear by ChatGPT and why not, but it doesn't seem to save much effort, as compared to actually writing the code. You can get it to write cover letters (which I guess serves recruiters right for requiring them in the first place )

I would say LLM search is one level deeper than a Google search. You can use it to search 5 or 6 queries, read the pages, find the info relevant to your natural language query, summarize the results and present them in a couple of tables and paragraphs, complete with sources. o3's web search is quite useful, especially if the results aren't that important. I haven't seen it hallucinate so far, but I'm sure it will one day. I've even had Google Maps hallucinate on me one time around 2016 or 2017, sending me on a 2 hour walk off into the countryside, where I was picked up by the police.

[Raphael]

because Malloc seemed to be arguing that, without a need for "a soul" or other spiritual element absent from machines, anything can develop intelligence - including machines...possibly that machines must develop overwhelming intelligence.

So I pointed out the rarity of intelligence on Earth,

Again, rarity doesn't imply impossibility.

[Ares Land]

I have consumed the content of all the largest online leftist commentators, and what I say is largely true. I've waded through too much trash content to calm down.

OK then. If you want to pick up a fight, fine -- but I'm not interested.

[zompist]

Third, well yeah, plenty of people are suspicious of AI. It's not the end of the world. Nobody is cancelling AI projects anyway. If AI turns out useless, the bubble will pop sooner or later. If AI turns out useful, well, the bubble may still pop but the technology will stay on.
AI companies are adequately funded; they're not exactly Galileo and they're not really in danger of being burned at the stake. They're perhaps not terribly popular, but by ignoring the worst abuses they did not exactly try to win hearts and minds.

That's even worse. Not only are leftists canceling the wrong things, they're too incompetent to cancel the right things; namely, all for-profit use of LLMs.

Are you preparing a heel turn where you turn right-wing? That's usually what happens when someone starts strawmanning the entire left.

In this timeline, "leftists" do not have the capability to "cancel" for-profit LLMs, becaue they are not in power. It's you who can't figure out the right people to oppose. Your constant assertions that you and only you know The Truth are just getting weird. Sure, you're an unsung genius, but stop thinking you're the only one.

I don't know what the open source alternatives to ChatGPT are, if any.

Then find out. Maybe ask ChatGPT since you think its an accurate search engine. I could ask a guy I know at Mefi who's always talking about open-source and non-corporate AI projects, but since you've read every leftist in the world you undoubtedly know him already.

It can solve protein folding.

Also: jeez, no one is criticizing protein-folding AIs, except maybe malloc. No one's banning your precious protein folders.

[rotting bones]

Nevertheless, if you're in the concrete situation where you need a specific piece of information, and you need it to be accurate, then you can't rely on AI, because it gets stuff wrong so often. This means that, in that situation, if you insist on using AI, afterwards you have to check independently whether what the AI told you is true. Which, in turn, means it might well be simpler and easier and less cumbersome to just do the research yourself without AI.

This is not true. If you study the P vs. NP problem, you will find that it's easier to verify an answer than to come up with it. An AI is often able to come up with a close enough answer that you can verify. This saves a lot of time.

That might be part of why business leaders and their friends in politics love AI so much: they rarely ever face any bad consequences for making bad decisions. Therefore, they don't have to care how good or bad their decisions are; and therefore, they don't have to care how reliable or unreliable the information on which their decisions are based is.

I'm shocked they're willing to share the credit for their "brilliance" with AI.

So AI works fine as long as you jump through enough hoops in exactly the right way when using it. Problem is, once the effort required to jump through enough hoops in exactly the right way becomes too big, it's less effort to simply do the thing in question without AI.

That's not true. It's not possible to find the same math proofs with AI as without it. It's not easier to train yourself to read Karl Marx's writing than using an LLM to summarize his blather.

It isn't? I had the impression that that's one of the things as which its promoters promote it.

No, what it can do is summarize and elaborate on large amounts of text in natural language.

