DaseindustriesLtd
late version of a small language model
Tell me about it.
User ID: 745
DaseindustriesLtd
late version of a small language model
2d ago
There had just been massive protests, which the Iranian regime has drowned in blood. I am suspicious of exact numbers, but it sounds more that people were beaten into temporary submission with overwhelming violence than that they simply gave up. Uprising is plausible I think.
Have you done research? It's not even about the tedium of grant-writing or whatever, people are doing a tremendous amount of routine work in data analysis and literature search, and scientific standards for programming are very low, "Ph.D code" is a meme. Even if AI doesn't contribute to the process of "discovery" per se (such as reasoning about hypotheses and planning experiments – though it definitely can plan experiments at this stage), it can trivially take over 90% of pure cognitive work-hours.
I recommend trying out some of the hot new models, with high reasoning settings. Opus 4.6, ChatGPT 5.2, Gemini 3.0 pro and so on, or even DeepSeek-Speciale. They're starting to make progress on really hard research-level physical reasoning tasks even without human guidance, and in a structured environment they are a great help to researchers.
P.S. Just an example of people's opinions, one of hundreds.
DaseindustriesLtd
late version of a small language model
19d ago
What is the point of this obstinacy? They're all correct, it is an act of war. We constantly commit acts of war by bombing the shit out of Ukrainian residential blocks, and this gets called terrorism because civilians become collateral damage, even if terror is not the point (terror is the point in human safari and arguably in infrastructure destruction though). If Russia could surgically annihilate Ukrainian generals no matter where they are, that's be merely war.
Ukrainians do commit terrorism, but not in this case.
DaseindustriesLtd
late version of a small language model
20d ago
Is it just me, or do the Olympics feel like they are far less culturally relevant than they used to be?
At least seems to be true in Korea:
The Milan Cortina Winter Olympic Games, which officially opened with preliminary events on Wednesday (local time) before the formal opening ceremony on Friday, have drawn the lowest level of Korean public interest ever recorded for a Winter Olympics. Google Trends data shows domestic searches for "Olympics" have fallen below 10 on a 100-point scale — down from 30 during the 2022 Beijing Games and a peak of 100 when Korea hosted the PyeongChang Games in 2018.
The disengagement is not uniquely Korean. U.S. broadcaster NBC's prime-time viewership for the Beijing Olympics averaged just 11.4 million — a 42 percent drop from PyeongChang. Ticket sales for Milan Cortina reached about 75 percent of capacity by early February, with nearly 1.2 million of roughly 1.5 million tickets sold, though organizers had relied on late surges and NHL star power to close the gap after a sluggish start that saw only 613,000 tickets sold through October.
"There was a time when families sat together in the living room to watch, even during hard economic times. That era is over," Yu said. "Everyone now consumes whatever content they want on their own smartphones."
Yu added that Koreans' emotional investment in national representation has weakened. "People are less inclined to feel that someone else's achievement on the international stage is somehow their own. Korea already has so much cultural content representing the nation globally that the Olympics no longer hold that singular status."
Ilia Kostaoinov Belov
The ID (which identifies his Bulgarian citizenship) says BELOV Mr. Ilia Kostadinov. This straightforwardly means he's Ilia Belov and his father's given name was Kostadin. There seems to be a lot of these guys. Like, here is the youtube of a Bulgarian guy named like this, but it's ancient.
How he can also be Fatos Ali Dumana, is beyond my Slavic knowledge, I guess that's just his nickname on FB. «Fatos is an Albanian masculine given name, which means "daring", "brave" or "valiant"». (Bulgaria and Albania are separated by North Macedonia). The caption on the video means something like "hey ladies, congratulations". He's listening to this crap from a duo of rappers, Turkish and German (I guess also Turkish). The ladies, surprisingly enough, do congratulate him, they seem to be family (at least one is clearly some auntie). The account is low-activity and consists of typical slop you might expect of a young low-IQ Southern Slav with Global Black characteristics trying to show off clothes and shit, or perhaps really just a Gypsy, though neither of his names is Gypsy-coded.
