AI/Automation Megathread

Morphic Tide

Well-known member
In a year or two it won't be an AI narrating your fanfic, the fanfic will be written, narrated, and illustrated entirely by AI.
And just like that, rather than creativity being decentralized and democratized, the ability to shape culture is concentrated in the hands of a few cloud server owners.
The way the technology works is too shallow for that. The reason why AI text generation works decently across scripts is because that's measured in kilobytes with a very well-defined finite set of output elements. A 50-page comic issue has enormous interdependency of dozens to hundreds of discrete elements with billions upon billions of logical output permutations, with only fractions of a percent of fractional percentiles making any logical sense and many times over again that few being any good.

That's why there's so much trouble with hands. The basic methodology is reconstructing 3d objects from 2d images by brute-force pattern recognition, and thus small details with enormous variability are obscenely complicated to do anything with. And that's within a single still image, nothing compared to keeping the entire character design consistent across hundreds, possibly thousands, of instances.

You two are fearmongering about AI taking over all creativity, while I'm sitting here and immediately being ejected by the mangled false-text in every damned one of the things. To be blunt, these neural network models are fundamentally incapable of the required context formation, because the way they work is purely blind fuzzy logic. Tell me when somebody has made an AI to generate TF2 Source Filmmaker shitposts without Source Filmmaker or TF2 assets directly in the generation pipeline, then I'll take you seriously.
 

Aaron Fox

Well-known member
One of the things that fucked us royally over the decades is the Heinleinian mentality that 'humans aren't ants.' Basically, specialization is bad and unnatural. He also fucked us over by entrenching the insane mentality that 'soft science isn't really science'.

Problem is, as complexity increases, so do the requirements of specialization. It's like how the economies of scale are baked into the universe; the same happens regarding specialization. Also, soft science is science, just more similar to quantum physics than regular physics.
 

Marduk

Well-known member
Moderator
Staff Member
People have yet to comprehend what this means for art and culture in general. What it means is that our creative faculties will atrophy, because people will be so used to just asking AI to amuse them in a very specific way, they won't have any interest in human-generated creative works any longer. It also means that whoever controls these AI content generation platforms will get to shape the culture by deciding what sorts of materials these AIs can actually produce. Ask an AI to produce a decidedly unwoke swashbuckling pulp novel with lots of sex and profanity? We're sorry, that violates our TOS. Would you like this ESG-approved pap instead?

And just like that, rather than creativity being decentralized and democratized, the ability to shape culture is concentrated in the hands of a few cloud server owners. Everyone's wondering, "How the hell are artists supposed to make a living?", and meanwhile, I'm wondering, "How the hell are humans supposed to be creative or communicate novel ideas with each other at all when we have to shout over a sea of generated content?"
Unless it gets much better and not too censored, very little. This exact problem has already existed for about 2 decades due to the sheer amount of creative content on the internet. People have to learn to find good stuff and filter out heaps of shovelware already. So far AI only makes the shovelware effectively free and infinite, with a temporary novelty bonus.
 

Morphic Tide

Well-known member
One of the things that fucked us royally over the decades is the Heinleinian mentality that 'humans aren't ants.' Basically, specialization is bad and unnatural.
Problem is, as complexity increases, so do the requirements of specialization. It's like how the economies of scale are baked into the universe; the same happens regarding specialization.
Factory workers are being expected to learn more of how their machines work to perform basic maintenance because the time delay for a specialist is too much, scientists are being expected to learn how to program so the code is made with a solid understanding of what it's for, AI researchers increasingly need to understand the hardware so they get a model that can run outside the trainer.

We're seeing complexity grow in a fashion unfriendly to delegation of sub-units, and thus it contra-indicates increased specialization of labor and machine learning. And no, the first example is not a farce, my brother's workplace has had contracted temps forced to rebuild file directories to get the CNC machines working again, because hiring a software engineer to do it is genuinely more of a bother.

More importantly, human instincts are shaped for highly multifunctional assistance from relatives, so the sort of specialization you mention is in fact unnatural to us and bad for our mental health. On top of being the easiest stuff to throw automation at by its very nature of extreme narrowness in function.

