➴➴➴Æ🜔Ɲ.Ƈꭚ⍴𝔥єɼ👩🏻‍💻

@riverpunk@defcon.social @alice@lgbtqia.space Google intentionally made their search engine shitty so that people would have to make multiple searches. They are the _least_ incentivized to be accurate.

So, turns out #AI actually did plagiarize its solution to a #math problem. I don't like their characterizations as plagiarism machines, but in this case it applies.

I want to clarify something. I'm not trying to allow people to search through Fedi content, though I guess that would be possible. I'm trying to use AP to replace search engine's with FOAF bookmarking of websites.

What is EHS? Well, let's say you were trying to determine the average height. You could formulate a hypothesis that everyone Is 6 feet tall. Then measure people against that, and then use the error to update the set of reasonable hypotheses. This is basically what Bayesian data analysis is. EHS is similar, but you're basically forming a hypothesis, attempting to prove something by it, finding the contradiction, and then using that to form the next one. This existed before #LLMs, the first time I was aware of it was for the four-color theorem.

So what does that mean for AI art in general? Well, you probably already noticed that most #AI art looks like AI art slop. Why? Because it was specifically trained by asking people what they liked more. So, the #AIArt style is a distillation of the median taste of the demographics of the sorts of people that use AI Art generators. (Though, I'm pretty sure the yellow comes from being trained on a lot of movie data.) But, you probably also know it doesn't have to look like that. 2/

Thinking you are doing something by refusing to use #AI is like convincing yourself you're saving the planet because you buy your produce at the farmer's market.

@meatbag@dragonscave.space @Robinzz3@aus.social @3goodthings@fedigroups.social @DigitalCoup@fedigroups.social @meditation@fedigroups.social @economics_that_works@fedigroups.social @startrek@fedigroups.social The best dictation tools use generative AI in order to correct the text afterwards.

The other is Concrete Harm. Now, don't get me wrong. Supporting Anthropic, Microslop, Meta, and OpenAI produces concrete harms. I'm less clear on whether or not supporting Alibaba is producing concrete harms. Especially since it's actively harming those four. What I can say definitively is that running an open weights model locally provides zero material support to any of these companies. So, what externalities am I causing by using a Qwen 3.6 model on my MacBook? 18/

@dahukanna@mastodon.social Oh, the 'making millions of requests' is a problem. It's also one that can be addressed by... not doing that. It's almost entirely bigtech doing that, not someone's local model. But that is a valid point.

1. Challenge expertise with substance. A person should not be believed merely because of their credentials. However, you need to address the substance of the topic and not evade the central theme. A technical response identifies what variable, mechanism, or inference is missing.
2. I don't engage with Sea Lioning. Shifting the goalpost is a form of Sea Lioning. Bringing up irrelevancies is Sea Lioning. Insisting that I have to engage with you is Sea Lioning.

4/

1. Dispute the actual claim I am making. Ideally quote it. Do not substitute it with a nearby claim that someone from an opposing coalition might make. There is a good chance I don't agree with that coalition. I endeavor to say exactly what I mean.
2. Do not flatten the logical form. If I put a statement in a conditional, then I am not asserting the consequent unless the antecedent is true. An analogy is not an equivalence. A critique of evidence is not a critique of the problem.

2/

@ohir@social.vivaldi.net @primeapple@polymaths.social @tivasyk@awawa.club You have quite simply shifted the goalpost. The issue of hallucinations is entirely separate from the issue of cognitive alliance. Your other statement was a nonsequitor.

@alice@lgbtqia.space What did the sticker on a different pair of shoes say? Stickers, by their nature, are fungible.

@murdoc@autistics.life Yes, but it'll be a minute. I have witchcraft to attend to.

@dahukanna@mastodon.social If I buy fast fashion second hand, then I'm not supporting fast fashion. If I'm using an open weight model, likewise, I'm not materially supporting its creation. Does that make sense? If I'm paying for a model (which I have) then I am materially supporting the harms. But an open weight model doesn't contribute anything to its originator.

