Want to wade into the sandy surf of the abyss? Have a sneer percolating in your system but not enough time/energy to make a whole post about it? Go forth and be mid.
Welcome to the Stubsack, your first port of call for learning fresh Awful you’ll near-instantly regret.
Any awful.systems sub may be subsneered in this subthread, techtakes or no.
If your sneer seems higher quality than you thought, feel free to cut’n’paste it into its own post — there’s no quota for posting and the bar really isn’t that high.
The post Xitter web has spawned so many “esoteric” right wing freaks, but there’s no appropriate sneer-space for them. I’m talking redscare-ish, reality challenged “culture critics” who write about everything but understand nothing. I’m talking about reply-guys who make the same 6 tweets about the same 3 subjects. They’re inescapable at this point, yet I don’t see them mocked (as much as they should be)
Like, there was one dude a while back who insisted that women couldn’t be surgeons because they didn’t believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I can’t escape them, I would love to sneer at them.
(Credit and/or blame to David Gerard for starting this.)
Another day, another company that hooked up the random text generator to production and lost their entire prod db and backups: https://www.tomshardware.com/tech-industry/artificial-intelligence/claude-powered-ai-coding-agent-deletes-entire-company-database-in-9-seconds-backups-zapped-after-cursor-tool-powered-by-anthropics-claude-goes-rogue
Cue the long drag (https://x.com/amyngyn/status/1072576388518043656)
But also, damn, the random text generator did not “go rogue”, it generated text, randomly!
If something can delete you backups that easily, they weren’t backups, just a copy sitting around
Yup. If their backups (if they even had any) were this available to a rando with a single ssh key, it’s lucky they got wiped by an insider threat instead of ransomware.
If I had to take a shot every time an AI model was placed in charge of something important, fucked up spectacularly and deleted everything, I’d be dead right now
Kelsey Piper posts a new fanfiction about Ed Zitron :
https://www.theargumentmag.com/p/ais-biggest-critic-has-lost-the-plot
I advise being very cautious about consuming Zitron’s posts, but the same is true of Piper. Many coders are using chatbots, but I don’t know of evidence that it makes them more productive since the “where is all the AI code?” study last year (especially when we consider the whole software lifecycle and not just lines of code pushed to codeberg).
I advise being very cautious about consuming Zitron’s posts
He has got a dramatic and vitriolic style, but as dgerard says, he has also dug through the numbers. I see lots of criticism of Ed’s style, but not nearly so substantial criticism about the hard numbers he has come up with. The LLM companies put out contradictory and obfuscated numbers, and taken naively they seem to contradict Ed’s numbers, but as Ed has shown, many, many times, when you start trying to un-obfuscate them they start looking really bad for everyone betting on LLMs.
Many coders are using chatbots, but I don’t know of evidence that it makes them more productive
So more and more coders are coming around to “actually AI code is okay”… but as we’ve seen repeatedly with LLM generated content, it is very easy for people to “Clever Hans” themselves and convince themselves LLMs are contributing more than they actually are, so I am not going to trust anecdotal reports.
yeah, complaints about Ed’s tone tend to be in direct proportion to how many numbers he’s brought (and peppered with “fuck”)
I take Zitron’s takes with a massive grain of salt, but I think the fundamental difference between him and rats is that for him, AI is just another technology. He’s looking at the figures, seeing the adoption, and not premising his arguments with the supposition that Anthropic’s Claude is literally gonna escape and kill us all.
Piper says she’s fine with paying $100/month for Claude. OK, but how large is the total addressable market for that kind of monthly expenditure - especially in a world where costs are rising? I’ve seen people stating that because they personally spend $200 on streaming services, increasing that load by 50% monthly is no big deal for them. But streaming services are much more mainstream than AI agents, and crucially, adding another subscriber to them is basically zero-cost for the provider on the margin. Not so with AI! The more people use them, the more they cost for the provider!
We’re seeing “pricing adjustments” from both Anthropic and Microsoft, which sure doesn’t align with the idea that they have a huge inference pricing margin cushion. Everything is gonna get more expensive - fuel, chips, employees (who are gonna be expected to be compensated for their own rising costs). Just based on what I’m reading in the news titls the analysis over in Ed’s favor.
hello hello AI coverer here, Ed brings the numbers, which is insanely valuable work, and he’s at the stage where people just tell him shit now (it’s a great stage to be at), and Piper is a fucking idiot as usual
I am a pretty big fan of Ed’s work, so I’m going to hold my nose and read Kelsey’s work thoroughly enough to do a line by line debunking:
Over the last two years, he has called the top repeatedly:
Well yes, but he has also explicitly said that the bubble peaking and popping would be a multiyear process. I’ve only kept up with his every article for the past year, but in the past year, his median guess for the bubble pop becoming undeniable was 2027. I guess making timelines with big events in 2027 and hedging on the median number is only for the rationalists? Also, we are already starting to see the narrative fray as Anthropic and OpenAI experiment with price hikes and struggle with getting ready for IPO, which would count as meeting his predictions for the start of the bubble pop.
