BANNED USER: temperamental dysregulation
DaseindustriesLtd
late version of a small language model
Tell me about it.
User ID: 745
Banned by: @Amadan
DaseindustriesLtd
late version of a small language model
1yr ago
I really appreciate that Americans have been so fed up with a demented president that they elected a retarded one instead. It's clear that Trump is simply illiterate – and perhaps literally so, do we know for sure that he can read English, has anyone seen him read a book or something to that effect? At the very least his takes aren't better than what I'd expect from an average voter with a stubborn child's notion of economy. Give me money, take my goods, very short words, very petty grievances.
The awesome thing about it is that with a stunted child in total control, you can truly just do things. He could threaten allies, the way he's doing it with Europe: buy our oil (more than we produce… pay in advance, 10 years on year, or whatever!) or we drop out of NATO and leave you to deal with Putin. He could threaten with direct military aggression, like the Greenland case. He'll feel very clever and smug in doing so, and his base will also gloat about his genius.
He may well succeed. American negotiating position (as the world's greatest military power) is strong. We may see democratic nations transform into communist dictatorships liquidating their pension funds to fund American war factories, or taking IMF loans to buy Teslas. There is no ceiling to winning here.
DaseindustriesLtd
late version of a small language model
1yr ago
Looming? What do you mean concretely?
There's reportedly been 9.5 million babies born in China last year. Something like 3.6 million in the US; of those, 1.8 million white babies. Accordingly, in 18 years there will be 5 times as many Chinese 18 year olds as there will be their White American counterparts. Yes, they have lower TFR, but at current trends they'll have a vastly larger workforce for many decades.
DaseindustriesLtd
late version of a small language model
1yr ago
The likelihood of winning a conflict has little relevance to whether that conflict should be waged in the first place.
It actually has a lot of relevance. The real reason you act like it doesn't is that you do not seriously engage with the possibility of losing, and losing badly (losing what? To what degree? How many cards do you have left at the point of losing, and what terms can be negotiated?). People make unreasonable maximalist demands when they are assured of their invulnerability. You treat a great power conflict like another Middle Eastern adventure, “oh we found WMDs in this shithole, our Democracy will perish if we do not conquer it hue hue!”. It's an instinct that's hard to overcome after a century of uninterrupted wins and cost-free losses. The same Main Character Syndrome, coupled with low human capital in Trump team, explains decidedly suboptimal and cost-insensitive means that were chosen for prosecuting the conflict. Americans think they can afford anything, because that's recorded in their institutional DNA. But they have never fought a superior power, due to it never having existed prior to this day. So they have developed an auxiliary belief that the very fact of them antagonizing any power confirms it is inferior. It's hard to feel pity for such a narcissistic people.
it is the serf who acts in accordance with prudence and rationality. The serf is a serf precisely because he correctly calculates that servitude is what gives him the best odds of continued survival. The nobleman, in contrast, acts in accordance with virtue, even when the outcome is certain destruction.
In Imperial Russia, there was a trend when mujiks, LARPing as nobles, initiated duels over petty spats, murdering each other with axes; eventually the state had to put its boot down. Due to extremely low literacy rates they couldn't have plausibly cited Nietzsche when doing so, but I believe that they'd have appreciated your quote.
Self-serving, petulant, handwavy, shallowly aesthetic notions of virtue are cheap and easy to brandish in defense of one's animalistic impulses; any kind of impulsive retardation can be dressed up as a calling of aristocratic, virile masculine nature, there's a whole genre of extremely popular Western music about it, authored by the impromptu warrior aristocracy of the streets. Your own elite has been wiped out to such a degree that this whole discourse is vacuous, we can't consult with a living bearer of a tradition, only speculate. It is plausible that I am wrong and there's just never been any substance to the whole fraud.
DaseindustriesLtd
late version of a small language model
1yr ago
No, you're correct. In fact, had the US continued the course it had just half a year ago, I'd still be largely holding this opinion. But the election of an illiterate boomer strongman does change matters. Xi also has managed to not do anything too self-defeating long enough. I admit I was wrong: the US does not have a functional elite to make appropriate use of its genuine (if transient) political, economic and technological advantages and keep the Chinaman down.
ASI may restore my faith in the previous model, but this is looking like a remote possibility.
DaseindustriesLtd
late version of a small language model
1yr ago
After all, if the American GDP is in some way fake how come the median American can buy so much Chinese production with his or her dollars?
