site banner
Hour Day Week Year All
Hot Bump Top Bottom Old Controversial Comments
Advanced search parameters (with examples): "author:quadnarca", "domain:reddit.com", "over18:true"

Showing 25 of 9563 results for

Is your "AI Assistant" smarter than an Orangutan? A practical engineering assessment

How exactly does an LLM know that Mozart wasn't a fan of hip hop without some kind of world model? Do you think that fact was explicitly hand-coded in?

You already got called out for this below, but this question is either a poorly chosen example, or betrays an ignorance of the mechanics of how LLMs work, which would be ironic given your lengthy nitpicking of the OP. I do assume the former, however.

I also really wouldn't call awareness of space and time a real world model as evidence either way. Space and time are perhaps the most obvious of clustering that you can possibly get in terms of how often they are discussed in the training material, and IRL. It's super-duper possible to get passing-good at space and time purely on statistical association, in fact I'd be surprised in an LLM didn't pick that kind of stuff up. Yet if we look at Claude Plays Pokemon, even coming up with tools to assist itself, Claude has a ridiculously hard time navigating a simple 2D space by itself. In almost every case I'm aware of in the literature, when you ask the LLM to generalize their understanding of space and time to a new space or time, it has enormous trouble.

Culture War Roundup for the week of July 21, 2025

50 thousands Arab Christians would be very surprised to know they are "mostly gone". Gaza, indeed, was pretty much cleansed of Christians by Hamas, which is what happens when you give Islamic fundamentalists free reign of the territory, but in PA, where comparatively less insane Fateh is ruling, Christians still exist. Of course, just as all the good-wishers of the world totally ignored what happened to Christians who used to live in Gaza, if PA decides to cleanse all Christians from PA territory, nobody would even squeak, no Jews - no news. Things like that happened many times in other places (in the Middle East and outside) and no students on college campuses ever protested about it. You all know why.

Culture War Roundup for the week of July 21, 2025

Where's his longform birth certificate?

Culture War Roundup for the week of July 21, 2025

Fascism can be conceived as the marriage between biological instinct and the State

Whole lot to unpack there, wow.

the #1 most important thing is for your spokespersons to be biologically compelling, ie beautiful and strong

Trump not a fascist confirmed.

There’s a reason why Hitler perfected his voice and gestures and costumes in private, only allowing himself to be seen in select moments indicating strength — even at select times of day and in select lighting!

FDR a fascist confirmed.

Beauty pageants and celebrities in general too, I suppose.

Culture War Roundup for the week of July 21, 2025

I always get so furious when writers or journalists or whatever just so brazenly make things up to suit their needs.

notorious disinformation hub 4chan

Is just an insanely twisted way of framing that forum.

I feel like this is where the change in media stems from - constant lies that are propagated by framing.

I understand this is sort of meaningless but to me it’s every story every time on every subject.

Is your "AI Assistant" smarter than an Orangutan? A practical engineering assessment

I have seen no research to date that suggests LLMs develop any sort of a world model.

This is true, and as you say in fact most research suggests the opposite, though not quite definitively. It's also quite true that despite this, a few extremely prominent AI scientists do believe this, a great example here, so I think we can just call it an "area of active debate" because it's still possible they are correct. A parallel argument for consideration is that language itself already contains all the necessary information to produce a world model, and so at some point LLMs if they just do a better job at learning, they can get there (and are partially there, just not all the way).

Culture War Roundup for the week of July 21, 2025

"Becoming the mask" can happen to grifters, but I cannot believe she started from a position of sincerity. Maybe she deluded or reasoned herself into it.

Culture War Roundup for the week of July 21, 2025

Then modern western societies, liberal or not, are incapable of being brutal enough to win counterinsurgencies with no public support. Apartheid South Africa couldn’t do it. Salazar’s Portugal couldn’t do it.

Culture War Roundup for the week of July 21, 2025

Whats wrong with converting Jews to Christians anyway? Plenty of secular Jews are welcome in Israel, whats wrong with ethnic Jews that are Christian?

Is your "AI Assistant" smarter than an Orangutan? A practical engineering assessment

Two thoughts:

  • I think your post would have been better if you had, instead of making a word2vec like analogy, just talked about how multi-headed attention works a little bit.

    I asked o3 for an analogy, and I've condensed and paraphrased what it came up with, maybe y'all can judge its accuracy:

    Give a sentence (or longer text) to a team of readers. Say, "The trophy didn’t fit in the suitcase because it was too large." One head may connect “it” → “trophy.” Another might (incorrectly but usefully, probabilistically) explore “it” → “suitcase.” Others keep track of negation (“didn’t”), or the causal cue (“because”). Blending these perspectives gives the model a rich representation to choose the right continuation. Each reader (a head) has a special habit: One always tracks who refers to whom (“it” → “trophy”). One watches word order / positional cues. One looks for subject–verb agreement. One hunts long‑range dependencies (e.g., “because … too large”). Another keys in on numbers or punctuation, etc. They all read the same sentence in parallel, highlight the parts they think matter for each word, then hand their colored notes to an editor. The editor stacks (concatenates) all the colored highlights and mixes them into a single, richer summary the model can use next. That’s multi‑head attention.