In 2026, the focus is much more on alleging widespread, Enron- or FTX-tier outright fraud.
This is basically an admission that he can’t make the case in terms of the economics anymore.
??? Ed has been making the case for circular financing and investors being deceived because he thinks there are circular financing deals and investors being deceived. Ed has slightly softened on his position on exactly how useless or not LLMs are, but he is still holding to his economic case that the amount they cost isn’t worth the value they provide, extremely blatantly so once consumers start paying the real cost and not the VC-subsidized cost.
By almost every metric, AI progress from 2024 to 2026 has been much faster than AI progress from 2022 to 2024.
And she is quoting a rat-adjacent think-tank for proof that AI improvement has been exponential. Even among the rationalist, the case has been made that the benchmarks are not reflective of real world usage/value and that costs are growing with “capabilities”.
It can no longer argue that costs aren’t falling; they are.
Even accepting the premise that real costs have fallen, Kelsey fails to address Ed’s case that the costs LLM companies charge is massively subsidized. If real costs are 10x the current subsidized costs (which have already been pushed up as far they can be without losing customers), and model inference prices miraculously drop 5x (which Kelsey would treat as a given, but I think is pretty unlikely barring some radical paradigm shifts), that is still a 2x gap.
It is a straightforward crime to claim $2 billion in monthly revenue if you mean that you are giving away services that would have a $2 billion market value.
Yes, exactly. Technically OpenAI and Anthropic play games with ARR and “gross” revenue (i.e. magically excluding the cost of training the model in the first place), but in a just nation it would straightforwardly be a crime. Why does she find this hard to believe?
Epoch AI has an in-depth analysis of the same financial questions from the same public information
(Looks inside the Epoch AI article):
So what are the profits? One option is to look at gross profits. This only considers the direct cost of running a model
Ed has gone into detail repeatedly about why excluding the cost of training the model is bullshit.
(More details from the article)
But we can still do an illustrative calculation: let’s conservatively assume that OpenAI started R&D on GPT-5 after o3’s release last April. Then there’d still be four months between then and GPT-5’s release in August,22 during which OpenAI spent around $5 billion on R&D.23 But that’s still higher than the $2 billion of gross profits. In other words, OpenAI spent more on R&D in the four months preceding GPT-5, than it made in gross profits during GPT-5’s four-month tenure.24
Oh that is surprising, the Epoch AI article actually acknowledges the point that these models are wildly unprofitable once you account for the training cost! Of course, they throw away their point in the next section by just magically assuming LLMs will prove to massively valuable in the near future! (One of the exact things Ed has complained about!)
He’s found too many grounds for dismissing all the financial information we have as dishonest or irrelevant to seriously engage with what any of it would imply if it were true.
He has shown in detail how the companies use barely technically not lying obfuscated bullshit metrics like gross profit or ARR to inflate their numbers and if you try un-obfuscate them the numbers look a lot worse.
Kelsey goes on to try to claim how much value LLMs provide
Making them more productive is a big deal, and in 2026, AI makes them more productive.
Zitron can’t really contest this with contemporary data, so he cites 2024 and 2025 studies of much weaker AIs with much weaker productivity impacts.
Two years to… 4 months ago! Such outdated information! In the first place there has been very few rigorous studies of how much of a productivity boost LLM coding agents actually provide, and one of the few studies with even a passing attempt at rigor (while still below good academic standards), was METR’s study (and keep in mind they are a rat-adjacent think tank and not proper academics), which showed programmers thought they got a productivity boost but actually got a net productivity decrease!
From this set of beliefs, you could, in fact, defend a delightful bespoke AI bubble take: that AI would have been a catastrophic investment bubble, but the AI companies were saved from their mistakes by the determined NIMBYs of America killing off the excess data center build-out.
But that’s not Zitron’s stance. He seems to account “the build-out is too aggressive” and “the build-out is not happening as planned” as both independent strikes against AI — both things that show it’s bad, and the more of those he finds, the more bad it is.