Largely because China (like everyone else) is buying your assets and the USD is the global reserve currency.
Trump is doing what he can to fix this pathological situation, by being laser focused on goods.
DaseindustriesLtd
late version of a small language model
1yr ago
But he actually argues that the Chinese services sector size is underestimated, by Chinese accounting mainly.
We believe China’s GDP and PPP GDP are lowballed by an incomplete transition from the Material Product System (MPS) of national accounts, which excludes services by design. The World Bank is likely dutifully doing its sums with goods consumption in China multiples of the US but measuring services consumption as a fraction of the US.
The United Nations System of National Accounts (UNSNA) provides voluntary guidelines and specifically states that nations should base their national accounts on local conditions. What that has meant in the West is to adopt all UNSNA “innovations” introduced over the years.
Items like imputed rent, legal fees and R&D are now all included in GDP. The UK went hog wild with both illegal drugs and prostitution as now part of their GDP because… hey, why not? UNSNA’s 2008 guidelines explicitly recommend that illegal market activity should be included in GDP.
China’s NBS stood its ground on a conceptual level. Rightly or wrongly, the Leninist MPS considers services necessary costs of material production rather than real value creation. In China’s first attempt at converting MPS to SNA in 1985, it tacked on a ludicrously low 13% to the MPS number and called it China’s services GDP.
Over the years, the World Bank has twisted the arm of the NBS for modest increases to China’s services GDP – with limited success.
The affordability crisis in Western economies, the US in particular, is largely driven by inflation of necessary services – rent, healthcare, education and childcare – not by manufactured goods. While these costs have also gone up in China, they have increased less and much are left out of GDP anyway.
Also not captured by the ICP survey conducted in 2021 are the price and service wars that have broken out across industries and products – a bane on businesses but a boon for consumers.
This is most visible in China’s car market with OEMs either cutting prices to the bone (Hyundai Sonatas down to $17,000 from $42,000) or offering cutting-edge technology for peanuts (a 2,000-kilometer range BYD Q plug-in hybrid electric vehicle for $14,000). The price of solar panels fell 50% in 2023 and continues to trend down in 2024. CATL has announced plans to cut lithium-ion battery prices in half by the end of 2024.
Restaurants are offering white glove perks like hot towels, lotion by the sink and snazzy remodeled decors. Hairdressers hand out bottled water and fruit plates. Tech companies have slashed large language model (LLM) prices to basically free. Service quality in China, impossible to quantify, is now head and shoulders above the West and probably even Japan.
Adherence to UNSNA has caused a breakdown in the meaning of GDP. As necessary services become an ever larger share of Western economies, their growth does not appear to result in discernable improvements in living standards.
It's not a workshop country. The thesis isn't that they have a strong industry - everyone knows that.
DaseindustriesLtd
late version of a small language model
1yr ago
This would actually make more sense yes. Allegedly, Trump’s Love for Tariffs Began in Japan’s ’80s Boom:
WASHINGTON — Donald J. Trump lost an auction in 1988 for a 58-key piano used in the classic film “Casablanca” to a Japanese trading company representing a collector. While he brushed off being outbid, it was a firsthand reminder of Japan’s growing wealth, and the following year, Mr. Trump went on television to call for a 15 percent to 20 percent tax on imports from Japan.
“I believe very strongly in tariffs,” Mr. Trump, at the time a Manhattan real estate developer with fledgling political instincts, told the journalist Diane Sawyer, before criticizing Japan, West Germany, Saudi Arabia and South Korea for their trade practices. “America is being ripped off,” he said. “We’re a debtor nation, and we have to tax, we have to tariff, we have to protect this country.”
Thirty years later, few issues have defined Mr. Trump’s presidency more than his love for tariffs — and on few issues has he been more unswerving. Allies and historians say that love is rooted in Mr. Trump’s experience as a businessman in the 1980s with the people and money of Japan, then perceived as a mortal threat to America’s economic pre-eminence.
That's from 2019. If China is just a stand-in for Japan (which he also tariffed anyway), it is no wonder that he acts so brazenly. You can bully Japan with no repercussions. He can finally get back at them for that piano humiliation.
