Part Two: Tech Bros Have Built A Cult Around AI | BEHIND THE BASTARDS 

Roko's Basilisk has always been a hilarious demonstration of the ridiculousness of these subcultures.

If I had a nickel for every time a Harry Potter fanfiction author started a cult, I'm pretty sure I'd have at least 50 cents by now.

The amazing thing about Roko's Basilisk is that it's the only philosophical proposition I'm aware of where "I think you're an idiot" functions not only as a dismissal of the idea, but also as an effective counter-argument.

"Cultural Necophilia" would be an awesome name for a death metal band.

That guy accusing Carlin's daughter of greed while defending AI grifters has so little self-awareness that he might not be aware of his own existence.

I hate how so many groups call themselves names like "rationals" or "objectivists" or claim to use "facts and logic" or say that "facts don't care about your feelings" and then go on to use the language of rationality to justify their opinions and prejudices. It means whenever you try to /actually/ be rational and skeptical and evidence based etc. you end up aesthetically similar to the Ben Shapiros of the world, which ironically poisons the well of whatever point your trying to make.

"When art is released, it belongs to the world." (~37:30) Honestly, the British Museum had similar thoughts about 300 years ago.

Wait, Roko was just some rando on the Internet the whole time? I thought he was some obscure philosopher or something, that makes the whole thing SO much dumber than it already was.

Wait... I Have No Mouth But I Must Scream is a Harlan Ellison story? I thought it was Hello Kitty's autobiography.

Rocko's Modern Basilisk

As a DM and a person who thinks a lot about RPGs, I kind of want to include a plot in a game of D&D where the party are investigating a Conceptual Necromancer. They don't reanimate people's bodies, they study their writings and performances to conjure a magical facsimile of them. It would be just as horrifying as it sounds - basically creating AI Carlin but as a real being - and no one likes it. Just a rambling abomination with too many teeth and too many fingers, body parts bleeding into each other, parroting the words of dead men back at their loved ones. Peasant: "I would genuinely rather that guy just puppeted my father's corpse, like a regular necromancer. It would have been less of a defilement of his memory than this. Please, for the love of the gods, kill this man and banish his creations."

7:12 it's even worse than that. Elon and Grimes first interacted when they found out they had independently come up with the same PUN on Roko's basilisk: the rococo basilisk! 🙈🙈🙈🙈🙈

Ok, pissing off the AI Orthodoxy sounds like a great way to get your kneecaps confiscated by the Algorithm Crusaders, but how sick is "The First Cyber Heretic" as an Epitaph?

I think those Carlin AI guys only fessed up to lying about the AI part because George Carlin's family was threatening to sue

Roko's Basilisk is a term I've heard thrown around so much but never really looked into the meaning behind it because most explanations seem to focus on the "basilisk" part of it moreso than the original theory which I am dying at, oh my god these guys were kept up at night because some random dude was like "I have no mouth and I must scream is actually a documentary"

The problem with Pascal's Wager is that there's literally no reason whatsoever to think that an omnipotent, omniscient god would accept such a horrible deal. I mean.. what do they even get out of it? Literally nothing.. so they'd just say no.

nightshade is a thing in cyberpunk too! if someone has it on their character, scanning that person will blur them and hide any info that can normally be taken from people without it.

Will AI be deciding who gets to hunt children on a far off South Asian island?

Blaise Pascal is guy who came up with the wager. He did a LOT of mathematics and physics back in the day, then he went full Christian.

Thank you Robert for independently making a recent realization I had that Roko's Basilisk is just new Pascal's Wager. This is my second nickel and it tickles me slightly on my right shoulder. just above the scapula before the clavicle.