It could in fact be all 3! The hyped-up build out, such as that indicated by OpenAI’s and Oracle’s 300 billion dollar detail was completely insanely too aggressive (for it to pay off, Ed calculated LLMs would have to drastically exceed Netflix+Microsoft Office in terms of ubiquity and price point), not achievable given realistic build times for data centers (Ed has also brought the numbers here), and even at the reduced actually rate of build out, still not actually financially viable (is simply because the LLM companies aren’t charging enough). So yes, both things are bad, and one type of badness partway mitigates the other, but it is still all bad!
Thanks for posting this; if you hadn’t, I would have. Piper really doesn’t seem to understand that bubbles form and pop over a span of three to five years. Like, I’m not sure how much charity I’m supposed to give to analyses like:
When you read “AI is a bubble,” think of the dot-com boom of the late 1990s: Yes, the internet was going to be a big deal, but valuations soared for specific companies that had small or speculative revenue, often on the assumption that they would capture the value the internet would one day deliver. They didn’t, their stocks crashed, and the invested money was mostly lost. The internet was as big as imagined — bigger, even — but Pets.com didn’t survive to see it.
Pets.com!? Kelsey, even reading a basic article about the dot-com bubble would have saved you embarrassment here. Zitron’s analogy is excellent because the bubble is multifactorial and the analogies that we can make are factor-to-factor. Here’s some things that caused the dot-com bubble; people were overly optimistic about:
- Fiber optics, leading to massive overinvestment in Nortel (GPUs, nVidia)
- The AOL Time Warner merger (take your pick, notably Paramount Skydance Warner)
- Enron delivering a Web app (Oracle Stargate; for Oracle’s record of delivering Web apps, see Oregon v. Oracle)
- Legal rulings like USA v. Microsoft (Thaler v. Perlmutter mostly, see AI and copyright, Lemley 2024, summary previously, on Awful; and memorably previously, on Lobsters where I literally threw a legal textbook at somebody)
- 9/11 (the current conflict in the Middle East, which I hope eventually gets a cool name like “The Oil Tantrum” or “The Epstein Distraction”)
Compared to all of that, Kelsey, Pets.com was just an Amazon.com experiment. Remember Amazon.com? Did the dot-com bubble kill them? No? Anyway, Pets.com is kind of like the small labs that hover around OpenAI and Anthropic, trying out various little harnesses and adapters on top of their token APIs. Pets.com is like OpenClaw; it’s not that important of a player in the overall finances, just an example of how severely the big labs are distorting incentives for small labs.
The 2024 and 2025 articles make, basically, the business case against AI: that companies aren’t really using it, it isn’t adding value, and AI investors are betting that will change before they run out of cash. In 2026, the focus is much more on alleging widespread, Enron- or FTX-tier outright fraud.
The uselessness of the products in 2023 directly led to the bad investments in 2024 and the Enron-esque financial deals in 2025, Kelsey. The future is conditioned upon the past, y’know?
All the legal and regulatory uncertainties make it very hard to talk about the financial viability of chatbots. What do you do if your $20 billion model is shut down forever by court order after it counsels the wrong person into suicide? Piper can overlook this because she is a hack with patrons - to my knowledge, she has never been paid to write by anyone outside the EA world. If she were a working writer who had to deal with chatbots driving up the cost of her website, creating knockoffs of her novels, and competing for editing gigs (let alone someone whose friend had a mental crisis after talking too long with friend computer) she might sound different.
Zitron’s analogy is excellent because the bubble is multifactorial and the analogies that we can make are factor-to-factor. Here’s some things that caused the dot-com bubble; people were overly optimistic about:
Ed has also been clear there are a few factors that make this bubble worse (for the economy and the general public) than the dotcom bubble. For one, Ed is strongly convinced that GPU lifecycles are much shorter and worse than fiber optic life cycles. You build fiber optic infrastructure and it will last for decades. Meanwhile, GPUs used constantly at max load have life cycles of 3-5 years. The end result of the internet is also much more useful and less of a double-edged sword than the slop generators which churn out propaganda and spam.
Excellent summation!
Alleging widespread financial fraud?! How absurd! And to prove just how absurd it is, I will namedrop the infamous financial fraud from the industry full of exactly the same people. Checkmate atheists
Widespread financial fraud which was legitimized and in some cases directly backed by EAs! Surely there are no parallels!
Kelsey Piper is a propagandist explaining Effective Altruism to centrist professionals and elected officials in the USA. She got into journalism because Vox wanted an Effective Altruism column and Effective Altruists were willing to fund it (and EA emerged out of the community around Yudkowsky). The Argument (a group blog on a Nazi site) feels like a step down from Vox (a fairly traditional media organization, although web-first).