DaseindustriesLtd
late version of a small language model
1yr ago
I can only say that engaging with the Chinese, and with people like you, has gradually convinced me that White People (Hajnali European stock specifically) are basically jumped-up serfs, the confused lower caste of prawns from District 9, with little more to offer to the world sans stale kanging and hollow, corporate-coded pretense of “soul” that, if it ever existed, resided in your currently extinct owners. You don't even notice my point about simple economics and logistics, so lost you are in your racial superiority masturbation. But of course those issues are related.
if Japan were in China's position instead
But it isn't, and you are largely responsible for that, because your previous generation had the exact same attitude towards the Japanese. Deaths from overwork, rigid hierarchy, soulless collectivist automatons cheating and copying to flood the markets and dispossess our Christian Germanic workers – this can't be allowed, can it? Oh, what a pity that now that we know them better, Japan is a geriatric country of no ambition, that mainly produces anime to give you some respite from the toxic antihuman sludge of your own media. (Presumably this is the fault of Joos. Somehow for all your natural nobility of spirit you are not capable of resisting a tiny tribe of natural wordcels. At least the Chinese managed to overthrow the Manchu).
Regrettably, China is 10 times larger and the same tricks won't work.
A change in American economic policy sent global markets into a tailspin, so objectively speaking, America is in fact a big deal.
Yes, you can do a great deal of damage to humanity. This is akin to the bafflingly swinish line of argument that “China needs us more than we need them, because they need to sell their valuable manufactured goods to someone; our consumption is more valuable than production”. We shall see how well this philosophy works.
DaseindustriesLtd
late version of a small language model
1yr ago
My impression after obsessively monitoring this situation for days (of course) is that neither side will fold, tariffs are here to stay, and everyone will be poor and mad for it. China of course won't fold, the idea that they're at risk is preposterous, they can well weather complete cessation of export to the US.
Broadly I have concluded that the main problem the US faces is racism towards the Chinese; the ill-earned sense of centrality and irreplaceability. I believe that Trump, Navarro and the rest of that gang are as misinformed as the average MAGA guy on Twitter, given how they speak and that amusing formula. Americans still think that their great consumption is the linchpin of Chinese economy, 10-30% of their GDP (it's more like 3%); that the Chinese produce apparel, “trinkets” and low-quality materials (they also produce things that Americans plausibly cannot start producing at the same quality in years); that American IP is vital for their industry (they're making their own software, OSes, CPUs…) and so on. The idea that American de-industrialization is a product of betrayal by Wall Street Elites who offshored jobs to China also feeds into the delusional notion of possible parity – but the truth is that there has never been a point in history where American industry had scale or diversity comparable to what's going on in China now. The issue with their bad financials is also overblown; as for losing markets, they have the capital at hand for consumption stimulus. This guy from Beijing writes:
China’s PPP GDP is only 25% larger than that of the US? Come on people… who are we kidding? Last year, China generated twice as much electricity as the US, produced 12.6 times as much steel and 22 times as much cement. China’s shipyards accounted for over 50% of the world’s output while US production was negligible. In 2023, China produced 30.2 million vehicles, almost three times more than the 10.6 million made in the US.
On the demand side, 26 million vehicles were sold in China last year, 68% more than the 15.5 million sold in the US. Chinese consumers bought 434 million smartphones, three times the 144 million sold in the US. As a country, China consumes twice as much meat and eight times as much seafood as the US. Chinese shoppers spent twice as much on luxury goods as American shoppers.
…It is prima facie ridiculous that China’s production and consumption, at multiples of US levels, can be realistically discounted for lower quality/features to arrive at a mere 125% of US PPP GDP. … Similarly, analysts who lament that China accounts for 30% of the world’s manufacturing output but only 13% of household consumption are far off the mark. China accounts for 20-40% of global demand for just about every consumer product but much of the services it consumes have been left out of national accounts.
Accordingly, with a higher real GDP, their effective debt to GDP ratio may be as low as 150%, not 200-300%. They have US assets to sell too.
So China can trivially absorb half of the overcapacity freed by reduced trade with the US, and might find buyers for the rest.
My thesis is that in picking this fight, Americans don't understand that they're actually not that big of a deal. Unfortunately, their delusions are globally shared and become reality in their own right. But perhaps not enough to offset the gross physical one.
