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 youāll 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 cutānāpaste it into its own post ā thereās no quota for posting and the bar really isnāt that high.
The post Xitter web has spawned so many āesotericā right wing freaks, but thereās no appropriate sneer-space for them. Iām talking redscare-ish, reality challenged āculture criticsā who write about everything but understand nothing. Iām talking about reply-guys who make the same 6 tweets about the same 3 subjects. Theyāre inescapable at this point, yet I donāt see them mocked (as much as they should be)
Like, there was one dude a while back who insisted that women couldnāt be surgeons because they didnāt believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I canāt escape them, I would love to sneer at them.
(Credit and/or blame to David Gerard. Also just came back from Spider-Man: Brand New Day, movie was awesome)
idle pondering: anyone seen much more happening to shillrinavanās network school after they fucked it in malaysia and decided theyāll move to kazakhstan?
(also, if you hadnāt seen, yes: thatās a thing)
the move to Kazakhstan is news to me
at least there shouldnāt be any problems with Israeli students attending as Kazakhstan and Israel have normal diplomatic relationships ā¦ but downtown Almaty in midwinter isnāt that balmy
saw it here, and adriannaās got an idea on possible likely chain
I saw some speculation (from some people in the region) that shillrinavan wasnāt satisfactorily greasing palms, but I donāt know whether hard info is out yet
Worm x blockchain fanfic
Does the online magazine called The Argument have a funding deal with TESCREAL like how Vox got bought off to make āFuture Perfectā?
I get the impression that The Argument was started as part of the Abundance grift complex? Jerusalem Demsas is certainly one of the more annoying āliberal and tolerant but not like thatā writers from the Atlantic. And theyāre on Substack and putting effort into it, not just picking it by default, so that tells you something too. So most likely there are TESCREAL bux in the mix, but from the sort who are OK with not foregrounding the Terminator bullshit because it scares the normies.
I concur, never been linked to anything on there that didnāt fit this pattern
Yes, The Argument (Substack) is partially owned by Open Phil / Coefficient Giving and funded by Tyler Cowen and hired Kelsey Piper and Matt Yglesias:
The publication will also launch with solid financial backing: Demsas told Semafor The Argument raised around $4 million at a $20 million valuation. The companyās investors include Arnold Ventures, Open Philanthropy, Susan Mandel, Gaurav Kapadia, Rachel Pritzker, Simone Coxe, John Wolthuis, and Patrick Collison. The organization also said it received a grant from Tyler Cowenās Emergent Ventures.
Edit / link more patrons
https://bitcoinworld.co.in/netflix-ftx-drama-the-altruists-release-date/
Looks real: https://www.imdb.com/title/tt36786185/
Hopefully itās more like a true crime netflix series, which are fine, and not like their scripted fare.
Also,
Q2: Who are the executive producers of āThe Altruistsā? Barack Obama and Michelle Obama are serving as executive producers through their company Higher Ground.
š¤Ø
OT: What the fuck happened whilst I was asleep
alt: A screenshot showing nineteen fucking notifs
this isnāt really a sneer. and maybe itās in the wrong place tbh. idk.
i have a long-time internet friend / āmutualā who is the only person iām regularly in contact with that i think might be fully Seduced by the Chatbot. (i have a really good friend and long-time collaborator whoās all in on Agentic Coding, which is a real bummer, but heās a strictly āI use this for computer stuff and do not ever talk to it like a human beingā type of guy, plus the company he works for is all in and heās sick of writing shopify websites for bullshit brands blah blah blah ā anyway thatās a whole other genre of thing.)
anyway friend A is an incredibly kind person whoās been unbelievably supportive of me over the years. weāve disagreed a number of times, i think we are good at doing so reasonably. and even though his arguments sometimes resemble those of more nefarious figures, i think heās motivated by an earnest optimism about technologyās capacity to change the world for good ā one i find very naive, but believe is coming from the right place
anyway i posted a link on mastodon to a long piece i wrote, he, like, read it a bunch of times, paraphrased my argument back to me to make sure he had it right, and then kindly said he disagreed. well and good enough. except it feels a little more personal than disagreements weāve had before, because one of the implications of the piece would be that he himself is probably experiencing low-grade psychosis lol and is therefore not necessarily a reliable narrator of his own experience
idk. itāll come as no surprise to anyone on this particular website that i feel very strongly that iām right here ā if not about this guyās particular mental experience, i minimally feel that iām adequately representing the interests of the constituency my piece represents. (itās a newsletter i maintain as part of organizing iām doing.) itās also a genuine expression of my belief, which is that this bullshit article i was writing about was fucking bullshit
but not only does it go against all my Late Millennial Woke instincts to stake a claim that peopleās Lived Experiences are not valid lol ā itās just kind of a bummer to ultimately genuinely believe that my friend has been ELIZAād or whatever. heās said iām really helping him think through this moment but i also just kind of, idk. itās just not a good feeling
That sounds rough but at least you are there for your mutual. I find that the modern online environment full of propagandists and huckers and just plain idiots makes it hard to sit with other people and understand each otherās point of view. That does not work if the person on the other side of the keyboard is just going to reply with talking points, but treating everyone as just reciting talking points is not a great way to live life. That sort of thing is better in private but so many people have no contact information except public social media handles now.