He also fucked us over by entrenching the insane mentality that 'soft science isn't really science'.
Also, soft science is science, just more similar to quantum physics than regular physics.
The "soft sciences" are worse than alchemy in rates of applicability and with lesser tangential learning. The vast majority of such fields are nearly totally compromised by political interests due to not having gotten nearly solid enough to give any pushback. Hermetic magic is more useful than most of the "soft sciences" at this point with how totally they've been gutted by the Long March.
 

Aaron Fox

Well-known member
Unless it gets much better and not too censored, very little. This exact problem has already existed for about 2 decades due to the sheer amount of creative content on the internet. People have to learn to find good stuff and filter out heaps of shovelware already. So far AI only makes the shovelware effectively free and infinite, with a temporary novelty bonus.
That isn't entirely the case. The problem with our current Art AIs is that the creators tended to gimp them in some way because they can make outputs that scare the shit out of their creators. If what I've heard is true, you can get pretty decent art from an unimpeded Art AI... meaning it will be pretty quick to have only the most talented of the talented to even compete in such an environment.

Face it, people, for pretty much everything, we're starting to outdate ourselves rather thoroughly at best. We're the horse when the viable car came out.
 

Morphic Tide

Well-known member
We're the horse when the viable car came out.
The key difference that comparison constantly misses is that the economy is not for horses, it's for humans. And thus if you put any significant fraction of the humans out of work, you've stripped much of the economy of income, and shortly thereafter the affected portions implode under the weight of their debts.

If some government, somewhere, tries to enact your nightmare scenario without a wide range of fundamentally different technology, it dies in extremely short order as things rapidly fall apart to the great challenge of central planning that is "making sense of the full picture".

Absolutely anything being missed dooms the entire system, and given that literal salt circles jam self-driving cars because "solid white line" means "don't cross" and most of the systems treat written text of a thing as the thing itself due to being designed to compare to the text string...

This is a critical problem of all hyper-impoverished dystopian scenarios arising directly from capitalist economies. It does not matter how "late stage" it is, how much it's mutated into Corporatism, how incredibly powerful the people at the top are, the basic methodology needs fools to have money to separate them from.

And it can only last a few months without that before megacorps start buckling.
 

LordsFire

Internet Wizard
Also, soft science is science, just more similar to quantum physics than regular physics.
The soft sciences absolutely are sciences, but from their inception they've been prone to being completely overtaken by hardened ideologues less interested in learning, and more interested in cloaking themselves in the mantle of 'science' to give the bullshit ideology they already hold more credibility.

Any science that is not linked to direct, immediately measurable consequences, or at most consequences just one step removed, is going to require anyone to become competent in the field themselves to know whether or not a given 'scientist' is spewing nonsense.

Put another way, human corruption makes 'the humanities' and all adjacent scientific fields corrupt.
 

hyperspacewizard

Well-known member
One of the things that fucked us royally over the decades is the Heinleinian mentality that 'humans aren't ants.' Basically, specialization is bad and unnatural. He also fucked us over by entrenching the insane mentality that 'soft science isn't really science'.

Problem is, as complexity increases, so do the requirements of specialization. It's like how the economies of scale are baked into the universe; the same happens regarding specialization.
probably not the thread for it but I can’t disagree more with this the science community and medical community has gone so far into their specialization they have a hard time even thinking outside it when confronted with a problem which leads to all kinds of bad outcomes. We need more generalists so the knowledge and skills of specialists are used in a synergistic manner.
 

Aaron Fox

Well-known member
probably not the thread for it but I can’t disagree more with this the science community and medical community has gone so far into their specialization they have a hard time even thinking outside it when confronted with a problem which leads to all kinds of bad outcomes. We need more generalists so the knowledge and skills of specialists are used in a synergistic manner.
That's literally outright impossible.

I'm serious.

We already need some 30 years (from birth to graduation) to get someone with the required skills to be decent at a specialty (things like surgery and studying quantum physics), and it will only get WORSE. Period. This line of thinking is outright dangerous, if not self-destructive. To get someone to generalize? Add another 20 on top of that. Minimum. This is just a historical trend that got supercharged after the 1970s. Not to mention that -also historically- the ability to specialize increases economic productivity.
 

Cherico

Well-known member
That's literally outright impossible.

I'm serious.