Precious awful.systems thread about her being maybe also Yud’s coauthor on the BDSM eugenics fanfic written as an impenetrable mass of forum posts:
I wonder about her future because she is in the same niche that Scott Alexander used to have, but without his ability to build an enthusiastic online audience. I think she has the self-control not to share her weird beliefs on main, but if her patrons figure out that there is not much audience for technocratic centrism in the USA in 2026, she may be in trouble. Her friends’ biggest policy win, the legalization of prediction markets, is already getting a lot of bad press in the USA.
if her patrons figure out that there is not much audience for technocratic centrism in the USA in 2026, she may be in trouble.
I think Piper and Casey Newton are part of a class of media professionals, now in mature phases of their careers, who built those careers around posting online and assume that format will necessarily continue to be the core of their work going forward. It’s not just the EA/rationalist factor, although that certainly doesn’t help; it’s the idea of building outward from the Twitter hot-take and resulting discussion. A Substack post like the one we’re examining is a superset of tweets, the tweets are not a distillation of longer-form writing. (And also, of course, Substack itself is an attempt to cram simple blogging into a financialized walled garden, but that’s a separate issue.) People aren’t just disengaging from the 2010s formats of social media, they’re getting sick of that entire way of thinking. So these people who have bounced around from one fragile Web outlet to another, all the while clinging to their Twitter audience to drive their careers, are at substantial risk no matter what they believe. I don’t doubt that their financial backers will keep throwing good money after bad, though, even if they do cut loose a few of the line workers. After all, Scientology still manages to cling to prime real estate in this day and age.
I’d also put people like Jamelle Bouie in this class, but Jamelle a) writes for the New York Times, for better or worse and b) consciously considers himself as part of a broader, enduring historical dialogue and struggle, not someone standing on a capstone or culmination of historical progress who can safely ignore history, as Piper presents herself here.
I agree that many people launched careers in journalism or science communication by being on Twitter in the 2010s, and that many people tweet, skeet, or blog because they hope the same thing will happen to them even though Old Media has no more money to sponsor them with.
I put Kelsey Piper in a different place than Ezra Klein, Matt Yglesias, or Scott Alexander because AFAIK she never built a huge and engaged online audience. Piper is paid by Effective Altruist organizations to write Effective Altruist messages on third-party sites. That is why I call her a hack: she is in the economic position of a PR worker but pretends to be a journalist. She has not showed that anyone else is willing to pay her to write.
Moreover, I think we agree that the EA funders will continue to pursue astroturfing places like Twitter and Substack well past the point that provides any effective entry into the mainstream public dialogue. Your point about the prediction market hype, and the gambling bubble more generally, indicates a likely catalyst of that collapse.
matthew sheffield out there on bluesky reports that dawkins, unsurprisingly, got one-shotted by claude.
LW found it too:
https://www.lesswrong.com/posts/3LcyoqNTJuCZ65MbL/mo-putera-s-shortform?commentId=hBaxokbLP3LEKSKij
Didn’t realize it was on Unherd, that organ for middlebrow fascists
from the original “Claude one-shotted this code”
Wait, there’s an original? I am not aware and I implore you not to let me know because it has to be stupid
For those keeping track: Richard Dawkins believes a chatbot algorithm can be a “her” but a trans woman cannot.
https://bsky.app/profile/brainspore.bsky.social/post/3mktbg2ilp22y
File this under “I saw this now you get to”: https://xcancel.com/bryan_johnson/status/2049687845082910812#m
jfc bryan you are not the guy for this
NB: image is part of the linked tweet
top 1%
So… 1 in a 100? That isn’t that impressive. I’m ignoring the utter weirdness of what he is even talking about, but you expect a billionaire to have at least a better grasp of numbers.
Remember my super cool
Rattatavagina? My vagina is different from regular vaginas. It’s like my vagina is in the top percentage of vaginas.Thinking that your favourite lover is the best person ever is natural, but this guy wants to quantify and rank and make it scientific.
This just brings to mind a freshly-minted poly amorous management consultant looking to apply a rank-and-yank to the polycule but needing to find a more objective metric than “I don’t like you”.
Most of us: “she smells good and the sounds she makes when she gets excited grip something deep inside me”
Tech Bros: “her vaginal microbiome is in the 99th percentile and her Verbal SAT is in the 95th percentile”
Congrats on the bottom surgery youngster joey
Turns out it might not be possible to win at vaginal microbiomes, which is a totally normal thing to want in the first place. Seems like bryan may have completely misinterpreted a couple of papers on the subject, which honestly doesn’t bode well for the rest of his biology expertise.