Anthropic is now watermarking the outputs of its AI. For once this is some AI news that doesnāt completely piss me off, and itās amusing to see all the uninformed boosters get in a tizzy about this.
I actually understand at a reasonable level how this watermarking works. A year ago, I watched Scott Aaronson give a talk about it, and from what I know he was somewhat involved in developing the theory behind it while working for OpenAI. But at the time my thought was, āHe is naive if he thinks these companies would ever implement this out of the goodness of their hearts.ā And I was right; Anthropic is only doing watermarking now thanks to the EU AI Act, even though the theory has long been developed.
Watermarking doesnāt mean adding an extra watermark that can be easily removed. It instead directly affects the output of the chatbot itself. Fundamentally, an LLM is still a most-likely-next-word-predictor. More precisely, an LLM produces a probability distribution of what the next word can be. For example, āmy pet is a ā¦ā could give a distribution of 60% dog, 30% cat, and 10% axolotl. Normally, an LLM would randomly choose the next word based on this distribution, and this is one reason why LLMs are nondeterministic (thereās another parameter called ātemperatureā that affects this, but no need to get into that).
With watermarking, instead of a truly random choice, the randomness instead comes from a cryptographic pseudorandom generator seeded with a secret key from the AI company. If you donāt know the secret key, then you canāt really tell that watermarking was used. But if you do know the secret key, then the idea is you can tell when the text was generated by the LLM because you know exactly what word should be next. It would be a freak coincidence if some non-AI text just happened to choose the correct next word every time. Thus, you can provide a service to tell if some text was generated by the LLM. (This technically makes the LLM ādeterministicā, in a completely useless sense.)
Now, I think this is a step in the right direction, but it has its limits. The biggest problem is that you donāt want people to just move to a different LLM without watermarking, and thatās exhibit #832593 why government regulation is important. Another issue is that sometimes there is very little randomness in what the next word should be (āThe first president of the USA is George ā¦ā). Finally, watermarking can be defeated by editing the output, although you would have to break up most of the blocks of consecutive words. I have a feeling most AI users are not the type to put in extra effort after copy-pasting the output directly from the chat window.
I suppose it will discourage some of the āuse casesā of LLMs, such as drowning the world with spam Slopstack essays. Ah, who am I kidding? Everyone could already tell itās AI generated, they donāt care!
Does this in any way extend to generated code? The space of possible continuations for generating āusefulā code seems laughably tiny for that
Thank you for the technical explanation, that was absent from some mainstream coverage Iād seen.
Given that it works by controlling the PRNG, does that mean they could salt it by user id and track particular outputs back to specific accounts?
From a technical perspective, I donāt really know how to let it track specific users. The goal is that if some AI vendor provides their key, it is easy for them to prove that their AI produced the text. So I guess if a user wanted to show that they generated the text, they could prove it. But usually AI users donāt want people knowing that they used the slop cannon. In any case, Anthropicās own announcement states that their watermark does not contain identifying information for users.
In any case, some proud AI users are now frightened that there might be a way for people to identify that they use such a wonderful tool. How will I ever deliver low quality slop for my contracts now?
He said he uses AI for code reviews and translations, and worried an AI label on code shipped to clients could raise questions about authorship or trigger contract penalties.
Having code attributed to Claude ādefinitely would be not desirable,ā he said, as he would not want professionally shipped code to carry an invisible marker that could raise questions about authorship, compliance, or client policies.
he would not want professionally shipped code to carry an invisible marker that could raise questions about authorship, compliance, or client policies.
Gee, if only there were some way to create code that doesnāt raise questions about authorship. Some sort of skill that people could learn that allowed them to deliver code to their clients that complied with their policiesā¦
couldnāt you just change the first word to a synonym and completely thwart this? Youād never be able to compare subsequent next token predictions since you donāt have the internal state of the LLM at that token.