Most of them have been refuted by people here. Remember when you tried to explain away the case where an AI had turned a low-res photo of Barack Obama into an high-res image of a white man?

None of them have been refuted. It was all the usual human meat flapping. AI always hallucinates. So do humans. So do all knowledge representation systems. The point is that it's not true that AI models always hallucinate Barrack Obama into a white man. Super resolution is trained on its performance compared to the original hi res images.

I know it is hard for us humans to accept factual information, so I'll show you.

I take this image: https://upload.wikimedia.org/wikipedia/ ... _Obama.jpg

I shrink it to get this: https://drive.google.com/file/d/1am2rHS ... sp=sharing

Put it through this website: https://imggen.ai/tools/upscale-image (warning: ad hell)

And get this result: https://drive.google.com/file/d/1_Q1XBR ... sp=sharing

See? What appeared as a refutation turned out to be the usual "leftist" cleverness that brought Trump to power twice. It's not real. Snap out of it.

[rotting bones]

Are you preparing a heel turn where you turn right-wing? That's usually what happens when someone starts strawmanning the entire left.

What, for free? The last time I checked, right-wingers were humans too. I admit this info might be out of date.

Seriously though, online leftist commentators only.

In this timeline, "leftists" do not have the capability to "cancel" for-profit LLMs, becaue they are not in power. It's you who can't figure out the right people to oppose. Your constant assertions that you and only you know The Truth are just getting weird. Sure, you're an unsung genius, but stop thinking you're the only one.

I'm the stupidest possible phenomenon.

Edit: They're not opposing the projects they should be opposing.

Then find out. Maybe ask ChatGPT since you think its an accurate search engine. I could ask a guy I know at Mefi who's always talking about open-source and non-corporate AI projects, but since you've read every leftist in the world you undoubtedly know him already.

I've run open source LLMs before. I mean I don't know the best, most usable alternatives. The answer is somewhere in here: https://huggingface.co/spaces?q=leaderboard&sort=likes

o3 says: https://chatgpt.com/share/682f128a-1b68 ... 8eed9471f3 Should I trust this? I wouldn't in my research.

Also: jeez, no one is criticizing protein-folding AIs, except maybe malloc. No one's banning your precious protein folders.

More's the pity. Why aren't commercial LLMs being replaced by open source, energy conscious alternatives?

[malloc]

More like defending the interests of humans against an interloper.

against an interloper? you refuse to defend the interests of humans against bad people, so why should we believe you would defend human interests against a nonhuman?

I am still trying to figure out the best way to resist AI apart from refusing to use it myself. Half the battle is simply convincing people of its power and ability to compete with humans. So many people still dismiss the whole concept of AI as smoke and mirrors or else hopelessly rife with hallucinations. One may presume that government regulations are necessary, although given the current regime that seems unlikely for the US, and that we ought to push AI into the realm of taboo. Grim as the situation may appear, humans do have the advantage of incumbency. We need only defend our position in the world from the intrusion of AI whereas AI must force us out of the way. Currently its handmaidens in the tech industry are attempting to bribe us into leaving the gates open by offering to eliminate the drudgery of thought and creativity. We must resist all such bribes and tell the techies in no uncertain terms: to hell with your thirty pieces of silicon.

although I really don't think far right billionaires should be the ones to design what comes after Homo sapiens.

don't worry, they aren't.

Clearly they are given all their talk of artificial general intelligence and the billions they are throwing at such projects. They have said many times that they want to bring about the Singularity in which AI becomes so powerful that exceeds our ability to comprehend it and radically transforms the entire world. Given the incredible success of AI over the past few years, they are evidently progressing toward this goal at a remarkable pace.

[Raphael]

although I really don't think far right billionaires should be the ones to design what comes after Homo sapiens.

don't worry, they aren't.

Clearly they are given all their talk of artificial general intelligence and the billions they are throwing at such projects.

They are trying to. Have you done everything you tried to do in your life? I sure haven't.

They have said many times that they want to bring about the Singularity in which AI becomes so powerful that exceeds our ability to comprehend it and radically transforms the entire world.