The actual dangerous thing for China here is that Trump seems determined to immiserate the whole planet, completely irrespective of any geopolitical rivalry, because he's an illiterate anarcho-primitivist and thinks that all trade is theft unless it's barter, basically. America vs. The World, especially with a chain reaction of tariffs on Chinese (and likely also Vietnamese etc…) capacity spillover, results in massive reduction of productivity for everyone. For now, nations like Vietnam are unilaterally dropping tariffs on American crap, but that can't be a big effect because their tariffs were low to begin with, and Americans just don't and cannot produce enough at price points that people of those nations can afford. (We may see IMF loans for 3rd world countries importing overpriced American beef or Teslas or whatever to placate Don, but I doubt it'll be sustainable). I suppose in the long run the idea is that Optimus bots will be churning out products with superhuman efficiency, at least Lutnick argues as much. But that's still years away. Perhaps this extortion of zero balance trade (so in effect, the demand that trading partners buy non-competitive American products) is meant to finance the transition to posthuman automated economy. Bold strategy.
I am of course very amused and curious to see how it'll go. Will Fortress America intimidate the rest of us into submission, likely forever? Or will it be so stubborn, brutal and ham-fisted that humanity will finally rebel and ostracize the rogue state, letting it broil in its own supremacist fantasies? Can Bessent et al. turn 1D “trade le bad” checkers of the King of Understanding Things (懂王) into 4D chess? We shall see.
DaseindustriesLtd
late version of a small language model
1yr ago
All I can say based on my sources (which we probably share) is that Hanania, if anything, understates his case, fearful of NDAs, and so focuses on blatant but not very consequential demonstrations of ineptitude. The brain rot in this administration is unbelievable, it's on par with the goofy early Communist governments I've marveled at in historical materials. People here who eloquently steelman them have their work cut out for them. They'll easily exceed the wokes in concocting justifications for pure primitive malice and barbarity in the coming years.
DaseindustriesLtd
late version of a small language model
1yr ago
It's pretty clear from his chart the tariff above 10% is a rate equal to half of the other nation's tariff on US goods.
are you seriously missing the point? How does this brazen lie pass the most basic sanity check? Do you think Vietnam tariffs the US to the tune of 92%?
DaseindustriesLtd
late version of a small language model
1yr ago
If you're really an SWE, I must presume that you're not speaking in good faith here.
Asking it for a gear setup for a specific boss results in horrible results, despite the fact that it could just have copied the literally wiki (which has some faults like overdoing min-maxing, but it's generally coherent). The net utility of this answer was negative given the incorrect answer, the time it took for me to read it, and the cost of generating it (which is quite high, I wonder what happens when these companies want to make money).
You must know that GPT 4.5 is pretty mid as far as instruction models of this generation go. DeepSeek's latest is close in performance and literally 100-200x cheaper. More importantly, what do you think would be a random college-educated human's score on Runescape questions? It is so trivial to grant these systems access to tools for web browsing as to not be worth talking about.
The rest of your comment is the same style. What is amazing and terrifying about LLMs is not their knowledge retrieval but generality and in-context learning. At sufficient context length and trained to appropriately leverage existing tools, there is nothing in the realm of pure cognitive work they cannot do on human level. This is not hard to understand. So tell me: what are you going for? Just trying to assuage your own worries?
DaseindustriesLtd
late version of a small language model
1yr ago
From what I can tell it is a giant improvement over V3 and a substantial one over R2. People from 4chan are saying it's suspiciously good at following their outlines and even uses a similar format, so they guess it's been trained on their writing (DeepSeek openly trains on user data). Generally it's in the same league as Sonnets. I recommend giving it a try.
DaseindustriesLtd
late version of a small language model
1yr ago
Consider that the gap between “me and a woman I have had sex with” may not generalize to World Cup men vs. women, logically speaking. Yes, it actually does (assuming me and the woman are average) but we know that [male] athletes in strength sports are basically superhumans, they are incomparably stronger and more fit than the general population. If you do not pay any attention to sports and aren't very observant, it's easy to believe that there exist outliers, massive 250 lb Amasons with six-packs and 20 inch biceps, somewhere out there. They exist in fiction, after all…
The biggest reason that the strongest women are much weaker than the strongest men is not the average gap. It's that they have a much lower ceiling.
DaseindustriesLtd
late version of a small language model
1yr ago
You are noticing that none of these companies want to race. The whole competition to build Sand God is largely kayfabe. Western AI scene is not really a market, it's a highly inefficient cartel (with massive state connections too), which builds up enormous capacity but drags its feet on products because none of them ultimately believe their business models are sustainable in the case of rapid commoditization. This is why DeepSeek was such a disruption: not only was it absurdly cheap (current estimates put their annual operations cost at like $200M), not only were they Chinese, but they dared to actively work to bring the costs of frontier capabilities to zero, make it logistically mundane, in alignment with Liang Wenfeng's personal aesthetic and nationalist preferences.
