It’s hard to get into the article’s mood when I know that Lexis not only still exists but is now part of the Elsevier family; this is far from the worst thing that attorneys choose to do to themselves and others. Lawyers have been caught using OpenAI products in court filings and court appearances, and they have been punished accordingly; the legal profession does not seem prepared to let a “few hallucinated citations go overlooked,” to quote the article’s talking head.
Yeah, that’s the most surprising part of the situation: not only are the SCP-8xxx series finding an appropriate meta by discussing the need to clean up SCP articles under ever-increasing pressure, but all of the precautions revolving around SCP-055 and SCP-914 turned out to be fully justified given what the techbros are trying to summon. It is no coincidence that the linked thread is by the guy who wrote SCP-3125, whose moral is roughly to not use blueprints from five-dimensional machine elves to create memetic hate machines.
Thanks for linking that. His point about teenagers and fiction is interesting to me because I started writing horror on the Internet in the pre-SCP era when I was maybe 13 or 14 but I didn’t recognize the distinction between fiction and non-fiction until I was about 28. I think that it’s easier for teenagers to latch onto the patterns of jargon than it is for them to imagine the jargon as describing a fictional world that has non-fictional amounts of descriptive detail.
The orange site has a thread. Best sneer so far is this post:
So you know when you’re playing rocket ship in the living room but then your mom calls out “dinner time” and the rocket ship becomes an Amazon cardboard box again? Well this guy is an adult, and he’s playing rocket ship with chatGPT. The only difference is he doesn’t know it and there’s no mommy calling him for dinner time to help him snap out of it.
I’ve done some of the numbers here, but don’t stand by them enough to share. I do estimate that products like Cursor or Claude are being sold at roughly an 80-90% discount compared to what’s sustainable, which is roughly in line with what Zitron has been saying, but it’s not precise enough for serious predictions.
Your last paragraph makes me think. We often idealize blockchains with VMs, e.g. Ethereum, as a global distributed computer, if the computer were an old Raspberry Pi. But it is Byzantine distributed; the (IMO excessive) cost goes towards establishing a useful property. If I pick another old computer with a useful property, like a radiation-hardened chipset comparable to a Gamecube or G3 Mac, then we have a spectrum of computers to think about. One end of the spectrum is fast, one end is cheap, one end is Byzantine, one end is rad-hardened, etc. Even GPUs are part of this; they’re not that fast, but can act in parallel over very wide data. In remarkably stark contrast, the cost of Transformers on GPUs doesn’t actually go towards any useful property! Anything Transformers can do, a cheaper more specialized algorithm could have also done.
That’s first-order ethics. Some of us have second-order ethics. The philosophical introduction to this is Smilansky’s designer ethics. The wording is fairly odious, but the concept is simple: e.g. Heidegger was a Nazi, and that means that his opinions are suspect even if competently phrased and argued. A common example of this is discounting scientific claims put forth by creationists, intelligent-design proponents, and other apologists; they are arguing with a bias and it is fair to examine that bias.
You now have to argue that oxidative stress isn’t suffering. Biology does not allow for humans to divide the world into the regions where suffering can be experienced and regions where it is absent. (The other branch contradicts the lived experience of anybody who has actually raised a sourdough starter; it is a living thing which requires food, water, and other care to remain homeostatic, and which changes in flavor due to environmental stress.)
Worse, your framing fails to meet one of the oldest objections to Singer’s position, one which I still consider a knockout: you aren’t going to convince the cats to stop eating intelligent mammals, and evidence suggests that cats suffer when force-fed a vegan diet.
When you come to Debate Club, make sure that your arguments are actually well-lubed and won’t squeak when you swing them. You’ve tried to clumsily replay Singer’s arguments without understanding their issues and how rhetoric has evolved since then. I would suggest watching some old George Carlin reruns; the man was a powerhouse of rhetoric.
Rick Rubin hasn’t literally been caught with a dead woman like Phil Spector, but he’s well-understood to be a talentless creep who radicalizes men with right-wing beliefs and harasses women. Nobody should be surprised that he’s thrown in with grifters yet again, given his career.