Looking up "Ali Dumana" floods the search with this Ilia. It's a very unusual string of tokens. If I restrict the search to a period before this scandal, I only get nonsense like this (an independent sexual allegation in Dundee, no Dumanas), somehow.
Now theCourier publishes propaganda about our "Dumana":
Bulgarian dad says his life has been shattered since a video of a Dundee street confrontation went viral after being shared by right-wing figures including Elon Musk and Tommy Robinson.
A 12-year-old girl, who cannot be identified for legal reasons, has been charged in connection with the alleged possession of a knife and an axe.
Speaking at his home, with his wife Fetka Fatosh, 19, and eight-month-old son Kostadin beside him, Mr Dumana said the abuse has left him in fear of leaving the house.
Although he speaks good English, his interview with The Courier was carried out in his native tongue and then translated by our journalist, who speaks both languages fluently.
"Fetka" is Slavic, "Fatosh" is some dimunitive in Arabic/Turkic I guess?
I particularly like this detail:
Far-right activist Tommy Robinson, whose real name is Stephen Yaxley-Lennon. Image: DC Thomson
So we get the name of the irrelevant right-winger, but the Mr Dumana remains an enigma. Brits are quite provincial, this is not exactly Soviet but pretty crude. Did they do any actual investigation?
Anyway, he's a Bulgarian citizen named Ilia Belov, he's got this weird Islamic pseudonym, he looks quite brown (without throwing any shade – that entire region is brown, I can't pin him to a specific country, between Bulgaria/Albania/etc), so I guess the girls could have panicked/reacted racistly even if he is a peaceful "Bulgarian dad" (feels weird to identify someone aged 22 as primarily "dad") and has never hurt a fly.
Very low information situation.
DaseindustriesLtd
late version of a small language model
23d ago
Back in November, there was discussion about the imminent fall of Pokrovsk
This sort of nonsense is why I do not follow the war news. It's disgusting seeing people, especially foreigners, cheer on two Slavic teams slowly grinding each other into nothingness, hype up a minor breach (in reality "a group of Russians maybe spotted slightly ahead of their usual positions") or interpret troop movements in the rear as a sign of impending collapse.
My dad used to repeat that the strongest bet in WWII on a day-to-day base was «nothing changes». But WWII was quite dynamic compared to this. It's actually hard to take territory in this kind of a war. Most gains are ephemeral digital map-painting, but losses are very real, and yet very gradual and insufficient to undermine either side's long-term warfighting capacity. Of course there's no decisive defensive line or «logistics hub» the loss of which will doom Ukraine – they can retreat just a little, to a more thoroughly prepared set of fucking trenches, and continue eroding Russian troops with the usual drone-centric tactics. There won't be gallant armor brigades thundering over the steppe, armor burns easily these days. With steady Chinese support of Russian military industry (bought and paid for) and steady European life support for the entire Ukrainian state (presumably Russians will end up paying for that too), it can go on like this for many years more.
Any plausible upset can only come from those external forces – either China ramping up its engagement, actually selling military assets rather than just dual-use goods and some sneakily rebranded «civilian» lasers and such (at the cost of losing European markets and goodwill, won largely through Trump's buffoonery), or the EU/NATO committing forces, or providing Ukraine with F-35s or something to that effect, or maybe the US getting serious. Nobody seems interested, however.
DaseindustriesLtd
late version of a small language model
24d ago
but ontologically it matters whether the system has state that persists across interruption and is causally necessary for its future behavior. Its not an arbitrary boundary.
Why? It's a meaningful distinction on its own terms, but what does it have to do with experience, awareness or consciousness?
DaseindustriesLtd
late version of a small language model
24d ago
I don't understand why this is so hard to understand.