I think R1's release has sped up every Western frontier lab by 20-50% simply by denying them this warm feeling that they can feed the user base some slop about hidden wonder weapons in their basements, release incremental updates bit by bit, and focus on sales. Now we are beginning to see a bit more of their actual (still disappointingly low, not a single one of these companies could have plausibly made R1 on that cluster I think) power level.
DaseindustriesLtd
late version of a small language model
1yr ago
Again I have to quote Boaz Barak (currently OpenAI): AI will change the world, but won’t take it over by playing “3-dimensional chess”.
Consider the task of predicting the consequences of a particular action in the future. In any sufficiently complex real-life scenario, the further away we attempt to predict, the more there is inherent uncertainty. For example, we can use advanced methods to predict the weather over a short time frame, but the further away the prediction, the more the system “regresses to the mean”, and the less advantage that highly complex models have over simpler ones (see Figure 4). As in meteorology, this story seems to play out similarly in macroeconomic forecasting. In general, we expect prediction success to behave like Figure 1 below—the error increases with the horizon until it plateaus to a baseline level of some simple heuristic(s). Hence while initially highly sophisticated models can beat simpler ones by a wide margin, this advantage eventually diminishes with the time horizon.
Tetlock’s first commandment to potential superforecasters is to triage: “Don’t waste time either on “clocklike” questions (where simple rules of thumb can get you close to the right answer) or on impenetrable “cloud-like” questions (where even fancy statistical models can’t beat the dart-throwing chimp). Concentrate on questions in the Goldilocks zone of difficulty, where effort pays off the most.” Another way to say it is that outside of the Goldilocks zone, more effort or cognitive power does not give much returns.
Rather, based on what we know, it is likely that AI systems will have a “sweet spot” of a not-too-long horizon in which they can provide significant benefits. For strategic and long-term decisions that are far beyond this sweet spot, the superior information processing skills of AIs will give diminishing returns. (Although AIs will likely supply valuable input and analysis to the decision makers.). An AI engineer may well dominate a human engineer (or at least one that is not aided by AI tools), but an AI CEO’s advantage will be much more muted, if any, over its human counterpart. Like our world, such a world will still involve much conflict and competition, with all sides aided by advanced technology, but without one system that dominates all others.
In essence, irreducible error and chaotic events blunt the edge of any superintelligent predictor in a sufficiently high-dimensional environment.
What remains to be answered for me:
- Can AI planners interfere in the events with enough frequency and precision to proactively suppress chaos and reduce the world to a game of chess they can model to the draw?
- Is a decently superhuman prediction and execution not enough to eliminate warm, simply because humans are already close to this level and only initiate wars they won't win (instead of pragmatically retreating to some defensible compromise) in feats of retardation (see: Russia)?
DaseindustriesLtd
late version of a small language model
1yr ago
I think a case can be made that whenever the Chinese state is stable, it is very stable and very good at suppressing peasant rebellions. This is perhaps grounds for an unflattering stereotype about cruelty and power distance, but not so much about obedience of the masses.
DaseindustriesLtd
late version of a small language model
1yr ago
that I argue that you do not understand how well connected people like they are or are not to the dominant American (or western) political cultures.
So, how well connected are they? Enlighten me. Being a clueless Imperialist (or however you see me), I have developed the impression that Peter Thiel and his creatures, and Palantir specifically, are fairly well connected in the current American establishment.
DaseindustriesLtd
late version of a small language model
1yr ago
Functional American hegemony, whether as a means or as an end, has clearly lasted for decades. Do you simply concede that it's not going to survive and the US accepts that since its continuation is not worth or not feasible fighting for?
Do you argue that people like Palmer Luckey, Alex Karp, Alex Wang, Dario Amodei, Sam Altman are, similarly to me, clueless and in disconnect with your political culture? Because they definitely argue for the maintenance and indeed revitalization of hegemony, not some strategic retreat to domestic affairs. Says Amodei:
This means that in 2026-2027 we could end up in one of two starkly different worlds. In the US, multiple companies will definitely have the required millions of chips (at the cost of tens of billions of dollars). The question is whether China will also be able to get millions of chips9.
If they can, we'll live in a bipolar world, where both the US and China have powerful AI models that will cause extremely rapid advances in science and technology — what I've called "countries of geniuses in a datacenter". A bipolar world would not necessarily be balanced indefinitely. Even if the US and China were at parity in AI systems, it seems likely that China could direct more talent, capital, and focus to military applications of the technology. Combined with its large industrial base and military-strategic advantages, this could help China take a commanding lead on the global stage, not just for AI but for everything.