A lot of medical quacks have said that they want to cure cancer.

Given the incredible success of AI over the past few years,

Could I interest you in a valuable mountain chalet in the south of Florida?

[Torco]

i feel drawn to agree with the mountain chalet remark, but actually, i have to admit that generative models really have had extraordinary success lately... at generating AI slop. i'm afraid some version or other of dead internet theory will be true within the next 10 years or so, as the barriers to entry to generating AI slop keep getting lower and lower.

We must resist all such bribes and tell the techies in no uncertain terms: to hell with your thirty pieces of silicon.

I don't think the butlerian jihad will start before the internet as a whole is unusable for a number of years. even in that case, it's just so economical to keep using AI tools for various jobs that even after that it possibly won't come.

[keenir]

More like defending the interests of humans against an interloper.

against an interloper? you refuse to defend the interests of humans against bad people, so why should we believe you would defend human interests against a nonhuman?

I am still trying to figure out the best way to resist AI apart from refusing to use it myself. Half the battle is simply convincing people of its power and ability to compete with humans.

you may as well convince people of the horrible future we will suffer with antigravity.

. We need only defend our position in the world from the intrusion of AI whereas AI must force us out of the way.

I wonder if there was once, thousands of years ago, a Paleo-Malloc, who warned and raved about how cattle will force humans into slavery, and how to prevent that, we must never domesticate the cow.

Currently its handmaidens in the tech industry are attempting to bribe us into leaving the gates open by offering to eliminate the drudgery of thought and creativity. We must resist all such bribes and tell the techies in no uncertain terms: to hell with your thirty pieces of silicon.

bribes? thirty pieces of silicon? Malloc, AI is not poised to take over the world, and it never will be - no matter how much your handmaiden friends want to bribe you.

although I really don't think far right billionaires should be the ones to design what comes after Homo sapiens.

don't worry, they aren't.

Clearly they are given all their talk of artificial general intelligence and the billions they are throwing at such projects.

and people who do that, like Musk, are also throwing their fortunes at projects like getting humans on Mars -- which is actually more attainable...even with all the problems, like getting people to survive 3-5 years of radiation in a glorified tin can, and surviving the landing (which not even all probes can do)

if I have a hundred billion or more dollars, and I spend two or three billion on a given program...that isn't much dedication. it is, in fish terms, broadcast spawning, not something with great odds of success.

They have said many times that they want to bring about the Singularity

people have been announcing the immenant arrival of the Singularity since at least the 50s. danged slow Singularity, clearly.

in which AI becomes so powerful that exceeds our ability to comprehend it and radically transforms the entire world. Given the incredible success of AI over the past few years, they are evidently progressing toward this goal at a remarkable pace.

success?

someone learns how to teach AI to fold proteins...and you think AI is now ready to transform the world? by...folding the planet?

[zompist]

I know it is hard for us humans to accept factual information, so I'll show you.

I take this image: https://upload.wikimedia.org/wikipedia/ ... _Obama.jpg

I shrink it to get this: https://drive.google.com/file/d/1am2rHS ... sp=sharing

Put it through this website: https://imggen.ai/tools/upscale-image (warning: ad hell)

And get this result: https://drive.google.com/file/d/1_Q1XBR ... sp=sharing

This is really really duplicitous, or self-delusion.

The original picture is here. No one made it up; an AI really really turned a picture of Obama into a white man. Now you could point out that that was 5 years ago and image processing has improved, but you don't, you just cheat.

The original pic is 32x32 pixels. Your "shrunken" image is 242x302... literally 71 times the resolution. Even the face part of the picture is 100x100, or ten times the resolution. You haven't shown anything except how to fool yourself.

No one has ever said "AI models always hallucinate Barrack Obama into a white man"; you made that shit up. The original example was about bias due to incomplete training data. But it does also serve to show that your hallucination that AI can restore missing data as in Star Trek is also wrong. We went over this long ago: AIs produce statistically likely data. That's suitable for, say, upscaling images in a remastered video game. It is not reliable for, say, convicting someone based on a blurry photo.