@dahukanna@mastodon.social Yes, but how does using a model which has already done this things encourage more of those things?

1. State what you are actually disagreeing with. If you are taking issue with the implications of an analysis, don't attack the statistics. If you are disputing facts I am going to trust that the facts are the problem and not the moral premise or terminology.
2. Ask me questions about what I believe. Do not assume that just because my argument reminds you of some argument you didn't agree with that we actually disagree with each other.

3/

Okay. I have to leave the app again for the day. I should be applying to jobs right now, and I'm starting to pick fights again. Which I'm not supposed to be doing. I'm really trying #Hecate.

@tinker@infosec.exchange Oh, I relish the day. An implosion of big tech followed by massive amounts of affordable #ram and people fine-tuning open source models in their basements. Perfect.

If there are externalities can I compensate for them? In my opinion, the #AntiAI crowd is actually doing the work for #Anthropic and I've wondered how much of the rhetoric is pushed by them. Anthropic's business model is to make everyone afraid of #LLM #technology in order to regulate their #monopoly. The biggest threat to humanity is that the technology is controlled by the top of the K-shaped economy. So, real AI education is fighting the megacorps. 19/

I know someone else is going to come on here and be like "I have #ADHD and..." My dude (yes, I'm assuming this will be a male), my problems aren't your problems. Some people with EDS are wheelchair bound and others just can't play soccer. ADHD is literally the only reason I'm not a millionaire. I could have used my skills to trade of FTC filings before that became common place except my severe inattentive ADHD doesn't let me work for money. 17/

Okay, I'm going to wrap up. I'll pin this thread here and soon I'll launch a blog where I'll write about AI more. I'm not going to spend my time arguing here or anywhere else about #AI. Particularly with people who clearly lack any understanding of the technology. As soon as my ADHD lets me, I'll launch my blog and link to it here. /END

@ukrdef@mstdn.social I don't think Russia is going to be capable of fighting anyone right now.

So, I think it's really funny that you can tell if your research is really novel by feeding it to an AI and then having the AI state it's wrong because it doesn't look like anything it's read before.

@ohir@social.vivaldi.net @primeapple@polymaths.social @tivasyk@awawa.club How is that different from a calculator? The article you linked to, which I was already familiar with, does not address the clause "necessary" in my question to you even though you for some reason chose to highlight it. The citation specifically also shows some diminishment through use of a search engine. Does your opinion apply to that? https://lgbtqia.space/@AeonCypher/117003117215139601

Happy Lunar Lughnasadh

So, within the last month four different groups have come forward with positions on AI. You have Anthropic and OpenAI that want to stop open models. You have bigtech who has lost the AI battle to have open models. You have Meta who wants it to be radically unregulated. You have former researchers who want frontier models to be slowed down. There is definitely going to be shifts in the marketplace soon.

@dimitriselectif@libretooth.gr If you can't use anything produced by #AI you can't talk to me. Remember that whole thing about Sociotechincal Systems? Well, AI has fundamentally modified my cognition. My writing skills have improved dramatically from its use. So, even through this thread wasn't made by AI. The AI's taint now lives embodied within me. /END2

@juergen_hubert@mementomori.social I am objecting, because you don't get to take several hundred years of Philosophy Of Science and then start claiming that things are settled.

@ohir@social.vivaldi.net @primeapple@polymaths.social @tivasyk@awawa.club How does using an open weight model help anyone? What evidence is there that "stupider" is a _necessary_ consequence? If you are honest with yourself, this is entirely about tribal affiliation.

@juergen_hubert@mementomori.social This is a really new definition. A theory is a comprehensive explanation of a set of fact. I don't know when this definition showed up, but it was not the definition used by Popper of Khun. It is true that "It's just a theory" was never a correct sentence. To constitute a theory it has to explain the data.