Let’s do veganism now. I’m allowed to do this because I still remember what lentil burgers taste like from when I dated a vegan at university. So, as with most vegans, Singer is blocked by the classical counting paradoxes from declaring that a certain number of eukaryotic cells makes something morally inedible, and the standard list of counterexamples works just fine for him. Also, I hear he eats shellfish, and geoducks are bigger than e.g. chicks or kittens (or whatever else we might not want to eat.) I don’t know how he’d convince me that a SCOBY is fundamentally not deserving of the same moral insight either; I think we just do it by convention to avoid the cosmic horror of thinking how many yeast cells must die to make a loaf of bread, and most practicing vegans aren’t even willing to pray for all the bugs that they accidentally squish.
I agree with everything else he puts forward, but it boils down to buying organic-farmed food and discouraging factory farming. Singer is heavy on sentiment but painfully light on biology.
Singer’s original EA argument, concerning the Bengal famine, has two massive holes in the argument, one of which survives to his simplified setup. I’m going to explain because it’s funny; I’m not sure if you’ve been banned yet.
First, in the simplified setup, Singer says: there is a child drowning in the river! You must jump into the river, ruining your clothes, or else the child will drown. Further, there’s no time for debate; if you waste time talking, then you forfeit the child. My response is to grab Singer by the belt buckle and collar and throw him into the river, and then strip down and save the child, ignoring whatever happens to Singer. My reasoning is that I don’t like epistemic muggers and I will make choices that punish them in order to dissuade them from approaching me, but I’ll still save the child afterwards. In terms of real life, it was a good call to prosecute SBF regardless of any good he may have done.
Second, in the Bangladesh setup, Singer says: everybody must donate to one specific charity because the charity can always turn more donations into more delivered food. Accepting the second part, there’s a self-reference issue in the second part: if one is an employee of the charity, do they also have to donate? If we do the case analysis and discard the paradoxical cases, we are left with the repugnant conclusion: everybody ought to not just donate their money to the charity, but also all of their labor, at the cheapest prices possible while not starving themselves. Maybe I’m too much of a communist, but I’d rather just put rich peoples’ heads on pikes instead and issue a food guarantee.
It’s worth remembering that the actual famine was mostly a combination of failures of local government and also the USA withholding food due to Bangladesh trading with Cuba; maybe Singer’s hand-wringing over the donation strategies of wealthy white moderates is misplaced.
Humans are very picky when it comes to empathy. If LLMs were made out of cultured human neurons, grown in a laboratory, then there would be outrage over the way in which we have perverted nature; compare with the controversy over e.g. HeLa lines. If chatbots were made out of synthetic human organs assembled into a body, then not only would there be body-horror films about it, along the lines of eXistenZ or Blade Runner, but there would be a massive underground terrorist movement which bombs organ-assembly centers, by analogy with existing violence against abortion providers, as shown in RUR.
Remember, always close-read discussions about robotics by replacing the word “robot” with “slave”. When done to this particular hashtag, the result is a sentiment that we no longer accept in polite society:
I’m not gonna lie, if slaves ever start protesting for rights, I’m also grabbing a sledgehammer and going to town. … The only rights a slave has are that of property.
Youtube channel We’re In Hell has an exploration of the history of computers in war. As usual for this channel, it’s not a fun watch, but it does show the absurdity of war and AI fairly well.
We have EFTs via ABA numbers and they are common for B2B transactions. Retail customers prefer payment processors for the ability to partially or totally reverse fraudulent transactions, though; contrasting the fairly positive reputation of PayPal’s Venmo with the big banks’ Zelle, the latter doesn’t have as much fraud protection.
Now, you might argue that folks in the USA are too eager to transmit money to anybody that asks, and that they should put more effort into resisting being defrauded.
Side sneer: the table-saw quote comes from this skeet by Simon W. I’ve concluded that Simon doesn’t know much about the practice of woodworking, even though he seems to have looked up the basics of the history. Meanwhile I have this cool-looking chair design open in a side tab and hope to build a couple during July.