I have to say, I have studied AI before, and I can corroborate that the current technology cannot create a true AGI (artificial general intelligence, the sci-fi version of AI, basically), by its very technical characteristics. For starters, it is not general or generalizeable as is, so... yeah. At most, one could argue that you couldu se it as one of the basis for some hypothetical future method, but a) that would inherently mean a new method, whether or not it is partially based on an existing one doesn't change that; and b) there isn't even a well formulated theoretical method that we just couldn't build yet - it is pure wishful thinking, there is nothing concrete or even reliable. We don't have evidence that the current methods cannot be used as stepping stone for AGI, but we have any evidence that it could either. By default, that means we discard the hypothesis until decent positive evidence appears. In other words, it is on those people saying "we probably/totally are going to get AGI from these technlogies, babyyy" to demonstrate how their technology could possibly lead to AGI, or to at least possibly be used as a basis for true AGI. Until they do, it is just hogwash blind faith hidden behind technobabble and general ignorance on the topic.

Oh my god, finally someone else gets it, I took the deep dive into the LessWrong community before any of this AI stuff happened, and from the beginning I KNEW that they were behind it all. They are paranoid and delusional.

The podcast is 100% human written. In the reaction to the lawsuit from Carlin's Family, they also admitted that it always was 100% human written, with the "AI" premise being a comedic construct and about as factual as Craig Ferguson having had a sentient robot co-host.

Anyone complaining about Nightshade disrupting their model is stealing food to add to their food hoard and then complaining that they got poisoned. I also love that the same people claiming image generators "are just doing what human artists have always done" will say that they, the prompter, are "just an artist using a more advanced tool". You can't have it both ways.

Ai is great for people who don't want to be creative or think for themselves, but who do want the artifacts of thought to use in presenting themselves as creative thinkers.

There was some channel telling Warhammer 40k lore like it was David Attenborough. Did not go over well with the community.

Oh wow, thanks for mentioning Glaze and Nightshade!

If all art is open and free for public use, then so should their algorithms that they feed it into. If the artists can't make money of their art, then you shouldn't be able to make money off your machine learning algorithms either. Their entire position reeks of selfish incredulity.

Dunno if Robert scopes the comments on YT. But I just finished the Tommy Robinson video, and him despairing about the absolute state of war journalism. I'd recommend getting in touch with Dylan Burns for a relevant episode of BtB or It could happen here. I asked him once before and he said he doesn't want to be "that guy" pestering Robert for a collab. I told him he should stand outside your house with a boombox. He's currently getting hatefugged by the YT algorithm for his coverage, and so could use the spotlight. I consider him a voice of reasoned, passionate and ethical journalism in a world of shills and grifters.

it really does feel like a crusade against artists, hit the nail on the head. Great episode!!

If anyone thinks that a computer, a thing, unthinking unfeeling, unexperienceing can generate art, or intellect Then that person knows neither.

I haven't seen any of the context in the comments, so I'll provide it. Roko's Basilisk was created as a critique/edge case of a certain approach to decision theory. It was an "if this concerns you, you're an idiot" and not a serious argument. E. Y. probably got a bit offended by it since some of his core assumptions are called out by the basilisk. Furthermore, it illustrates the difficulty of modeling the collaboration between intelligent agents (people, companies, countries) in scenarios where one deters another by an act of pure revenge later (ie.: counter-value 2nd strike). The argument differs from PW in one important way: the suffering is self-inflicted, god doesn't exist until you build it. This is to illustrate that certain approaches to game/decision theory are broken enough to tell you to go ahead and do it.

Garrison’s opinion about Marietta is extremely valid and I second it.

For nightshade, like are ther any regulations about the bots imput? no, then its fair game, and it wont harm computer. If artists arent asked to share, they are goddamn free to do what the hell to alter their image, and thats literally it, altering, their own art. Which they are freely to do. If there were any agreement wiith artists there might be , but as long artists arent in any case anything in that agreement, their art, it literally is part off their damn art and people are free to alter it, and no computer harmed. Just an image alter , which is fair, in art. Also i am inherently suspicious who first goes after the artists. So yes, good reason to be suspicious.

The guy saying the program was like hacking seems to be getting confused between hacking/viruses and encryption.