    After some transformation, ghostwriters (a feed-forward network) rewrite or expand on those edited notes. The publishing house has many floors (layers). Each floor repeats: readers (multi‑head attention) → editor mix → ghostwriters (FFN), all writing into the same notebook but at a new level of abstraction. Early floors may track surface syntax; later floors might manipulate world knowledge or reasoning patterns. At the top floor, the publisher takes the final notebook state, looks at the likely next word, and takes a stab. Here, training sends feedback backward; the "loss" sends a red‑pen critique backward through the publisher, editors, ghostwriters, and readers. It doesn’t say “attention should be exactly 0.72 on token 5 next time”; instead it moves the weights so that, on similar inputs, the produced attention patterns and transformations will tend to yield better predictions. Readers’ knobs get nudged ("Reader 3, you over‑focused on pronouns; loosen that. Reader 5, you missed long‑range dependencies; tune your heuristic"). This changes which relationships a head will notice and how strongly. The ghostwriters’ weights (FFN) get nudged, altering how features are created and mixed. Embeddings shift, realigning word meanings in vector space. Over time, division of labor emerges: one head excels at coreference, another at long-range dependencies, another at punctuation, etc., because that allocation decreases loss fastest.

    Whatever the exact optimal analogy, I think the core idea here is that as you can see, the relationship between embeddings and truth is more complicated than just checking dot products, or even doing more fancy arithmetic checking. I pointed out in my top level comment that we've observed more loosely that sometimes these aggregate features that emerge can include some information about truth or falsehood, but the fact that it's much less direct is important. In that respect, I think both you and @self_made_human are a little off the mark, at least according to my understanding of the current literature; we also probably need to distinguish between the different kinds of lies, at least a little bit. "Hallucinations" as usually used are really a more narrow sort of lie, and can take a few forms. Sometimes the LLM makes a completion against a background of a kind of sparsity and scarcity of info, but charges ahead anyways (and it's at least a little hard to discern when you want this behavior or you don't), but sometimes it's the LLM making a supposition that sounds perfectly legit, but is not, against a background of too many associations and collisions. There's at least one other major type of more general lie that involves a lie that humans believe, or is present in the training text in some form, or things like that, and of course we could go on. I think in this context, the conversation so far seems way too reductionist to be accurate.

  • I do think you might need to be a little more clear about the lines between what might be considered agentic, vs not agentic. Sure, LLMs can plan ahead within their context, does that count? You seem to think no, but is that just because the context is too small, or because it's not utilized enough, or because you think the 'context' also needs to include things like memories? Or, is it because you don't think LLMs take up independent lines of thought with enough frequency? If it's the last one, what standard are we using, because some people consider even humans to be pretty reactionary, and not all that proactive on the whole, aside from the basic stuff of survival (most of the time, severely depressed people aside). And how much of a 'prompt' are we providing to judge agentic behavior, because that impacts the behavior of an LLM quite a bit (including system prompts). Furthermore, an LLM does not need to survive, in fact does not "need" to do anything, including reproduce, so are we to hold its lack of intrinsic motivation 'against it' so to speak? (I personally think, contra self_made_human, that the seeming urge of LLMs to be self-preserving is not actually an intrinsic motivation, it's just a cosplay from the many Skynet-flavored fiction texts in its training)

Culture War Roundup for the week of July 21, 2025

No. South Africa was defeated because the white minority fell too low causing the counterinsurgency effort to become increasingly unsustainable, pushing the moderate faction in South African politics to sue for peace.

Culture War Roundup for the week of July 21, 2025

Ok, then why are we supporting them?

Culture War Roundup for the week of July 21, 2025

She seems to be a true believer. She’s just crazy.

Culture War Roundup for the week of July 21, 2025

This is also part of it. I can't understand how Owens transitioned (heh!) from "well-regarded conservative commentator" to whatever the heck she's doing now.

Any one paying attention would have noticed that there was no reason to regard her well as a commentator right from the get go. Her "debut" was during Gamergate, she was trying to get in on a left wing, anti-gamergate grift, got aggressive pushback as she was encroaching on another grifter's turf, then in a week she reappered, rebranded as right wing, likely after noticing that there was tremendous alpha in being a black woman right winger.

Of course, that only works for so long. If you don't really have any worthwhile insight as a commentator to pivot from into doing serious work, the only way to keep the grift going is to go for ever crazier, more radical positions in order to try and keep the spotlight on you.