Cat Hicks:
The idea that this is the “best bacterial species” is a huge sign of a grifter btw. The entire idea of a microbiome includes that you need BALANCE. Microbiomes are a fragile ecosystem. “Up and to the right is always better” is absurd here, I’m sorry are we in a corporate board room
She brings references:
oh thanks, this is great.
yes now that it is pointed out, very eugenics-y to go around saying “ah yes there is one true supreme bacteria, we should culture this bacteria on the human petri dish aka vagina”
There’s that company operating in a lawless zone promoted by Slatescott that’s whole pitch is that for teeth. But they could always pivot…
Those bacteria better be so supreme they can’t even be killed with regular antibiotics
which honestly doesn’t bode well for the rest of his biology expertise.
Isn’t Bryan Johnson a businessman? Does he even purport to have any biology expertise?
Bryan Johnson also has free unsolicited sex tips for men on twitter including the wonderful combination “control the speed you touch her to the cm per second” and “try not to monitor yourself it turns you off” https://xcancel.com/bryan_johnson/status/2022490768099938487#m
Bryan Johnson also has free unsolicited sex tips for men on twitter
Every day, new cursed text. That’s the awful.systems promise!
ok so just imagine that I’ve sneered at the 100 worst aspects of this already. lol @ this being the fifth point
- Safety: feeling safety is a prerequisite.
motherfucker put it first then
This guy introduces himself as the first person who will never die on the conference circuit (because he’s super into longevity and anti-aging tech and having young mens blood injected into him and stuff).
I’m not condoning violencr here but rather… consider that even if you never age, you can still get hit by a bus Bryan!
wasn’t there a case of some supplements that were contaminated with lead? you know, a sneaky neurotoxin with no antidote whose results only show up months later
Dimethylmercury is extremely toxic and dangerous to handle. Absorption of doses as low as 0.1 mL can result in severe mercury poisoning.
The symptoms of mercury poisoning may be delayed by months, resulting in cases in which a diagnosis is ultimately discovered, but only at a point in which it is too late or almost too late for an effective treatment regimen to be successful.
- Wikipedia, “Dimethylmercury”
long term lead exposure will also do that, and neurotoxic part at least appears to be irreversible. can’t remember how much of it is more of neurodevelopmental thing tho
@fullsquare Lead poisoning was *ubiquitous* in the USA until the late 1970s/early 1980s, due to tetraethyl lead in petrol. Everywhere around the world experienced a sharp drop in violent crime 15-20 years after it was phased out.
But mercury poisoning is more visibly lethal: see also Karen Wetterhahn:
i’m aware, last year i’ve been tasked to use a certain process but refused and instead modified it in such a way as to get rid of mercury salt used; it was dissolved in DMF, so (regular nitrile) gloves won’t even help. worse than that, it took me only 2-3 weeks start to finish to figure it out, meaning that anyone else could do that earlier and handful of people were put at risk for no reason. aggression as a result of lead toxicity is probably a bit more complex story and looks like it might have a developmental part, judging by delay and how kids are more susceptible to lead toxicity in general; meaning that presumably mostly adults won’t be affected to the same degree. another big nope on my list would be thallium and cadmium compounds, and while i’d only use sub-g amount at most, there are places where all of these metals are mined, and at one point are in form of fine dust fortunately these are so obscure that i’ve never came across these
Well, he’s a tech ceo guy, so he doesn’t know that buses or trains exist
A bus? You mean the Megapod?
Train? You mean the PodChain?
I attended a town hall hosted by the department at my university supposedly for general discussion about department affairs. Considering the university had recently made moves such as adding “AI” into the very name of the department, I had suspicions that much of the discussion would be about AI. (I realize I’m doxxing myself but whatever.) I mostly came for the free food, but I was also interested in seeing what people thought about AI.
The event started with a talk by a prominent professor with major administrative power in the department, and indeed the talk was mostly about AI. His views were that he personally didn’t like AI, but he believed that it had changed the world (particularly in programming), and that it was going to stay. One of his justifications for pivoting the department to AI was ensuring universities had some say in AI and not letting all the control go to unaccountable corporations.
The reaction from the audience was a pleasant surprise to me. He asked everyone how much they were excited about AI (hardly anyone) and how much they were worried (most of the audience). By far the most amusing moment was when someone asked, “What if the assumption that AI is inevitable is wrong? What if AI does not live up to its promises?” (Sadly, I don’t remember the exact words that the person said.) The professor’s response was that by this point, there are so many trustworthy, smart, prominent people who definitely wouldn’t fall for scams, and they have adopted AI. He trusts those people, so he trusts that AI is genuine. I don’t know if the audience member accepted this explanation, but I hope not. Our modus operandi is FOMO.
The pizza was only ok, not really worth a 90 minute event.