Oh god every time I hear Les Wrong’s name I groan inside. That fan fic was a trip

To be clear, the reason why it's called a Basilisk is because, as the theory goes, if you aren't aware of the theory, then the AI would not punish you for your ignorance; it would only be actively choosing to not fund the AI that would incite its wrath. Therefore, you are only vulnerable to the basilisk if you read about it, making it dangerous information to read in the first place (the term 'basilisk' also taken from older, better sci-fi, the writings of David Langford). This is the context as to Yudkowsky banning discussion of it; even if he claims otherwise, on some level, at some point, he did believe that this was dangerous information that needs to be suppressed because it endangered people to even contemplate. Which is both much crazier and much more ridiculous than it just being a panic-inducing idea on its own mereits.

Thanks, this helped put my mind at ease, it's a much-needed conversation. The god damn George Carlin thing... I had forgotten about that nauseating bit of techno necromancy.

I need that plain tasteless pizza yesterday. "What if I told you, that the same finger... you use to order pizza... you could pick your nose with... our A.I. experts will show you HOW!"

I've started listening to the audiobook of that HP fanfic out of morbid curiosity. It's the most irritating dork shit, couldn't get through it. The only thing better in it is that the step parents aren't abusive for no reason.

About the copyright thing: As a leftist with a strong opposition to copyright law in general... I do think there are a lot of concerns around closing off lines of derivative art, and I worry about that in connection with AI. But at the same time, I may think we shouldn't have copyright laws... But until we DON'T, then we can't just argue that we can give stuff over to AI without any limitations, because right now, those laws DO exist. We can consider doing something like that (and I do mean consider, because there's plenty of reasons not to) only once the laws are gone for everyone, not just artists and workers.

We've seen that the corpos seem to have no clue what the limitations of generative ML are, and are incredibly lazy about actually checking the output, so I figure it's only a matter of time until a massive error in something sets off a bunch of big lawsuits, and that will be the end of that. Any guesses on the timeframe?

"Whoever Pascal was" Robert pains me sometimes. To me, that statement is on the level of the Arminian/Armenian debacle.

I'd heard Roko's Basilisk before, but yeah, that's where it came from. I had to check out wikipedia to verify it. And yes, it's a Pascal's wager variation. Oh, and thank you for the reference to Nightshade. I'm going to start poisoning images to hopefully hilarious results down the road.

The idea that an AI can't operate from its own whims is one I'm interested in. Trouble is, the questions you'd ask of it to test its original thought are going to have answers pillaged or planted externally. In any case, has anyone asked a chatbot to provide an example of an interesting question to put to an AI chatbot? Extra question - why are AIs becoming bad at spellcheck, when the permutations are complex but the words are finite and the effects of grammar structures are very mappable in the present day?

The wildest thing learned is that HPMOR was actually serious about rationality in any way. What in the actual fuck.

I'm pretty sure the Rationalist Harry Potter fanfiction mentioned is Harry Potter and the Methods of Rationality, which is as described, but I did enjoy reading, but to base a belief structure? That's wild.

Love the Harlan Ellison shout-out! When I was reading him in high school (1970s) and recommending him to people I would either get converts or horrified responses. Almost as fun as when I was teaching high school chemistry in the 2000s and giving aspiring song writers Tom Waits. I created a hardcore fan club of 16 year-olds that will never settle for pop music.

Saying "bad use case" is pretty good evidence that he doesn't know what a use case is.

I just feel like we don’t even understand our own brain, why do we need to add this to the equation. It’s like we wanna go colonize mars cause earth sucks, but we don’t care about Earth. I may be stupid though.