Here’s a better take! Slop-bots are like wood glue: a slurry of proteins that can join any two pieces of wood, Whatever their shapes may be, as long as they have a flat surface in common. (Don’t ask where the proteins come from.) It’s not hard to learn to mix in sawdust so that Whatever non-flat shapes can be joined. Or, if we start with flat pieces of Whatever wood, we can make plywood. Honestly, sawdust is inevitable and easier than planing, so just throw Whatever wood into a chipper and use the shards to make MDF. MDF is so cheap that we can imagine Whatever shape made with lumber, conceptually decompose it into Whatever pieces of MDF are manufactory, conceptually slice those pieces into Whatever is flat and easy to ship, and we get flat-paks.
So how did flat-paks change carpentry? Well, ignoring that my family has always made their own furniture in the garage, my grandparents bought from trusted family & friends, my parents bought from Eddie Bauer, and I buy from IKEA. My grandparents’ furniture was sold as part of their estate, my parents still have a few pieces like dining tables and chairs, and my furniture needs to be replaced every decade because it is cheap and falls apart relatively quickly. Similarly, using slop-bots to produce software is going to make a cheap good that needs to be replaced often and has high maintenance costs.
To be fair to Simon, the cheapness of IKEA furniture means that it can be readily hacked. I’ve hacked lots of my furniture precisely because I have a spare flat-pak in the closet! But software is already cheap to version and backup, so it can be hacked too.
Frankly this isn’t even half as good as their off-the-cuff comments two years ago. There’s a lot of poser energy here as they try to invoke the concepts of “senior engineer” and “CEO” as desirable, achievable, precise vocations rather than job titles. In particular, this bit:
Look, CEOs, I’m one of you so I get it.
This is one of the most out-of-touch positions I’ve ever seen. In no particular order: CEOs generally don’t understand, CEOs form a Big Club and you ain’t in it, CEOs don’t actually have power in their organization but delegate power flowing from the board of directors, CEOs are inherently disrespectable because their jobs are superfluous, and finally CEOs don’t take business advice from one-person companies unless it’s through a paid contract.
The job title naturally associated to a one-person limited-liability company is usually “manager” or “owner”, and it says nothing about job responsibilities.
Finally, while I think that their zest for fiction is admirable, it would help to critically consider what they’re endorsing. Dune’s Butlerian Jihad resulted in neo-Catholicism which effuses the narrative; it’s not a desirable outcome. Paraphrasing the Unabomber is fairly poor taste, especially considering that they are sitting in a city in Canada and not a shack in the wilderness of Montana.
Well, yes. It’s not a new concept; it was a staple of Cold War sci-fi like The Three Stigmata, and we know from studies of e.g. Pentacostal worship that it is pretty easy to broadcast a suggestion to a large group of vulnerable people and get at least some of them to radically alter their worldview. We also know a reliable formula for changing people’s beliefs; we use the same formula in sensitivity training as we did in MKUltra, including belief challenges, suspension of disbelief, induction/inception, lovebombing, and depersonalization. We also have a constant train of psychologists attempting to nudgelord society, gently pushing mass suggestions and trying to slowly change opinions at scale.
Fundamentally your sneer is a little incomplete. MKUltra wasn’t just about forcing people to challenge their beliefs via argumentation and occult indoctrination, but also psychoactive inhibition-lowering drugs. In this setting, the drugs are administered after institutionalization.
Read carefully. On p1-2, the judge makes it clear that “the incentive for human beings to create artistic and scientific works” is “the ability of copyright holders to make money from their works,” to the law, there isn’t any other reason to publish art. This is why I’m so dour on copyright, folks; it’s not for you who love to make art and prize it for its cultural impact and expressive power, but for folks who want to trade art for money.