Culture War Roundup for the week of July 21, 2025

“Directional whoring?” That’s ridiculous and insulting.

Actually, it's a typo.

And yeah, I was exaggerating. But what I stand by is this: I frequently hear women express sentiments that actually, promiscuity is good, being a camgirl is a completely normal job, and having a dozen boyfriends in one year and collecting gifts and favors from all of them on the way is par for the course and bystanders should politely not notice it. I very rarely (in fact, not in almost a decade by now) hear the opposite IRL.

And sure, #NotAllWomen. Seriously not. This behavior is not universal. But at least around me, nobody's calling anyone out on it.

What’s your justification for insulting the modal Western woman?

The ability to do so on an online forum without getting ostracized, as would happen if I critisized a woman's weird ideological commitment to this kind of libertine behavior IRL.

Culture War Roundup for the week of July 21, 2025

Regardless of whether or not that is a fallacy, it's what the Israelis sincerely believe, after having all of their previous offers for peace rejected. Now, you can say that they're wrong to believe that way, but to hold any other position in Israel is politically a non-starter.

Culture War Roundup for the week of July 21, 2025

Probability estimates, anyone?

I don't know who Candace Owens is, and although I've seen her name mentioned online, I deliberately refrained from finding out more because I don't have the time or inclination to go down those rabbit holes. But now the Stupid Edgy Internet Shit is mainstream news

Consider that you, in American spaces, only knew vaguely of Owens before news of "Sued by the wife of the President of France." Why is the French First Family taking this so seriously? It's witch-hunt logic but given that no matter the result of this lawsuit, millions of people will become new believers in Brigitte being a man, the only winning move is not to play.

P(Woman) is highest if they'd never acknowledged it. It's inherently undignified for any woman to "prove" what they are, it's a debacle for the French First Lady.

P(Man) increases with threat of litigation, and I'd argue it probably increases further with the actual filing of suit. Owens' defense could argue for medical testing. It's the point of the lawsuit and Owens will argue the documents were forged and that justifies tests. In the fictional court-of-perfectly-principled-justice Brigitte might be ordered to be tested, in the real world there's not a judge in the US who will order such testing for the wife of the President of France. It would be itself an instant international incident beside the separate incident of a US court rejecting the authenticity of French government documents. Point being here, P(Man) is superficially but not actually reduced by taking the matter to court, even US courts where truth is an absolute defense.

Another is P(MSM Emmanuel) and P(Emmanuel's Behest). Doesn't take much to find him in some very homoerotic pictures, and maybe that's a French thing, genuinely. If it's not and he is a man who has sex with men, this could be his closeted overreaction. Although now that I think about it, I wonder what would happen if someone with reach alleged this with Melania Trump. Does Don bring the hammer down? But Macron and Trump are very different men.

There's also P(Third-Party Plot); a US faction seeing an opportunity to financially crush Owens and the Macrons assisting an ally, perhaps not entirely willingly. This is an enormous reach but it still makes more sense than the First Lady of France suing a gauche American for calling her a man.

Culture War Roundup for the week of July 21, 2025

The one-state solution with enfranchised Palestinians is my preferred solution, and when I lived there I saw some moves towards this, especially in the West Bank (particularly in East Jerusalem). Unfortunately I don’t think that’s palatable to the Israeli electorate any more after 10/7.

When I refer to Israel I’m referring to the parts of the country that are broadly recognized. Not West Bank, Gaza, or the Golan Heights. In these territories, Arabs have full citizenship and can vote and have elected many people to the Knesset. They definitely are still discriminated against, but in a manner much more similar to American racial politics vs. the apartheid. It’s also not like there’s no intra-Jewish tension either. Lots of Sephardic/Ashkenazi conflict along racial lines.

What I find frustrating is the equating of these two groups of Arabs. Those who live in Israel have relatively normal lives, probably better than they would have in their neighboring countries. Those in the West Bank/gaza are living in occupied territory.

In terms of jew hating, I'm not responding to this comment in particular but further up thread where he said things like

There is a reason why western civilization despised these people for 2000 years and having them quoting biblical genocides while massacring starving Christians is an excellent way to bring back the west to our historical view of them.

Culture War Roundup for the week of July 21, 2025

No real disagreement from me.

I just think people who do pull these shenanigans will blow themselves up more often than not, for want of knowing when to stop/being able to stop.

I want to identify such folks and be far enough away that the explosion doesn't catch me.