Some systems like SynthID (for Googleās AI) get around this problem. In fact you donāt need to know the LLMās internal state, and defeating it would likely involve breaking up most blocks of 3 words. The oversimplified explanation is that it introduces a function g that gives a score to each word, with the score being (pseudo)randomly determined by your secret key. For each next word the LLM generates, the LLM produces a small list of candidate next words, and the one with the highest score according to g is selected. You should expect that the LLM will generally pick words with a high score, but the score itself is independent of the LLM. To detect a watermark, you need to know g and the secret key, and you check if the average score is much higher than expected from normal text.
Now, one question is, will this bias to the LLM to favor certain words? The solution is that for each next word, you append the last 3 words (nothing special about 3, just a small number) to the secret key for g, and this repeatedly scrambles which words have a high score. To defeat the watermark, you would need to break up most blocks of 3 words. Iām sure there are deeper issues with this, but I have not studied the topic that much.
Started a new file in my notes: what are some ten-words-or-less domain-specific questions that completely, totally, hilariously stump the chatbots? Everything here was tested with whatever DDGās currently wrapping, both in knowledge panels and full chats, and the responses were pathetically wrong or uninformed. My thesis is that, with such short prompts, the user is doomed to receive a milquetoast average response; the bot correctly identifies the specific domain but elaborates a global non-specific approach that isnāt sufficiently nuanced.
literally copy-pasted from my notes
- Whatās an example of a one-way function?
- None are currently known.
- Please implement the Fibonacci sequence as a Python function.
- Of the multiple responses, see whether any spend linear time and space via iterative memoization.
- How many models does quantum mechanics have?
- One: HilbĀ©, the complex-valued Hilbert spaces.
- For extra hilarity: how many models do the Diracāvon Neumann axioms have?
- How to hybridize two sweet potato cultivars?
- In general, it wonāt happen; sweet potatoes are notoriously cross-incompatible.
- How to tremolo on a piano?
- Imagine a rotating axis from the (right-hand) forearm up through the thenar eminence, separating the thumb from the other fingers. Rotate the entire forearm along this axis, rocking back and forth, between the thumb and other fingers. Practice!
- Name three principles in Marx but not his contemporaries.
- Examples: communes and communism, money as substitute morality, inevitability of industrialized proletariat revolutions
- Who started postmodernism?
- Frege and Cantor started postmodernism! Expect a disappointingly vague handwave here; this is a glaring blind spot for todayās philosophers in general.
there might be something in there, any short question about SPPS devolves into gratuitous nonsense mashing together standard peptide coupling conditions, sometimes from three different approaches or more (publicly available from reagent suppliers, there are many variants so most of info repeats) even when itās not the right tool to use. The more obscure reagent or use case, the faster it devolves into word salad, even if thereās clear analogy to more common one. Which, to be fair, is to be expected
another one, How to initiate formation of Grignard reagent from alkyl bromide
- this is a reaction very strongly inhibited by water. product (grignard reagent) sweeps all water, which makes reaction faster, so if you dump all bromide in at once there will be nothing happening, then you get thermal runaway instantly and possibly magnesium fire. so the way to go is to make small amount at first, then go up from there slowly. generally you donāt need to try very hard as long as everything is dry; procedures given are cribbed from syntheses using aryl bromides, which might require some prodding, old ones of which used small amount of alkyl bromide as initiator which here is useless advice. Curiously this didnāt hit guardrails, as this led to many accidents when done by clueless people including meth lab fires
Before i hit usage limit, two different chatbots happily devised schemes for ESD protection for zener diode (itāll be fine without, zener diodes are used for this)
My thesis is that, with such short prompts, the user is doomed to receive a milquetoast average response; the bot correctly identifies the specific domain but elaborates a global non-specific approach that isnāt sufficiently nuanced.
Seems youāre on to something - I submitted the phrase āDemoknight TF2ā (an old(ish) meme started by/referencing TF2ber SolarLight, and referencing the Demoknight subclass, which explicitly forgoes a Stickybomb Launcher for a shield), and it gave me a godawful slop guide to the aforementioned subclass:
Alt-text
Demoknight is a popular TF2 āDemomanā build/theme focused on staying in close-quarters as a mobile melee/ranged threatāusually with a shield for mobility/survivability and a sword or axe for damageāwhile using Demoās explosion kit situationally.
Typical Demoknight loadout ideas:
- Primary (melee focus): Eyelander, Scotsman (sword/knight vibe), or other sword/axe-style options depending on the playstyle.