(Also, everyone in AI probably read that article back in 2020, and made sure their training data had pictures of Obama. Not every problem in AI can be solved this way. Both critics and proponents of AI fail to understand the power of a training set. E.g. AI guys think that answering test questions is a good test of AI... because it's a good test for humans. But training sets include reams of test questions and their answers. Whatever corpus you've got, AI is good at creating stuff that sounds or looks like it.)

[rotting bones]

The original picture is here. No one made it up; an AI really really turned a picture of Obama into a white man. Now you could point out that that was 5 years ago and image processing has improved, but you don't, you just cheat.

I already did. I answered in the past about how engineers try to debias models. Raphael's point was about my post. I only said it's amazing that AI can interpolate details to form crisp images from blurry images. I shrank the image to a quarter size. That's enough for a comparison to Star Trek's Enhance, right?

I agree that AI has significant hallucinations. I reported some of them here when ChatGPT first came out. IIRC It was unable to compare two dates and figure out which item is older. This was before any of the massive media blow up. I think recent models would be better at the date test I posted here.

I have repeatedly stressed that the models are unreliable, should not be used for profit and businesses should hire humans. What do you want from me?

The original pic is 32x32 pixels. Your "shrunken" image is 242x302... literally 71 times the resolution. Even the face part of the picture is 100x100, or ten times the resolution. You haven't shown anything except how to fool yourself.

What results you get by extrapolating from microscopic images is an interesting experiment, but it has nothing to do with my original point about super resolution or AI upscaling.

Go ahead if you want to shrink the image even further. I would if I were on the computer. I'm sure the results will be amusing.

No one has ever said "AI models always hallucinate Barrack Obama into a white man"; you made that shit up. The original example was about bias due to incomplete training data. But it does also serve to show that your hallucination that AI can restore missing data as in Star Trek is also wrong. We went over this long ago: AIs produce statistically likely data. That's suitable for, say, upscaling images in a remastered video game. It is not reliable for, say, convicting someone based on a blurry photo.

(Also, everyone in AI probably read that article back in 2020, and made sure their training data had pictures of Obama. Not every problem in AI can be solved this way. Both critics and proponents of AI fail to understand the power of a training set. E.g. AI guys think that answering test questions is a good test of AI... because it's a good test for humans. But training sets include reams of test questions and their answers. Whatever corpus you've got, AI is good at creating stuff that sounds or looks like it.)

I don't know if you remember, but the person who said you shouldn't convict someone based on an upscaled photo was me, not you or anyone else in the conversation. This position is perfectly compatible with describing AI upscaling as Star Trek's Enhance.

"AI models always hallucinate Barrack Obama into a white man" is the conclusion you get if you oppose my original post with that article:

"AI can enhance images."

"Obama was turned into a white man. Lol"

How is the second sentence relevant to the first of it's not being used as an example to prove that AI can't enhance images?

There are detailed studies about the mathematical structure of models and how to debias them. Unfortunately, that's all I have time for right now.

[malloc]

I wonder if there was once, thousands of years ago, a Paleo-Malloc, who warned and raved about how cattle will force humans into slavery, and how to prevent that, we must never domesticate the cow.

Nobody was trying to breed superintelligent aurochs though. Quite the opposite, cows were bred as food and draft animals under the absolute dominion of humans. Artificial intelligence by contrast is specifically designed around maximizing agency and intelligence with the intent of replacing humans in the economy and ultimately civilization in general.

someone learns how to teach AI to fold proteins...and you think AI is now ready to transform the world? by...folding the planet?

Except that AI has mastered numerous other tasks besides protein folding, everything from writing and drawing to playing chess to proving mathematical theorems. Every day the abilities of AI expand and if this trajectory continues, it will match or even exceed humans at everything. Nobody has explained why AI would suddenly hit a wall in its intellectual development that would prevent this from happening.