So... no big deal, but I'd like to push back a bit against slamming the 'Brights'. I never identified as one, but I did have a bit of a rep in the atheist community, and I can tell you, that name wasn't about bigoted self-aggrandizing. Back in those days, there was a lot of debate among atheists on the question of what to call ourselves. Because the fact was that *theists* had had the main influence over the dictionary definition of 'atheist' for centuries, and it was pretty tiring to have to keep pushing back against that, telling people no, we're not baby eaters or amoral or any of that. *My* view, the reason why I never called myself a 'Bright', was that re-naming was a waste of time since *whatever* we called ourselves, believers were going to use it was pejorative. So why bother? But I knew those people, and that's all it was. What *is* true, is that some atheists are assholes (some people in *any* group are assholes), and they've incorporated their atheism into their crap. Sigh. On a related note: I *did* used to call myself a Transhumanist. Despite the fact that Transhumanism had already become a haven for some pretty annoying Libertarians, I was totally on board with the goal of "let's augment ourselves technologically". I don't know when Transhumanism mutated from that into "let's steal everything from ourselves and give it to machines so they can murder us", but apparently it has, and I am not on board with that. So, I no longer identify with that term - assholes have ruined it.

Do Netanyahu next

I think you kind of glossed over a specific point... they also think that if they don't do it, other competing economies will happily pirate and do w/e they can to harvest all of the material necessary to make AI.. meaning that the US would be left behind. Though I think it is their strongest point, I don't think it is strong enough to just wipe out all of copywrite law just for them. The US military, whatever passes for homeland security ect.. probably already know this and are developing or have developed their own versions of it.

The only viable comedy I can think of AI generating based on that brady thing is dracula flow

The actual historical Crusades were mostly failures, so it's likely that the heretics will survive here, too. 38:30 I think the most telling part of the whole AI art thing is that the first people to sue over it were actually Free Software developers. The people who think software shouldn't be copyrightable sued Microsoft and OpenAI for training on freely-available and modifiable code that was licensed *to ensure that it remained freely modifiable*. If the arguments of the AI people were correct, you could strip code of this "must remain Free" clause just by running it through a training algorithm and then asking the model to spit it back out. Most of the artists victimized by AI are already acting outside the law. The AI cultists might still be able to claim fair use, but copyright already unambiguously criminalizes fanart, fanfiction, and all the other stuff that actual artists write and draw all the damned time. Copyright does not protect the creative class from publishers, it transfers wealth from the creative poor to the creative rich. AI also transfers wealth: from the creative poor to the technological rich. Who are increasingly part of the same company as, or at least able to talk to and socialize with, the creatively rich. Strict copyright enforcement is not going to stop AI once people agree on a licensing framework for it. 44:08 Nightshade can't be illegal because it's a form of DRM, and every country in the WTO has agreed that malware that protects the interests of a copyright owner is legally protected. If Nightshade is illegal then so is the DRM that keeps you from installing pirated iPhone apps.

I feel like it's worth pointing out that yudkowskey is no more an expert than you or I. He's just a nerd. Like, I don't actually dislike nerds, but he's not a genuine expert who knows anything about the production of AGI or anything; most experts with no vested interest consider AGI, if not impossible, so distant as to not be worth considering. That's not to say we can't commit mass fraud, or similar, but that's not AGI.

It would be hillarous and horrifying and probably extremely strange what a robert fed AI would say, or come up with sponsorship transitions. Would it useful. probably not. But for lols. Seriously ther should be a still mostly human made guarantee on anything comertial. Other than sph-e soph-e is good Also can ai please used for not creative stuff please, not for stuff that humans are just objective better, especially creative stuff or that needs ahuman hand. Because ther are pretty useful even ethical uses to do that with, as aid for research even as aid, not replace.

Just had this thought in light of the mention of the AI cult schism, but online forums/subcultures are basically the monasteries/temples of our day, and us users are basically cenobitic monks. We're all isolated at home, super pent up, and prone to get into elaborate arguments about dogma that get more and more detached from reality. The only difference is that the monks lived in the same place and had to do physical labor. Honestly, it explains why a lot of these monks' exegeses feel super deranged and like some proto-incel shit.

It is absolutely untrue that AI prompting is like writing the setup and letting AI write the punchline. For something like this it's certain that they let the AI writes a dozen to a hundred punchlines and then picked one they liked.