We already need some 30 years (from birth to graduation) to get someone with the required skills to be decent at a specialty (things like surgery and studying quantum physics), and it will only get WORSE. Period. This line of thinking is outright dangerous, if not self-destructive. To get someone to generalize? Add another 20 on top of that. Minimum. This is just a historical trend that got supercharged after the 1970s. Not to mention that -also historically- the ability to specialize increases economic productivity.

it gets worse, the baby boom generation started mass retirement last year, and their already starting to move money out of the stock market and into safer investments. So all of that free money that went into the tech industy? Its gone forever.
 

Aaron Fox

Well-known member
it gets worse, the baby boom generation started mass retirement last year, and their already starting to move money out of the stock market and into safer investments. So all of that free money that went into the tech industy? Its gone forever.
That isn't the case. The current automation revolution shares many similarities to the 2nd Industrial Revolution (i.e., what everyone calls the Industrial Revolution)... and we're already past the point where it can't stop, i.e., it's snowballing. That would have been true... if it was the 1970s or 1980s.
 

Cherico

Well-known member
That isn't the case. The current automation revolution shares many similarities to the 2nd Industrial Revolution (i.e., what everyone calls the Industrial Revolution)... and we're already past the point where it can't stop, i.e., it's snowballing. That would have been true... if it was the 1970s or 1980s.

Money still is needed to fund things, and there is no such thing as unlimited resources.

We have seen what tech is capable of when it has essentially free money, now were going to see what happens in a far more constrained environment.
 

Morphic Tide

Well-known member
We already need some 30 years (from birth to graduation) to get someone with the required skills to be decent at a specialty (things like surgery and studying quantum physics), and it will only get WORSE. Period.
Between the bullshit shoveled into K-12 in the interests of absolutely everyone getting a highschool education, the bullshit shoveled into colleges for political reasons, and the bullshit shoveled into the medical schools to navigate regulatory accretion, we could in fact have well-functioning 25 year old surgeons by focusing on teaching what's actually useful to them.

And with quantum physics, the theories are so interdependent that they're using fundamental particle research to generate Muon beams to create non-Baryonic matter to double-check that mass spectroscopy is right while simultaneously finally having an experiment where we can measure matter and antimatter together, which all requires a variety of engineers who need to be quantum physicists themselves to understand how to build the equipment in question, and the final researchers need to understand some of the engineering to understand the data they're getting.

Fact of the matter is that jobs are rarely so isolated that pure specialists are useful already, and this trend continues as the technological dependencies increase. Because the people making the technology need to understand what it's for to be sure it works, you really can't build something like the Large Hadron Collider with the people working on the electromagnets not understanding the edge-case relativistic effects at play.

That isn't the case. The current automation revolution shares many similarities to the 2nd Industrial Revolution (i.e., what everyone calls the Industrial Revolution)... and we're already past the point where it can't stop, i.e., it's snowballing. That would have been true... if it was the 1970s or 1980s.
No, it shares many similarities to the Dotcom Bubble, where a flashy new technology is being oversold utterly absurdly to the point of solid class-action fraud cases and hyped up for functions requiring a complete overhaul of the methodology to accomplish in any meaningful way, or just finally paying for shit we were perfectly capable of in the 90s.

Work from home is the best example. We could have done that in 2010, easily. Even with the ongoing economic trouble. But the managers insisted on in-person meetings, and so such continued until the lockdowns forced them to look for an alternative. Turns out we've been wasting billions of man-hours a year on completely fucking unnecessary commuting time.

Also, the thing about needing fools to part from their money applies. We simply cannot lay off half the workforce without collapsing the economy, and due to the combination of this being a gradual process and the enormous bulk of AI research being done by and for wholly unessential tasks like artwork or voiceacting, it's nearly guaranteed to kneecap itself. The capital to push it forward comes from the people it's putting out of work!
 

Bear Ribs

Well-known member
Tell me when somebody has made an AI to generate TF2 Source Filmmaker shitposts without Source Filmmaker or TF2 assets directly in the generation pipeline, then I'll take you seriously.
Can you point me at the human who's making TF2 Source Filmaker shitposts without either using assets or having at least looked at them?

The key difference that comparison constantly misses is that the economy is not for horses, it's for humans. And thus if you put any significant fraction of the humans out of work, you've stripped much of the economy of income, and shortly thereafter the affected portions implode under the weight of their debts.
Many powerful and influential people are talking about how they need to get rid of a significant fraction of the human race, which is pretty worrisome in this context.
 