@dzwiedziu@mastodon.social It's not an ad hominem, My supsicion earlier that you don't know what you're talking about is proven. "The Relationship Between Firearm Availability and Suicide" That is not the statement I made. Is it? It demonstrates your inability to understand the argument.

Oh... for the record. I think the #EU #AI labelling rule actually doesn't go far enough. The rule says the label is required for "Deepfake" material (trying to pass off a work as authentic) and for "public interest" work. It still allows you to sell unlabeled #AI slop. If there were a practical way to do it, I would demand all slop (e.g. content mills) to be labelled. Substantive human review and responsibility is the requirement for public interest work and that is below the floor.

@ohir@social.vivaldi.net @primeapple@polymaths.social @tivasyk@awawa.club I have to remind myself that I made this in order to avoid fruitless embattlement. https://lgbtqia.space/@AeonCypher/117003157784844728

@dimitriselectif@libretooth.gr Yeah, except the majority of widely shared rhetoric isn't reasonable, saying it's wonderful technology and can hep people.

The majority Anti-AI talking points are:

• It's dangerous
• It's useless
• "They took out jobs".

But perhaps most telling is that the recent discussions around it in open source where people are literally stating they won't use any software where AI has written any code.

Which has made this an issue of purity.

Which reminds me...

21/

So... #OpenAI is claiming their "containment breach" (which I continue to think is #bullshit) means they aren't training new #AI. It's awfully convenient that it's training new #LLM models that is the expensive part of their company.

There's a whole problem of what constitutes "AI Generated Work". The EU had to explicitly define the scope, and I don't think it goes far enough. I also think that requiring anything that has used AI in any part of the production anywhere as being therefore "AI Generated" is also wrong. Steam's policy right now is pretty good as it asks the author to explain exactly how it was used in the production of the game.

@tivasyk@awawa.club Are you seriously making this argument?

@flyingpenguin@infosec.exchange The design of the new fighter plane that they are building is actually proof of this. It's like an 8 year old boy wrote a wishlist. It reminds me of the Homer Simpson car. The USA is going to spend a ton of money on a bloated expensive POS.

@dzwiedziu@mastodon.social That makes it _even less interesting_. The fact that a think designed to kill happens to kill is _even less surprising_,. "Gun Availability Increases Gun Deaths" is a statistical statement of the for p(X) is conditioned on p(Y). If p(Y) increases the base rate of p(X) then the statement is just a normal statistical statement. The fact is uninteresting. Doesn't matter what guns are or what they do, the statement "Gun Availability Increases Gun Deaths" is literally self-entailed.

I have a controversial point. For the most part #Fuck #Farmers. By and large they are #millionairre #capitalists #welfare recipients who are obsessed with crying all day and blaming other people for their bullshit.

@juergen_hubert@mementomori.social The "Theory Of Evolution" is no more or less a theory than the "Lamarkian Theory Of Evolution" except that the latter theory proved to be both more complicated and less able to fit the available evidence. So the Darwinian Theory Of Evolution won out over the Lamarkian one, but the Lamarkian theory was still a theory.

Since I'm not supposed to be arguing with people I am going to need to start thinking about what I want to do with HTs I see. Just saw a thing saying that using AI is giving private corporations personal data. I wanted to point out that this isn't true with privacy first providers like tinfoil.sh, or with local AI. But that would be engaging, so I should... write it down as a thing to blog about later?

Mild tangent. People mostly don't like Throne Of Glass and #Romantasy because women like it. That's the reason. It's just #sexism.

@dzwiedziu@mastodon.social You can make a statement like "A gun is designed for violence as therefore..." that's fine and an interesting statement. Gun Availability increases Gun Deaths makes the anti-gun person making the statement look like an idiot that understands neither guns nor basic logical entailment. If you can't understand logical entailment, then you can't understand any argument whatsoever, and you can at best play the role of human stochastic parrot.