The Basilisk is just Pascals Wager but also guessing the "God's" 'commandments beacuase it doesn't exist yet. Atleast Christianity already came with an instruction document

It would be the Brady Bunch, surely

28:40 The Brady Bunch.

Harrison Ford and Nick Cage with new youthful shots = cgi, as were dead ghost busters. Are we confusing improved human tools with intelligence ? controlled cgi + image search engines + smashups with NLE+AfterEffects/Photoshop/Nuke merge tools + human curation = digital fakes.

Hail the Basilisk

When I hear something is out of lesswrong I assume it is dumb as fuck until shown otherwise.

Harry Potter and the principles of rationality?

"Harry Potter and the methods of Rationality" is a masterpiece. You come for the parody, but you stay for the actual dilemmas that Harry Potter raised on game theory actually has to deal with because of his choices. I genuinely went "Oh wow, I forgot about that crucial detail too!" when reading it.

It's absurd and depressing that we live within a system where a new tool being developed to potentially offload or replace some of the work that humans do is cause for consternation rather than celebration.

It's disingenuous to ignore open source AI in the discussion of interest groups in this conversation. It adds a wildly different option on the table for the future. Status quo: publishing companies own all cultural artifacts, Altman and compatriots vision of all cultural artifacts being made, curated and distributed from there control, and opensource models giving everyone in the world with the limits of hardware easier ability to access idea and information made in the world.

A I is here to help us. Scrotums are here to hurt us.

I think you guys are largely wrong, and also that it's very unlikely to matter.

I like ai generative art because it's great for non-commercial projects like modding a game where the quality of the art doesn't matter that much so long as it's better than nothing (/better than what you could have done yourself with minimal effort/time). Has nothing to do with hating artists or whatever.

For every AI company someone should use their AI to recreate the CEO of their company and use it to make a silly pitch for the AI as a way to mock them.

Training a model on publicly available works is not ‘stealing’. If you want to claim this, you are implying that viewing a piece of art is theft. You are implying that remembering that piece of art is a copyright violation. Both of those notions are absurd. Now, if you use a model to reproduce a creative work, then you might have violated copyright. But to be clear, the violation happened as a result of the output, not the input. We have an entire legal framework designed to discourage such violations and we should enforce them. I’m sick of the simplistic, reactionary mindset (uncritically parroted by everyone in this discussion) that says that models are stealing from artists by training on works they posted online. This is not a meaninglessly esoteric distinction that I’m making. In your zeal to stick it to those big bad capital owners, you’re going to assemble a culture that actively discourages creativity and will result if less art in the world. According to the worldview you are assuming, every single artist that has been moved by, or influenced by a previous creative work, has stolen from the creators of them. That’s absurd and counterproductive.

As with part 1 of this series, this discussion is a mess. You are conflating a number of different problematic things that have little to do with each other. For example, you state that AGI isn’t a thing yet, and that we don’t know if it will ever be a thing. You acknowledge that no one is claiming that it exists. You then proceed to provide demonstrations of AI generated crap that should be ridiculed for how uncreative it is. Ok, so what? What does one have to do with the other. What point are you making with this, and what does any of that have to do with being cultish? If awful people are probing to find the legal and cultural limits on fraud, and copyright violation, and identity theft, then we should punish them for committing fraud, copyright violations and identity theft. What does that have to do with ‘AI’? That stuff is already illegal, and if not, let’s enact better laws. Again, why is that an indictment of AI technology? Also, if the technology is so bad at all this stuff, then why should I be concerned about the implications? On the other hand, if you acknowledge that the technology might actually be good at these things, why are you spending time ridiculing the awful crap someone put out now? What purpose does this discussion have, other than jsut feel good nonsense? AI technology and the economic forces promoting it’s development are important topics, please approach them with the seriousness that their cultural implications demand.

yiekes! yiekes! that's just embarressing from Behind the Bastards