[Raphael]

Except that AI has mastered numerous other tasks besides protein folding, everything from writing and drawing to playing chess to proving mathematical theorems.

You're still way too much in awe of tasks you associate with stereotypically very smart human beings.

Every day the abilities of AI expand and if this trajectory continues, it will match or even exceed humans at everything. Nobody has explained why AI would suddenly hit a wall in its intellectual development that would prevent this from happening.

Well, for one, AI software runs on computer hardware, and there are physical limits to how fast computer hardware can get. That's why "the Singularity" as originally imagined can't happen; it would require the processing power of microchips to keep on increasing for the foreseeable future, which can't work. At some point, you get to individual electronic elements the size of a few atoms, and then you can't make them smaller any more.

[keenir]

I wonder if there was once, thousands of years ago, a Paleo-Malloc, who warned and raved about how cattle will force humans into slavery, and how to prevent that, we must never domesticate the cow.

Nobody was trying to breed superintelligent aurochs though.

you understand the concept of 'analogy' worse than an AI does; i didn't think that was humanly possible.

wait...

Quite the opposite, cows were bred as food and draft animals under the absolute dominion of humans.

show me an AI that can survive outside of the captive enviroment it depends upon.

Artificial intelligence by contrast is specifically designed around maximizing agency and intelligence with the intent of replacing humans in the economy and ultimately civilization in general.

the people who claim that AI will do all that - yourself and the 'techbros' included - are not the ones actually designing the AIs.

someone learns how to teach AI to fold proteins...and you think AI is now ready to transform the world? by...folding the planet?

Except that AI has mastered numerous other tasks besides protein folding, everything from writing and drawing to playing chess to proving mathematical theorems. Every day the abilities of AI expand and if this trajectory continues, it will match or even exceed humans at everything. Nobody has explained why AI would suddenly hit a wall in its intellectual development that would prevent this from happening.

we have explained it to you, MANY MANY TIMES.

AI is specialized, even beyond things like cheetahs. the protein-folding AI can't win at JEOPARDY, for example, nor play chess.

[malloc]

show me an AI that can survive outside of the captive enviroment it depends upon.

Sufficiently powerful AI could control all the industry necessary to maintain itself, though. Imagine robots mining and refining the ore, more robots manufacturing silicon wafers, and so forth all controlled by artificial intelligence. Nothing in principle prevents an entire civilization of intelligent machines with nary a human in sight.

we have explained it to you, MANY MANY TIMES.

AI is specialized, even beyond things like cheetahs. the protein-folding AI can't win at JEOPARDY, for example, nor play chess.

Sure but considered in the aggregate, artificial intelligence has mastered so much of human activity and continues to advance everyday. Even if no individual AI can do everything humans can, there are simply so many AI models out there that humans now face competition from at least one of them in an increasing number of fields. Presumably the next step is some kind of master AI that specializes in choosing the appropriate model for a given task. Imagine something like an LLM but instead of composing text in response to prompts, it chooses a particular AI and drafts a prompt based on stimuli.

Well, for one, AI software runs on computer hardware, and there are physical limits to how fast computer hardware can get. That's why "the Singularity" as originally imagined can't happen; it would require the processing power of microchips to keep on increasing for the foreseeable future, which can't work. At some point, you get to individual electronic elements the size of a few atoms, and then you can't make them smaller any more.

The tech industry already solved or at least mitigated that problem by using data centers and server farms full of hardware instead of trying to pack all the processing power necessary for AI into one small machine. They have massive warehouses with millions of chips all linked together to process far more data than any individual chip could manage. If necessary they could simply stuff more and more chips with maxed out processing power into the building. Imagine something like the Burj Khalifa but filled with billions and billions of GPUs and so forth.

Meanwhile we humans already reached the limits of our processing power hundreds of thousands years ago when our brains reached more or less their present size. Barring some truly radical technological advances, we don't have the option of stuffing more neurons into our skulls to improve our cognitive abilities. Artificial intelligence is racing past us while we remain stuck in quicksand.