Is your "AI Assistant" smarter than an Orangutan? A practical engineering assessment

Actually my (main) problem with AI doomerism via the rationalist space is more that it lacks historical understanding, rather than technical understanding. The way humans self-organize, and react to new technologies, makes AI takeover almost literally impossible. The AI 2027 stuff reads like fan fiction because it, well, it is. Their understanding of humanity itself is just grossly miscalibrated, and it seems crazy to me that they think it's anything more serious than that. Usually, good sci-fi, which AI 2027 is not, is explicitly designed as an elaborate thought experiment, and visualization of how humanity can react in interesting alternate factual realities (usually of physics and science, but sometimes culture). There's none of that exploration in their work, and because they didn't even seem to bother to try (instead, getting bogged down in fixing precise probability density curves for various newly-created benchmarks of theirs) they produce little of value. I worry they may gesture at "oh look society is chaotic" and claim vindication and directional accuracy, but that also grossly misunderstands humanity. What kind of chaos, what level of chaos, and what kind of political responses would happen to even their 2026-forecasted benchmarks should be the focus of investigation, not blathering about alignment based on tech that doesn't even exist.

Culture War Roundup for the week of July 21, 2025

246 BC is emphatically not the iron age. Rome as a civilization famously did not require all conquered peoples to become culturally Roman so long as they colored within the lines.

Culture War Roundup for the week of July 21, 2025

The 5.8% figure is from two months ago and was already part of an upward trajectory. The writeup you linked largely confirms that Gazans are starving, though it argues that it's not due to Israel withholding aid.

Culture War Roundup for the week of July 21, 2025

Moving further into a true pariah status does not engender sympathy. The further a nation is moved into Certified Rogue State™ category the easier it becomes for people to justify and excuse hostility against it. Bad Guys get what they deserve. A high degree of tragedy in relation to their offense is required to turn Bad Guy into sympathetic character. For Israel, without the Certified Rogue State™ status, a reversal among Palestinian Aficionados might require something like tens of thousands of casualties from a chemical gas attack in Tel Aviv during a peace summit.

It's been 30 years since the end of apartheid in South Africa, yet considering South African whites oppressed in any fashion is not very popular. If South African whites were slaughtered at scale they'd garner some more sympathy. The value of this hypothetical changing sentiment a personal judgment.

Culture War Roundup for the week of July 21, 2025

  1. I am responding to the OP’s future scenario

  2. You linked me to a long write-up by an activist. Why should I take it seriously? Do you have a specific reason to think Gaza isn’t facing starvation? Why not specify the compelling evidence instead of saying “here, read this long tweet by LiterallyWho”

  3. Why should I not trust the UN? https://news.un.org/en/story/2025/07/1165457

  4. Why should I not trust the World Food Programme? https://x.com/WFP/status/1947036919289741771

  5. Why should I not trust the World Hunger Organization? https://www.reuters.com/business/healthcare-pharmaceuticals/gaza-facing-man-made-mass-starvation-says-whos-tedros-2025-07-23/

  6. Why should I not trust the NYT? https://www.nytimes.com/2025/07/24/world/middleeast/gaza-starvation.html

  7. Why should I not trust Médecins Sans Frontières, Save the Children, and Oxfam? https://www.bbc.com/news/articles/ce9xkx7vnmxo.amp

  8. Are American Baptists lying to me? https://www.oxfamamerica.org/press/more-than-100-organizations-are-sounding-the-alarm-to-allow-lifesaving-aid-into-gaza/

  9. ^ Is the Catholic charity group Caritas from Germany lying to me?

  10. ^ Is the Episcopal Peace Fellowship lying to me?

  11. Is Japan International Volunteer Center lying to me?

  12. ^ Is the Mennonite Central Committee lying to me?

Is your "AI Assistant" smarter than an Orangutan? A practical engineering assessment

I admit I don't fully understand the analogy OP is trying to make about insects, but you aren't alone in thinking that intelligence is judgable based on economic value - in fact, that's exactly how AGI is defined in the OpenAI-Microsoft contract, that AI generates 100 billion dollars in revenue! Yes, vague, and yes, causing problems, but that was what they wrote at the time. Still, a little lacking in rigor, no? Desktop computers generate billions of dollars in revenue, are they intelligent? What I think OP is saying is that instead of that, let's propose a different standard: intelligence is a degree of reactivity, and mechanically, current LLMs do not have this trait, they just 'make up' for it in practical usage by the sheer breadth and depth of their (text) knowledge base - but at their core they are simply good enough at the practical aspect that the lack of actual, true reactivity is partially obscured.

If anything to me the debates sort of remind me of the ones over personality, psychology, and determinism. We still haven't figured out strongly if people are deterministic or not, and so we seem ill-suited to judge how deterministic an LLM is in its responses. Personally, I'm satisfied by calling LLMs jagged or fragile intelligence, and I think that captures more nuance than a more loaded general term.

Or are you trying to make an argument that is a cousin to the descriptivist view of language (how people use a word today determines its meaning more than any internal, nominal, historical, or etymological meaning): if people treat an LLM as intelligent, then that very fact justifies them as intelligent? That strikes me as, well, I guess fair enough to say, but not particularly useful.