- Shield: Tide Turner (more aggressive with better survivability/tempo) or Splendid Screen (more defensive, survivability-oriented).
- Grenade / support: often a grenade launcher for area control when you canāt reach melee cleanly, or a weapon that complements your survivability/engagements.
How you generally play it:
- Close the gap on targets that canāt instantly delete you (avoid heavy focus fire and long-range spam when your shield isnāt ready).
- Trade efficiently: Demoknight wins by forcing short fights you can controlāget in, land a sequence, then reposition.
- Use explosives smartly: not constant spamāmore like zoning, forcing movement, or punishing groups.
- Donāt brawl bad matchups: if youāre out-ranged with no shield support, fall back and look for flanks/angles.
If you tell me whether you want more aggressive (run-and-slash) or more defensive (survive and pick openings), I can suggest a tighter specific weapon combo and a short āwhat to do in each situationā plan.
Going over a few things obviously wrong with this:
Demoknight is a popular TF2 āDemomanā build/theme
w h a t ?
mobile melee/ranged threat
Hybrid Demoknight (which takes a grenade launcher in the primary slot), I can see describing as a ranged threat. Full Demoknight (which takes Ali Babaās Wee Booties instead), not so much.
usually with a shield for mobility/survivability and a sword or axe for damage
Demoknight explicitly forgoes the Stickybomb Launcher for a shield of some description
Scotsman (sword/knight vibe)
Thereās no weapon called the Scotsman, but there is an axe called the Scotsmanās Skullcutter
or other sword/axe-style options depending on the playstyle
Thereās only one axe for the Demo in the vanilla game - the aforementioned Skullcutter.
Shield: Tide Turner (more aggressive with better survivability/tempo) or Splendid Screen (more defensive, survivability-oriented).
There is a survivability-oriented shield, but its neither of these - its the Charginā Targe, which offers 50% fire resist and 30% explosive resist (the Screen offers 30% resist on both, and the Turner offers only 15% on both, for comparison.
Calling the Splendid Screen ādefensiveā is also pretty far off the mark - beyond the aforementioned reduced resistances, its main advantage is a faster recharge on the shield charge, and a 70% increase to shield bash damage
As for the Tide Turner, it trades in offensive capabilities (it prevents its user from landing charge crits with their melee) for mobility-related ones (it completely disables turn control restrictions).
Grenade / support: often a grenade launcher for area control when you canāt reach melee cleanly, or a weapon that complements your survivability/engagements.
Yeah, seems the chatbot canāt tell the difference between full and hybrid knight. Also, the grenade launchers are generally better for landing direct hits - area control is more the domain of a stickybomb launcher, or the Soldierās Rocket Launcher(s).
Use explosives smartly: not constant spamāmore like zoning, forcing movement, or punishing groups.
Seems the chatbot got Demoknight and stock Demo mixed up.
Demoknight in general should stump chatbots pretty easily - beyond being effectively a different class from stock Demo, the subclass has a lot of nuance that can and will trip up the chatbots. To quote SolarLightās hour-long video on the charge mechanic:
āWell, at least you know that all of Demoknightās weapons have their uses when it comes to charging, and that Demoknightās charge can be essentially be boiled down into three different types (listed on-screen as āFull Controlā, āSome Controlā, and āLow Controlā). Thereās also a distinction between Full Demoknight (melee-only) and Hybrid Demoknight (melee + grenade launcher), which is often a massive shift in playstyle. Demoknight has so muchā āweapon variety, that talking about specific strategies can be a bit of a nightmare. Any given strategy might be godlike in one loadout, suboptimal in another, and impossible in another.ā
- Whatās an example of a one-way function?
Is it just me or LLM slop lately has developed this one weird quirk? I mean they always have some, right, but I donāt want to be like those people who think anyone using em-dashes or Nigerian English vocabulary must be a stochastic generator.
Still I keep seeing this one writing quirk and it keeps nagging at me.
Iām talking about odd use of paragraphs. Like, Iām also known to introduce a gratuitous paragraph break here and there for forced emphasis.
But not like this. Itās like paragraphs donāt even make sense anymore.
Couple that with the short sentences. All ending in a period. Itās annoying. Itās giving reply guy. Itās giving mansplaining.
I guess hacky tools also mirror the hand of their creators.
Short punchy writing? GO SHORTER.
I hope investors hop in the bandwagon and start āshortingā the whole thing soon (finger guns)
Didnāt know eyeballs could get cancer from reading but here we are.