If China can't get millions of chips, we'll (at least temporarily) live in a unipolar world, where only the US and its allies have these models. It's unclear whether the unipolar world will last, but there's at least the possibility that, because AI systems can eventually help make even smarter AI systems, a temporary lead could be parlayed into a durable advantage10. Thus, in this world, the US and its allies might take a commanding and long-lasting lead on the global stage.
Well-enforced export controls11 are the only thing that can prevent China from getting millions of chips, and are therefore the most important determinant of whether we end up in a unipolar or bipolar world.
As you can see he deems bipolar outcome unacceptable, since it's merely a prelude to American (and all Western/liberal) defeat: either the US wins a “durable strategic advantage” by capitalizing on its compute edge, or China does by capitalizing on its industrial capacity. For my part I think he's wrong and dumb, the US is highly defensible and not at risk of Chinese unipolar dominance. But that is his argument, and others are making near-identical ones.
From CSIS, I don't know, maybe you hold them too in contempt, but they use the same terminology:
China’s success to date suggests that, at least for Huawei Ascend chips, the answer is that they will have millions of chips within the next year or two. Thankfully, these chips are, at present, dramatically lower performing than Nvidia ones for training advanced AI models; they are also supported by a much weaker software ecosystem with many complex issues that will likely take years to sort out. This is the time that the export controls have bought for the United States to win the race to AGI and then use that victory to try and build more durable strategic advantages. At this point, all the margin for sloppy implementation of export controls or tolerance of large-scale chip smuggling has already been consumed. There is no more time to waste.
All of this does not look to me like acceptance of coming multipolarity.
Do you write it off as inconsequential self-interest of individual players, because the vote of salt-of-the-earth rednecks is more influenced by price of eggs?
DaseindustriesLtd
late version of a small language model
1yr ago
I think this is still too self-serving and frankly racist a spin: “Chinese are robots, sure, but they can train robots on Western genius and follow their lead, still copying the West by proxy”.
I try to approach this technically. Technically you say that Asians are incapable of thought in full generality, that – speaking broadly – they can only “execute” but not come up with effective out-of-the-box plans; that their very robust IQ edge (Zhejiang, where DeepSeek is headquartered and primarily recruits, and where Wenfeng comes from, has average around 110 - that's on par with Ashkenazim!) – is achieved with some kind of interpolation and memorization, but not universal reasoning faculties. To me this looks like a very shaky proposition. From my/r1's deleted post:
The West’s mythos of creative genius – from Archimedes to Musk – emerged from unnaturally prolonged frontiers. When Europe lost 30-60% of its population during the Black Death, it reset Malthusian traps and created vacant niches for exploration. The American frontier, declared "closed" by the 1890 Census, institutionalized risk-taking as cultural capital. By contrast, China’s Yangtze Delta approached carrying capacity by the Song Dynasty (960-1279 CE). Innovation became incremental: water mills optimized, tax registers refined, but no steam engines emerged.
This wasn’t a failure of intelligence, but a rational allocation of cognitive resources. High population density selects for "intensive" IQ – pattern-matching within constraints – rather than "extensive" creativity. The same rice paddies that demanded obsessive irrigation schedules cultivated the hyper-adaptive minds now dominating international math Olympiads. China’s historical lack of Nobel laureates in science (prior to 1950) reflects not a missing "genius gene," but a Nash equilibrium where radical exploration offered negative expected value.
R1 might understate the case for deep roots of Western exploratory mindset, but where we agree is that its expression is contingent. Consider: how innovative is Europe today? It sure innovates in ways of shooting itself in the foot with bureaucracy, I suppose. Very stereotypically Chinese, I might say.
What I argue is that whereas IQ is a fundamental trait we can trace to neural architecture, and so are risk-avoidance or conformism, which we can observe even in murine models, “innovativeness” is not. It's an application of IQ to problem-solving in new domains. There's not enough space in the genome to specify problem-solving skills only for domains known in the bearer's lifetime, because domains change; Asians are as good in CTF challenges as their grandfathers were in carving on wood. What can be specified is lower tolerance to deviating from the groupthink, for example as cortisol release once you notice that your idea has not been validated by a higher-status peer; or higher expectation of being backstabbed in a vulnerable situation if you expend resources on exploration; or greater subjective sense of reward for minimizing predictive error, incentivizing optimization at the expense of learning the edges of the domain, thinking how it extends, testing hypotheses and hoping to benefit from finding a new path. Modulo well-applied and tempered IQ, this eagerness to explore OOD is just a result of different hyperparameter values that can also produce maladaptive forms like useless hobbies, the plethora of Western sexual kinks (furries?) and the – no, no, it's not just Jewish influence, own up to it – self-destructive leftist ideologies.