Because it's either a non sequitur or a completely bizarre theory of cognitive awareness.
LLMs, shouldn't be thought of as minds or cognitively aware "beings" or any other such "conceptions" because we know exactly, precisely, specifically what they are.
In other words, only things for which we do not have this exact, precise, specific understanding can be minds or cognitively aware beings? So cognitive awareness intrinsic to X is conditional on our ignorance of the nation of X? Or a mind is inherently not-knowable? Or what?
I repeat, what's your actual argument here? I gave you some options.
You see a house and say "That house is really a landscape for a family to build dreams. It's a compassion and bonding machine" Well, that's fine if it works for you, but what the house really is is a house
This condescension is not helping. You are apparently vastly overestimating the quality of your ontology and epistemology. I hope you realize how frankly childish it is, using my helpful examples. A house is a house rather than a landscape not because we can precisely define a house, but because we can precisely define both a house and a landscape – or at least train an LLM to investigate embedding similarity – and see how the definitions do not intersect, and so applying the token "house" to a "landscape" or vice versa is purely metaphorical speech. We have a definition of an LLM. Do you have a rigorous definition of a mind that excludes LLMs on principled grounds?
DaseindustriesLtd
late version of a small language model
26d ago
Why? Superdeterminism posits that everything is absolutely determined, including humans. Does this change the human condition? In general I have never seen the link between consciousness, agency, and randomness/unpredictability. I suppose seeing it requires a very practical, utilitarian mindset where a real agent changes the global state on its own, and if the state is evolving according solely to its intrinsic rules, there is no place for agency in the system and no need for the term. I believe that's too depressing a metaphysics. Agents are a class of events we observe inside the universe. They must be be definable, for they are observable, even if their substance is as immaterial as a sunbeam or a shadow. And AI agents just put our folk metaphysics to test.
DaseindustriesLtd
late version of a small language model
26d ago
That is not my definition, and I do not see how non-determinism is required at all.
DaseindustriesLtd
late version of a small language model
26d ago
LLMs don't have an internal state that I know of. If you have another article I'll read it, I do enjoy them.
https://arxiv.org/abs/2512.23675
https://research.google/blog/introducing-nested-learning-a-new-ml-paradigm-for-continual-learning/
Is merely making LLM weights dynamic at inference enough to challenge your model? KV cache is «external state» but weights must be internal I suppose, since LLMs have already been defined as weights above.
This is all an aesthetics-based argument with arbitrarily drawn categories. I don't see why we should care how particular matrices are stored and multiplied.
DaseindustriesLtd
late version of a small language model
26d ago
At what point does scaling up molecular dynamics result in agency? How many molecules does it take?
DaseindustriesLtd
late version of a small language model
26d ago
With AI models, you can describe behavior directly in terms of the underlying code
You can't. It's intractable. For example, one of the top 3 organizations pursuing AGI, the current leader in agentic coding, Anthropic, investigating misalignment:
New Anthropic Fellows research: How does misalignment scale with model intelligence and task complexity?
When advanced AI fails, will it do so by pursuing the wrong goals? Or will it fail unpredictably and incoherently—like a "hot mess?"
Finding 2: Scale improves coherence on easy tasks, not hard ones
How does incoherence change with model scale? The answer depends on task difficulty:
Easy tasks: Larger models become more coherent
Hard tasks: Larger models become more incoherent or remain unchanged
This suggests that scaling alone won't eliminate incoherence. As more capable models tackle harder problems, variance-dominated failures persist or worsen.
Why Should We Expect Incoherence? LLMs as Dynamical Systems
A key conceptual point: LLMs are dynamical systems, not optimizers. When a language model generates text or takes actions, it traces trajectories through a high-dimensional state space. It has to be trained to act as an optimizer, and trained to align with human intent. It's unclear which of these properties will be more robust as we scale.