On p3, a contrast appears between Chhabria and Alsup (yes, that Alsup); the latter knows what a computer is and how to program it, and this makes him less respectful of copyright overall. Chhabria doesn’t really hide that they think Meta didn’t earn their summary judgement, presumably because they disagree with Alsup about whether this is a “competitive or creative displacement.” That’s fair given the central pillar of the decision on p4:
Llama is not capable of generating enough text from the plantiffs’ books to matter, and the plaintiffs are not entitled to the market for licensing their works as AI training data.
An analogy might make this clearer. Suppose a transient person on a street corner is babbling. Occasionally they spout what sounds like a quote from a Star Wars film. Intrigued, we prompt the transient to recite the entirety of Star Wars, and they proceed to mostly recreate the original film, complete with sound effects and voice acting, only getting a few details wrong. Does it matter whether the transient paid to watch the original film (as opposed to somebody else paying the fee)? No, their recreation might be candid and yet not faithful enough to infringe. Is Lucas entitled to a licensing fee for every time the transient happens to learn something about Star Wars? Eh, not yet, but Disney’s working on it. This is why everybody is so concerned about whether the material was pirated, regardless of how it was paid for; they want to say that what’s disallowed is not the babbling on the street but the access to the copyrighted material itself.
Almost every technical claim on p8-9 is simplified to the point of incorrectness. They are talking points about Transformers turned into aphorisms and then axioms. The wrongest claim is on p9, that “to be able to generate a wide range of text … an LLM’s training data set must be large and diverse” (it need only be diverse, not large) followed by the claim that an LLM’s “memory” must be trained on books or equivalent “especially valuable training data” in order to “work with larger amounts of text at once” (conflating hyperparameters with learned parameters.) These claims show how the judge fails to actually engage with the technical details and thus paints with a broad brush dipped in the wrong color.
On p12, the technical wrongness overflows. Any language model can be forced to replicate a copyrighted work, or to avoid replication, by sampling techniques; this is why perplexity is so important as a metric. What would have genuinely been interesting is whether Llama is low-perplexity on the copyrighted works, not the rate of exact replications, since that’s the key to getting Llama to produce unlimited Harry Potter slash or whatever.
On p17 the judge ought to read up on how Shannon and Markov initially figured out information theory. LLMs read like Shannon’s model, and in that sense they’re just like humans: left to right, top to bottom, chunking characters into words, predicting shapes and punctuation. Pretending otherwise is powdered-wig sophistry or perhaps robophobia.
On p23 Meta cites fuckin’ Sega v. Accolade! This is how I know y’all don’t read the opinions; you’d be hyped too. I want to see them cite Galoob next. For those of you who don’t remember the 90s, the NES and Genesis were video game consoles, and these cases established our right to emulate them and write our own games for them.
p28-36 is the judge giving free legal advice. I find their line of argumentation tenuous. Consider Minions; Minions are bad, Minions are generic, and Minions can be used to crank out infinite amounts of slop. But, as established at the top, whoever owns Minions has the right to profit from Minions, and that is the lone incentive by which they go to market. However, Minions are arbitrary; there’s no reason why they should do well in the market, given how generic and bad they are. So if we accept their argument then copyright becomes an excuse for arbitrary winners to extract rent from cultural artifacts. For a serious example, look up the ironic commercialization of the Monopoly brand.
Top-level commenters would do well to read Authors Guild v Google, two decades ago. They’re also invited to rend their garments and gnash their teeth at Google, if they like.
Last Week Tonight’s rant of the week is about AI slop. A Youtube video is available here. Their presentation is sufficiently down-to-earth to be sharable with parents and extended family, focusing on fake viral videos spreading via Facebook, Instagram, and Pinterest; and dissecting several examples of slop in order to help inoculate the audience.
I have a slightly different timeline.
In general, I think that trying to frame our current century-long investigation into cybernetics as something recent, new, or unprecedented is ahistorical. While the general shape of AI winter can’t really be denied, it’s important to understand that it’s a cyclic system which will eventually yield another AI spring and AI summer. It’s also important to understand that the typical datacenter is not in financial trouble and there’s not going to be any great destroying-of-looms moment.