Marduk

Well-known member
Moderator
Staff Member
meaning it will be pretty quick to have only the most talented of the talented to even compete in such an environment.
Exactly my point. Any art copied and distributed by internet at least was already going there even without AI, because top 0.01% of talent in any field multiplied by billions of people can produce more material already than anyone has hours in the day to enjoy.
The other 99.99%? Left with scraps and niches.
 

Morphic Tide

Well-known member
Can you point me at the human who's making TF2 Source Filmaker shitposts without either using assets or having at least looked at them?
The point is that the AI are utterly incapable of the continuity required, because as it turns out that bloats the resource demands of the model to the point that Google couldn't run it for anything useful.

Many powerful and influential people are talking about how they need to get rid of a significant fraction of the human race, which is pretty worrisome in this context.
Their failure to grasp that the system cannot simply be shrunk on-demand does not tell well of their mental faculties, no.
 

Bear Ribs

Well-known member
The point is that the AI are utterly incapable of the continuity required, because as it turns out that bloats the resource demands of the model to the point that Google couldn't run it for anything useful.
And the first car did 2.5mph tops and could only run for 20 minutes before having to "rest" and repressurize its boiler. You're objecting to problems with an extremely new, extremely experimental prototype in an industry that's known for increasing in capabilities by orders of magnitude per year. We're already seeing advancements in novel-writing AIs learning to include continuity and foreshadowing, the rest won't be far behind. Estimates are most of the internet's content will be AI-generated by 2025, I've already linked to those reports.

Their failure to grasp that the system cannot simply be shrunk on-demand does not tell well of their mental faculties, no.
It can be shrunk... if AI and robots can do the jobs of the... downsized humanity.
 

Morphic Tide

Well-known member
And the first car did 2.5mph tops and could only run for 20 minutes before having to "rest" and repressurize its boiler.
To extend your own comparison, the internal combustion engine has not been invented yet, and you're the one promising wonders that boiler power proved incapable of.

in an industry that's known for increasing in capabilities by orders of magnitude per year.
Not anymore, we're running out of atoms to shave and the data logistics just don't pan out with the road left in the current hardware.

We're already seeing advancements in novel-writing AIs learning to include continuity and foreshadowing, the rest won't be far behind.
Text is literally thousands of times simpler, if not millions, than a 50-page comic keeping its main character on model.

It can be shrunk... if AI and robots can do the jobs of the... downsized humanity.
The current system runs on money constatly changing hands to the point that China resorts to expiration dates on digital currency. The "useless eaters" are the hot air keeping this afloat
 

Bear Ribs

Well-known member
To extend your own comparison, the internal combustion engine has not been invented yet, and you're the one promising wonders that boiler power proved incapable of.

Not anymore, we're running out of atoms to shave and the data logistics just don't pan out with the road left in the current hardware.
We're literally on the brink of a Quantum Computing revolution.

Here's a major breakthrough less than a week old.

Text is literally thousands of times simpler, if not millions, than a 50-page comic keeping its main character on model.
An AI was able to generate 14 days of its own voiced video, based on Seinfeld, without a problem. It's pretty primitive in both graphics and delivery but clearly lacks the problems you claim it would have to have with continuity, we see coversations that have back and forth and scene changes to areas the discussion was just about when the characters talk about visiting a restaurant, as an example. In fact the only issue was that the AI decided to make a transphobic joke eventually and got hit by the banhammer for not being woke enough. Your objections don't reflect the observable reality at hand.



The current system runs on money constatly changing hands to the point that China resorts to expiration dates on digital currency. The "useless eaters" are the hot air keeping this afloat
We're entering an increasingly post-poor-value world. As wealth continues to concentrate in the hands of the 1%, the poor increasingly have no currency to change hands with and no value. We had lockdowns for years with people laid off, and shows like Velma being made even though nobody wants them except the woke super-rich.

As for China's digital currency, that appears to be much more about being able to datamine and control people's spending than them worrying that the economy will disintegrate if poor people save money. The poor already can't save a significant portion of their income, they need it to pay for rent and food. The idea that China's reliant on them spending everything, so hard they need to have expiration dates to keep the peasants from having a saving's account, is clearly nonsensical. The ones this will hit are the middle-class, the traditional target of the rich for every class-warfare tactic to ensure the next generation of competition never arises.
 

Users who are viewing this thread

Top