One anecdote is illustrative of the conundrum, I think. Some time ago, a ByteDance intern came up with a very interesting image generation technique, VAR. It eventually won the NeurIPS best paper award! Yandex trained a model based on it already, by the way, and Yandex has good taste (I may be biased of course). But what else did that intern do? Driven by ambition to scale his invention and make an even bigger name for himself, he sabotaged training runs of his colleagues to appropriate the idle compute, fully applying his creative intelligence to derail an entire giant corporation's R&D program! Look at this cyberpunk shit:
- Modifying PyTorch Source Code: Keyu Tian modified the PyTorch source code in the cluster environment used by his colleagues, including changes to random seeds, optimizer's direction, and data loading procedures. These modifications were made within Docker containers, which is not tracked by Git.
- Disrupting Training Processes: Keyu Tian deliberately hacked the clusters to terminate multi-machine experiment processes, causing large-scale experiments (e.g., experiments on over thousands of GPUs) to stall or fail.
- Security Attack: Tian gained unauthorized access to the system by creating login backdoors through checkpoints, allowing him to launch automated attacks that interrupted processes of colleagues' training jobs.
- Interference with Debugging: Tian participated in the cluster debugging meeting and continuously refined the attack code based on colleagues' diagnostic approaches, exacerbating the issue.
- Corrupting the Experiments: Tian modified colleagues' well-trained model weights, making their experimental results impossible to reproduce.
Upon uncovering clear evidence, ByteDance terminated Tian's internship. Instead of taking responsibility, he retaliated by publicly accusing other employees of framing him and manipulating public opinion in a malicious manner.
This, I think, is peak of non-conformist genius, the stuff of the Romance of Three Kingdoms and warlord era. This is the essence of what the Confucian paradigm is trying to suppress, crushing benign self-expression at the same time.
But what if your peers cannot backstab you? What if resources are abundant? What if all your peers are rewarded for exploration and it clearly has positive ROI for them? It might not transmogrify the Chinese into archetypal Hajnalis, who engage in these behaviors without stilts, but the result will be much the same.
Only on greater scale.
R1:
Liang’s meta-derisking – making exploration legible, replicable, and prestigious – could trigger a phase shift. But true transformation requires more than outlier firms. It demands ecosystems that reward speculative genius as reliably as rice farmers once rewarded meticulousness. The question isn’t whether Chinese minds can innovate, but whether China’s institutional lattice will let a thousand DeepSeeks bloom – or if this lone swallow merely heralds a cultural spring that never comes.
China's fragile treasure
Tl;DR: after months of observation, I am convinced that DeepSeek has been an inflection point in Chinese AI development and probably beyond that, to the level of reforming national psyche and long-term cultural trajectory, actualizing the absurd potential they have built up in the last two decades and putting them on a straight path to global economic preeminence or even comprehensive hegemony. It is not clear to me what can stop this, except the idiocy of the CCP, which cannot be ruled out.
Last time I wrote on this topic I got downvoted to hell for using DeepSeek R1 to generate the bulk of text (mostly to make a point about the state of progress with LLMs, as I warned). So – only artisanal tokens now, believe it or not. No guarantees of doing any better though.
The direct piece of news inspiring this post is The Information's claim that DeepSeek, a private Chinese AGI company owned by Liang Wenfeng, is implementing some very heavy-handed measures: «employees told not to travel, handing in passports; investors must be screened by provincial government; gov telling headhunters not to approach employees». This follows OpenAI's new Global Policy chief Chris Lehane accusing them of being state-subsidized and state-controlled and framing as the main threat to the West, popular calls on Twitter (eg from OpenAI staff) to halt Chinese AI progress by issuing O1 visas or better offers to all key DeepSeek staff, and the sudden – very intense – attention of Beijing towards this unexpected national champion (they weren't among the «six AI tigers» pegged for that role, nor did they have the backing of incumbent tech giants; what they did have was grassroots attention of researchers and users in the West, which China trusts far more than easily gamed domestic indicators).