Constraining a generic dynamical system to act as a coherent optimizer is extremely difficult. Often the number of constraints required for monotonic progress toward a goal grows exponentially with the dimensionality of the state space. We shouldn't expect AI to act as coherent optimizers without considerable effort, and this difficulty doesn't automatically decrease with scale.
That's, like, the frontier of interpretability research.
Does this look like looking at the code and saying «Ah I get it, X does A»?
We're in a very similar epistemic position with regard to a tiger and to an LLM. The big difference is that with a tiger we have some very limited observation methods like electrocorticography or tomography or something, and with an LLM we can – in theory – deconstruct any particular causal sequence, every activation, every decoded token. But it won't become comprehensible to humans just because we produce another vast array of zeroes and ones from logging its activity.
They are parameterized conditional probability functions, that are finite-order Markovian models over token sequences. Nothing exists outside their context window. They don't persist across interactions, there is no endogenous memory, and no self-updating parameters during inference
Just a string of non sequiturs.
DaseindustriesLtd
late version of a small language model
26d ago
That's literally, exactly, precisely what they are.
So what?
@self_made_human proceeds to generate a lot of prose, but all he really needed to do was press for some substantiation of this argument. «Weights» is a word. What LLMs really are is information. Why exactly is this specific mode of information incompatible with having high-level properties like «personality flaws»? You accuse him of incoherence in the inane tiger side debate, but «models are weights, ergo anthropomorphized traits don't apply except as a loose metaphor» is basically schizophrenic in my book. What's the actual claim here? That anthropomorphic properties are substrate-dependent, that functionalism is wrong? Just say so instead of snarking and appealing to incredulity. Ideally with some defense for this opinion.
I'm afraid this jingoistic intoxication will get worse when the US proceeds to topple some other incompetent country (probably Cuba next). This is all fine and good so long as the actual decisionmakers are sober and don't infer they can start anything with China, but will they be sober? Currently you're burning through interceptors and, if this is not AI fog of war slop, even losing F-15s (alledegely to friendly fire, I presume due to lack of relevant training. Should have called upon Ukrainians to teach you guys air defense). But long term, Iran is poised to lose the war, of course, so the sense of invulnerability will be restored.
First of all Iran was not a "military ally" to China in a way that matters, this is just cope to inflate the sense of achievement, just like hyping up Iranian "gigantic military built over 40 years" (I see you double down on it). For reference, India, Kazakhstan and Pakistan are also there, are they Iranian allies too? Are India and Pakistan allies? They've just had a war. "Heavily involved in selling oil" just means that due to sanctions their oil was selling at a discount, which the Chinese opportunistically exploited. Here's what Foreign Policy had to say last September about the nature of the relationship:
That's diplomatic; on a personal level, Chinese consider Iranians worthless backstabbing third worlders. So, their lack of direct involvement is quite understandable. The article lists some symbolic gestures (Beijing inviting Iranians to Victory day parade, agreement on implementation of the 25 year investment pact) but that was transparently a panicked reaction to a crisis. Objectively they're given about the same treatment as Starmer, Macron, Carney and other foreign dignitaries. I don't want to say there's nothing to multipolar agenda, obviously China prefers Iran to remain a thorn in the US/Israeli side and also to buy cheaper oil. But that's a benefit of bounded and not great value, and ineptitude and duplicity of the mullah regime qualifies it further.
No, Iran is only about Iran and Israel, not China. Except psychologically (I'll return to this).
I think what we're learning is not so much that the US is a supreme military power but that it's been a very reluctant hegemonic empire indeed. Why do these shitholes even exist? Venezuela, Cuba? Seriously? The former is a pure petrostate that had failed to keep its oil rigs running due to decades of mismanagement and populism (and also sanctions). The latter is a country famous for sugarcane that's importing sugar now because Communists have ran the industry into the ground. Just months ago, Iran had almost collapsed due to a drought, not to mention that it's deeply infiltrated by Mossad. Why does the US tolerate such enemies instead of giving them a push? Why does it just allow the hostility to persist? Well, Trump has been asking himself just this, it seems. The answer is, there's no good reason. The US can afford to crush them, because it'll be pretty cheap (especially given the fixed costs of US military power).