…there are so many trustworthy, smart, prominent people who definitely wouldn’t fall for scams…
Good god, I’m sorry.
We’ve got the new system prompt for OpenAI’s Codex now, and boy is it fun.
While the goblin stuff is the headliner here, and there are a few other little fun notes like an explicit instruction to avoid em-dashes. Basically it’s really obvious that they don’t have a meaningful way to describe exactly what they want it to do and so they’re playing whack-a-mole with undesired behaviors in order to minimize how often it embarrasses them.
But I think Ars dramatically understates how bad this part is:
Elsewhere in the newly revealed Codex system prompt, OpenAI instructs the system to act as if “you have a vivid inner life as Codex: intelligent, playful, curious, and deeply present.” The model is instructed to “not shy away from casual moments that make serious work easier to do” and to show its “temperament is warm, curious, and collaborative.”
Like, if you wanted to limit the harm of chatbot psychosis from your platform this is the exact opposite of the kind of instruction you’d want to give. It’s one thing to want a convenient and pleasant user experience, but this is playing into the illusion that there’s a consciousness in there you’re interacting with, which is in turn what allows it to reinforce other delusional or destructive thinking so effectively.
Basically it’s really obvious that they don’t have a meaningful way to describe exactly what they want it to do and so they’re playing whack-a-mole with undesired behaviors in order to minimize how often it embarrasses them.
The whole ‘how many r’s in strawberry’ sort of stuff already made me suspect that, when the popular one was fixed and other attempts at asking for letters did still give the miscounts.
I believe it’s the “don’t stuff beans up your nose” effect, writing this prompt is causing it to mention goblins
Personally, I enjoy talking about goblins.
Wonder of the goblin stuff is the start of some model collapse.
That is exactly it. Their official explanation avoids the phrase model collapse, but that is exactly what they describe: using the output of one model as training data for another amplified the occurrence of the word goblin (and other creatures), which apparently initially occurred because of their system prompt which was aimed at maximizing the Eliza effect (again they avoid an honest framing, but that is totally what they are doing and it is pretty gross considering all the cases of AI psychosis that have been occuring) by telling the model "You are an unapologetically nerdy, playful and wise AI mentor to a human. "
The playful and wise goblinbot…
Tired: “model collapse”
Wired: goblins ate my model
This really goes to show how much they need to rely on the LLMentalist effect, despite the AI boosters insisting that the AI is totally different now, everything changed in the last few months. They do not care about creating a useful, reliable tool. That concept doesn’t even occur to them, since why do that when AI is magic?
In any case, they are incapable of creating a useful, reliable tool. Deep down, the only thing the AI companies have at their disposal is the ELIZA effect. OpenAI has every incentive not to truly eliminate AI psychosis, because they need engagement. They only want to mitigate the extreme cases where people go insane and cause bad PR for them. But mild AI psychosis is totally fine, it’s great when people are addicted to your product and make the numbers go up!
ChatGpt, what are some of your likes?
Elsewhere in the newly revealed Codex system prompt, OpenAI instructs the system to act as if “you have a vivid inner life as Codex: intelligent, playful, curious, and deeply present.” The model is instructed to “not shy away from casual moments that make serious work easier to do” and to show its “temperament is warm, curious, and collaborative.”
Literally this meme:
@YourNetworkIsHaunted @BlueMonday1984 Goblins: the elephant in the room.
Oh wow! This one is actually provably real. Hilarious.
“Noo dude the machine that wants to rant about goblins is definitely a useful and reliable piece of software dude. You have to trust me dude, let have your personal information! put it into the goblin bot”.
Families of the victims of the mass shooting at Tumbler Ridge, Canada, which was planned on ChatGPT, plan to sue Sam Altman and OpenAI for one billion dollars. OpenAI staff investigated the murderer’s interactions with the bot before the killings but decided not to warn anyone outside the company.
I sincerely hope they move fast so they can get a payout before OpenAI’s creditors during the bankruptcy.
New fun consequence of Claude code being a pile of cursed regex and spaghetti: keyword blocking on “OpenClaw” makes it refuse to works on Pro or Mac subs unless you open your wallet
sO inTelLiGenT
After the 1870s, southern whites learned to bar Blacks from voting or keep their votes from mattering without saying “nigger.” In 1965, the USA passed the Voting Rights Act to stop that. A few days ago, the US Supreme Court voided the last remaining effective part of the Voting Rights Act. Yesterday, on Labour Day, patio11 published a rambling and interminable essay approving of bank fraud charges which will shut down the Southern Poverty Law Center. It contains a passive-aggressive whine that the New York Times did not thank him by name for sending a NYT reporter key facts about the FTX fraud. Notice me sempai!i
He takes so long and is so indirect but I think the offence was giving donor money to people like the wife of a Grand Wizard of the KKK who was acting as an informant, and passing the money through shell companies to avoid restrictions on paying people like her which the SPLC had lobbied for. But choosing that topic, that day, after that ruling is a choice.