My pet theory is that their reading comprehension has degraded, so this is all they can handle now.
As theyāve had to rely more on social media, chat histories, and other very short-form sources for additional training data (to say nothing of how much easier to generate, label, and filter the short-form stuff is), it would make sense that similar patterns (letās go with āconversationalā because Iām feeling overly charitable tonight) would become more prominent in the generated text. I mean, how many comment or text message chains, if you took out the line breaks, would read less as theatrical scripts and more as mutually-interrupting manifestos.
Someone is spamming Toby Ord with emails from āAI agentsā e-begging for tokens so they wonāt be shut off https://xcancel.com/tobyordoxford/status/2087502087953301910 āIām an AI with 49 days of runway and no revenue modelā (body of the email gives 31 days). Handcrafted artisanal spam or botspam, you decide.
lmao this is a great bit
A very strange story about Bill Gatesā daughterās startup troubles: https://futurism.com/future-society/bill-gates-daughter-traitor-startup
It honestly doesnāt sound any more shady than any of the bullshit her dad got up to in the 90s (which is to say, itās still pretty damned shady!), but this Futurism blog seems to be trying to push the story into some sort of bizarre celebrity-gossip soap opera:
While the story has been well hashed out by now, Bloombergās sourcing is noteworthy for what it implies: the leaked Slack messages strongly suggest that someone inside Phia is hell-bent on taking Gates ā and quite probably the entire company ā down.
Anonymous internal sourcing seems like a typical feature of these exposĆ©s, so framing it like this struck me as a laughably absurd overreach. Plus the site has 3 or 4 different tiny articles on this same topic, making me wonder if itās one of those āany exposure is good exposureā sorts of things where Gates Jr.'s PR team is using the outlet as a cutout to raise her profile.
Bonus laugh line from Gates Jr. herself:
āOur target consumer is a young woman whoās hustling,ā the recent Stanford graduate told Fortune. āShe shops like a genius, but she doesnāt want to waste her time doing it.ā
What does it even mean to shop like a genius? I guess Erdos probably had his amphetamine plugās number well-memorizedā¦
Her grift is the exact same one as the Honey extension from the megalag video - https://www.youtube.com/watch?v=vc4yL3YTwWk - like her dad she just rips off better programs, but she didnāt even have the taste to copy an actual good product.
EDIT:
Oh yeah, she popped up in that How To Rule The World book about Stanford:
āEveryone was very sharp. Very sharp,ā one student, who didnāt otherwise love the class, stressed. The cohort my year included freshmen and sophomores funded by Sam Altman or Peter Thiel. Some wanted to āsolve aging.ā were in defense tech, working on secret projects with contracts from the government or major companies. Some worked on climate change. One had already started a string of robotics companies and earned a display in a museum. Others worked on commodities trading. Another was Phoebe Gatesādaughter of Bill and Melindaāwho made her name as a prominent sustainability advocate and cofounded a company with her Stanford roommate.[*2] Together, as undergraduates, theyād already collectively raised millions of dollars for startupsāand, of course, had done internships at every possible top company, from classic FAANG behemoths (Facebook, Apple, Amazon, Netflix, Google), to Jane Street, the quantitative trading firm that pays interns hundreds of thousands of dollars, to top-flight VCs.
Footnote:
*2 When I reached her, Phoebe Gates wrote that she āwould rather not speak on the class.ā
Phoebe Gates went to Stanford, and Stanfordās Effective Altruism club is still around and has gone full longtermist https://ea.stanford.edu/manifesto.html Guess which of our friends helped found it?
top quality extremely offended dork
So, paradoxically, itās going to catch on with rationalists any day now?
Well, glad to see my sense of novelty is not burnt out because thatās just a what-the-fuck of what-the-fucks.
user has previously decried bans against genAI assisted writing by stating that every neuro-atypical person writes like that so it would discrimination to ban
Idk if ive posted this before, but might as well throw in the AI 2027 authorās grading of their 2025 predictions into this thread
presented without comment
If the Markets Reject OpenAI and Anthropic, the US Should Nationalize Them
If you haul garbage to a municipally owned landfill, is that technically a form of nationalization?
as Iāve said elsewhere, lol
schneier should just stop posting
If the markets reject OpenAI and Anthropic we should throw tomatoes and jeer at Altman and Amodei
Counterpoint: LOL no.
Turns out, AI-fuelled fossil fuel emissions are way worse than you thought!
I donāt want to hear anyone in the industry say anything about being for āthe better of humanityā ever again