The problem comes with assuming that China is anything like them. I get it, too – Communists, enemies of America, poor, theoretically allies (though China has no real allies except for North Korea and informally Pakistan, to counter India). But it's dangerously delusional. At the end of the day, the reason America can do this, the reason it has all those stealth jets and satellites and AI and smart munitions and everything else is that it has a large, productive, complex, technologically advanced economy. Even the industrial sector, for all the talk of hollowing out, is the world's second largest (though it depends on how you treat value-added figures – in physical output, it might be closer to Japan than to China). And these guys are so far down the line they barely have an economy.
I like the measure called Economic Complexity Index, maintained by Harvard Growth Lab. In intention, it tracks how capable a nation is of mainitaining industries that generate globally competitive products, though in reality is just measures export diversity. It's not pefrect – for example, Australia and the US get punished by the predominance of a few commodities in their export basket – but it's a decent proxy if you keep that in mind. Say, in 2024, Venezuela was ranked #133 out of 145 countrires. Cuba is #122. Afghanistan is #110. Iran is #87. Russia is #67. Canada (commodity exporter) is #35. USA is #20, between Hong Kong and France. The top 10 all have negligible commodity endowment. The list is as follows: Japan, Switzerland, Taiwan, South Korea, Singapore, Germany, Czechia, Israel, Slovenia, China (climbed 7 points in the last 16 years; the US has fallen down by 7 again). Discounting the fraction of the economy involved in fossil fuels and agriculture (a generous choice), I'd say the US would end up roughly as complex as China. They have a nice visualisation, you can click through it, eg here's the structure of American imports from China, and here are exports. Or, here are Chinese exports to Japan which as we know is number 1. And the other way around. It's quite clear to me that the ranking is directionally aligned with reality. And it's a ranking of complexity. In terms of volume or human capital employed in militarily relevant fields, it's not close. China doesn't need "allies" because it surpasses the entire Western bloc in scale.
Reminder that China is already testing two 6th generation jets in the open, and given that you haven't resolved the issue of launching 5th generation from Ford in over a decade (their EMALS works flawlessly btw), there are hardly any grounds to expect the gap to widen (or even to exist).
How's that working out? I see Iranian ballistic missiles hit Israel online. Reminder that it's a barely functional theocracy, these aren't some fancy MIRVs or HGVs. Forget THAAD, actually, you're struggling wtih Shaheds.
Starship is a real argument for interception, but the gap in rocket technology is likely smaller than it seems. We'll see over 2026 if China can begin landing their boosters, and they won't need Starship unit economics to scale up production.
Currently estimated at 600 warheads, vs American stockpile of 3700. It's a completely sufficient deterrence. You glibly dismiss 50-90 million dead Americans, I suspect that's a lowball but the point is that you're unlikely to destroy China either, for all the memes about Three Gorges Dam. Their cities are denser but made of concrete far more resilient to nuclear flash than your suburbs, for starters. That said, we're all far from the genocidal peak of Cold War, and these assets on both sides would be used on counterforce strike.
What I want to say is that this isn't just a funny hypothetical. "How do we fight China" is the question on the mind of American planners, and the answer is "we don't, not really". China is your only rival and pacing threat, China is likely to take Taiwan in years, and there are no adequate answers sans praying to AGI and Elon Musk to bail the US out. Accordingly this showboating in hostile shitholes, while inflating their alleged capability to proportionally inflate American dominance, to the extent that it's not executing on prior plans and commitments – is best understood as procrastination in the face of unsolvable strategic dilemma, with a nice bonus of inciting this feverish national pride and maybe improving the GOP's chances in the midterms.
More options
Context Copy link