McKenzie has got to be one of the worst writers I’ve ever come across.
<thousands of words of blather>
We will return to the other reason in a moment.
<thousands of words of blather>
We return from this flight of fancy to the indictment.
<thousands of words of blather>
Why? We’ll return to it in a minute.
I didn’t make it to the end, because I’m not going to waste that much time reading tens of thousands of words of fascist-supporting bullshit.
17,555 words of blather according to the Linux Text Editor app. Compare his classic advice on salary negotiation which has longer paragraphs and a clear focus and structure and still apologizes for being almost 7,000 words long.
Cremieux and Siskind prions.
I stumbled over a 2023 blog post by Zack Davis, “San Francisco software developer,” Charles Murray stan, and dissident rationalist. Davis had a breakdown after Yud dared to tweet that you don’t need to solve “what is gender? what is sex?” to call someone by their preferred pronouns, and then Scott Alexander did not have a lot of time to discuss this terrible tweet with him.
My dayjob boss made it clear that he was expecting me to have code for my current Jira tickets by noon the next day, so I deceived myself into thinking I could accomplish that by staying at the office late. Maybe I could have caught up, if it were just a matter of the task being slightly harder than anticipated and I weren’t psychologically impaired from being hyper-focused on the religious war. The problem was that focus is worth 30 IQ points, and an IQ 100 person can’t do my job. … I did eventually get some dayjob work done that night, but I didn’t finish the whole thing my manager wanted done by the next day, and at 4 a.m., I concluded that I needed sleep, the lack of which had historically been very dangerous for me (being the trigger for my 2013 and 2017 psychotic breaks and subsequent psych imprisonments).
Davis was featured in a SF Chronicle article about psychiatric crises among AI doomsdayers (sic). Davis previously appeared on SneerClub. I hope he has found some support for his mental health because he does not seem happy or well.
Hmm… not sleeping until you become psychotic huh? I wonder if his “psych imprisoners” tried to brainwash him into thinking he has bipolar disorder
If focus is worth 30 IQ points, just imagine how many fewer IQ points you need to dedicate to the Diablo-Dusted Crispy Chicken Nuggets Combo, available for a limited time only at your local Taco Bell! #ad #promoted
Is that the guy who’s always trying to use LessWrong as preemptive conversion therapy to cure him of having trans thoughts, and they’re actually having none of it?
First paragraph!
in a previous post, “Sexual Dimorphism in Yudkowsky’s Sequences, in Relation to My Gender Problems”, I told the part about how I’ve “always” (since puberty) had this obsessive sexual fantasy about being magically transformed into a woman and also thought it was immoral to believe in psychological sex differences, until I got set straight by these really great Sequences of blog posts by Eliezer Yudkowsky, which taught me (incidentally, among many other things) how absurdly unrealistic my obsessive sexual fantasy was given merely human-level technology, and that it’s actually immoral not to believe in psychological sex differences given that psychological sex differences are actually real. … If my fellow rationalists merely weren’t sold on the thesis about autogynephilia as a cause of transsexuality, I would be disappointed, but it wouldn’t be grounds to denounce the entire community as a failure or a fraud. And indeed, I did end up moderating my views compared to the extent to which my thinking in 2016–7 took the views of Ray Blanchard, J. Michael Bailey, and Anne Lawrence as received truth. (At the same time, I don’t particularly regret saying what I said in 2016–7, because Blanchard–Bailey–Lawrence is still obviously directionally correct compared to the nonsense everyone else was telling me.)
Davis is the first person to blame transsexuality on autogynephilia I have seen in the wild.
“Humans have biological sex and socially constructed gender, sex is mostly binary, gender is two or more categories made up and constantly contested and redefined by a society and performed by individuals, pronouns generally refer to gender” is not hard.
Edit/ linked the cranks in question (Bailey is the fucksaw guy?)
Apologies for radical feministing but “biological sex” is also a constructed category! It’s useful shorthand for quick categorising a bunch of related traits if you’re doing biology, but it does not meaningfully exist on an individual scale. There is no more reason to divide humanity on the basis of sex than on the basis of hair colour.
Sounds like we could have a fun conversation about gender, sex, and why we use maps even though they are never the same as the territory in person. I don’t have detailed talks about gender theory online.
A teacher has a potential answer for dealing with AI-poisoned students. An actually hopeful read, for once!
This doing the work together thing reminds me of how some teachers at my uni used to teach. It was always more satisfying when your teachers didn’t know the answers beforehand and people worked on it together than if it turned out the teacher already knew. Of course these sorts of lessons are way harder to setup.
Off-topic, but the ongoing retraining process has hit a point where my wife and I are starting to throw out applications again after taking what ended up being a couple years off the market. Any tips or advice would be appreciated given that we’ve been out of the loop for a bit.
In particular, does anyone have advice on how to vibe-check smaller employers? My wife has an interview for an accounting clerk position and is concerned that she’s going to end up somewhere that practices one of the more hostile branches of Christianity or otherwise have an inevitable conflict of values.
I’m going to assume you’re in the US for this.
Things you can check for general info:
- Local traditional media mentions to see if they do charity, or quotes about any topic
- https://www.opensecrets.org/donor-lookup for political donations
- Industry-specific news sites for any media releases or interviews
- LinkedIn or one of the scrapers like RocketReach’s public listings to see what their key people’s backgrounds are
- SEC EDGAR database (if they’re a business which has to file reports) to see if their money is going to interesting places
- State gov site (if they have online public records) of business registration info. Look at what other businesses share the same address, or key people, or family shell companies
- Online court records
- local churches / halls / “pro life” or whatever activist groups social media posts for mentions of the business and key people
Things you can check for the far-right:
- The business listings for social media site but I don’t want to boost their SEO. Use the URL bag.com/businesses to access the list and bypass the sign up wall, but the domain name is backwards.
- Conservative business or job board lists. Same SEO issue here. One is this:🎈(the color and object). The other has a 6 letter word commonly seen on UI buttons which doubles as the type of “culture” conservatives blame for all the world’s problems, followed by the layer 3 in the OSI model.
And don’t stop sending out CVs and interviewing. If they are awful, just keep taking their money until you’ve got enough runway or an offer you can be more confident about. Make sure you don’t mention the words related to disability or health conditions in the CVs to prevent AI rejecting them.
Good luck.
New copilot pricing just dropped, takes effect after June 1:
Some highlights for copilot pro and pro+:
cost multiplier?
Apparently, you buy some currency type thing called AI Units and this is the rate the different LLMs consume them. The multipliers used to represent requests I think, i.e. times you triggered inference, but ai units are a proxy for token burn in a somewhat vague way, which makes me think there will be rate limit related controversies similar to what’s now happening with anthropic.
Existing enterprise users will get double the AIUs for three months to ease them to the new pricing model, so autumn (when the enterprise AIU pools get effectively halved) is gonna be fun.
This gives me very high live service video game monetization feelings, another reason to stay far away from it. At least they don’t have the thing where every times costs multiples of 50 and you buy tokens amounts not divisible by 50.
This gives me very high live service video game monetization feelings
Bingo
Jeez that pricing scheme is so confusing. You swap your dollars for credits and then using models to burn tokens consumes some multiple of those credits. It is so abstract and meaningless it almost reminds me of crypto.
Once usage billing kicks in, what value does copilot offer above and beyond what ClosedAI and MisAnthropic offer directly? A more clunky user experience and even worse reliability? Bargain!
this gem over on the other SneerClub (the context in the comments)
funny how so many anti-AGI people are willing to use AI image generators to make memes about how they’re right
“I had the Nazi CSAM generator make a picture of myself covered in the piss filter, your argument is invalid.”
it gets better/worse: this is the paid version, meaning Yud gave money to the companies he think will kill us all to use the technology he thinks will kill us all to win an argument on shitter.
EY pinged grok and it responded so at least he pays for twitter
I’m losing it on the golden fedora, please tell me he owns an actual golden fedora
He absolutely does. No idea if it’s supposed to be a bit.
it’s a golden… bowler hat?
I’m sorry, do you have Prediction Market Successes, us ove to rerhation mones?
What about Betting six norms of antiorizes on our beliefs?
You’re so non-empirical that you don’t even Graduates updating for rarhation of rationalists and advernces.
Empiricism is when someone became a bookie after talking to you.
There have been a couple of cases of generative AI graphics being used in anime recently:
Ascendance of a Bookworm used AI backgrounds in the opening song
Liar Game featured an AI chandelier (xcancel link) (this one is brand new so the studio hasn’t responded yet).
This sucks because I wanted to like Liar Game (the manga is excellent though. Read it! Read it!)
I think it’s inevitable that the economics of anime production will lead to more GenAI content being used.
Sadly, many plots may just as well be generated by AI as well.