I am not sure if this is true, possibly it's more FUD, like the claims about them having 50K H100s and lying about costs, claims of them serving at a loss to undercut competition, about compensations over $1M, and other typical pieces of «everything in China is fake» doctrine that have been debunked. But China does have a practice of restricting travel for people deemed crucial for national security (or involved in financial institutions). And DeepSeek fits this role now: they have breathed new life into Chinese stock market, integrating their model is a must for every business in China that wants to look relevant and even for government offices, and their breakthrough is the bright spot of the National People’s Congress. They are, in short, a big deal. Bigger than I predicted 8 months ago:
This might not change much. Western closed AI compute moat continues to deepen, DeepSeek/High-Flyer don't have any apparent privileged access to domestic chips, and other Chinese groups have friends in the Standing Committee and in the industry, so realistically this will be a blip on the radar of history.
Seems like this is no longer in the cards.
Recently, @ActuallyATleilaxuGhola has presented the two opposite narratives on China which dominate the discourse: a Paper Tiger that merely steals, copies and employs smoke and mirrors to feign surpassing the fruit of American genius born of free exchange of ideas etc. etc.; and the Neo-China coming from the future, this gleaming juggernaut of technical excellence and industrial prowess. The ironic thing is that the Chinese themselves are caught between these two narratives, undecided on what they are, or how far they've come. Are they merely «industrious» and «good at math», myopic, cheap, autistic narrow optimizers, natural nerdy sidekicks to the White Man with his Main Character Energy and craaazy fits of big picture inspiration, thus doomed to be a second-tier player as a nation; with all cultural explanations of their derivative track record being «stereotype threat» level cope – as argued by @SecureSignals? Or are they just held back by old habits, path-dependent incentives and lack of confidence but in essence every bit as capable, nay, more capable of this whole business of pushing civilization forward, and indeed uplifting the whole planet, as argued by Chinese Industrial Party authors – doing the «one thing that Westerners have been unwilling or powerless to accomplish»?
In the now-deleted post, me and R1 argued that they are in a superposition. There are inherent racial differences in cognition, sure, and stereotypes have truth to them. But those differences only express themselves as concrete phenotypes and stereotypes contextually. In the first place, the evo psych story for higher IQ of more northern ancestral populations makes some sense, but there is no plausible selection story for Whites being unmatched innovators in STEM or anything esle. What is plausible is that East Asians are primed (by genetics and, on top of that, by Confucian culture and path dependence) towards applying their high (especially in visually and quantitatively loaded tasks) IQ to exploitation instead of exploration, grinding in low-tail-risk, mapped-out domains. Conformism is just another aspect of it; and so you end up with a civilization that will hungrily optimize a derisked idea towards razor-thin margins, but won't create an idea worth optimizing in a million years. Now, what if the calculus of returns changes? What if risk-taking itself gets derisked?
And I see DeepSeek as a vibe shift moment nudging them in this direction.
The Guoyun narrative around DeepSeek began when Feng Ji 冯骥, creator of the globally successful game “Black Myth: Wukong,” declared it a “national destiny-level technological achievement.” The discourse gained momentum when Zhou Hongyi 周鸿祎, Chairperson of Qihoo 360, positioned DeepSeek as a key player in China’s “AI Avengers Team” against U.S. dominance. This sentiment echoed across media, with headlines like “Is DeepSeek a breakthrough of national destiny? The picture could be bigger” The discourse around 国运论 (guóyùn lùn, or “national destiny theory”) reveals parallels to America’s historical myth-making. Perhaps the most striking similarity between China and the US is their unwavering belief in their own exceptionalism and their destined special place in the world order. While America has Manifest Destiny and the Frontier Thesis, China’s “national rejuvenation” serves as its own foundational myth from which people can derive self-confidence.
And to be clear, DeepSeek is not alone. Moonshot is on a very similar level (at least internally – their unreleased model dominates LiveCodeBench), so are StepFun, Minimax and Alibaba Qwen. Strikingly, you see a sudden formation of an ecosystem. Chinese chip and software designers are optimizing their offerings towards efficient serving of DeepSeek-shaped models, Moonshot adopts and builds on DeepSeek's designs in new ways, Minimax's CEO says he was inspired by Wenfeng to open source their LLMs, there are hundreds of papers internationally that push beyond R1's recipe… the citation graph is increasingly painted red. This, like many other things, looks like a direct realization of Wenfeng's long-started objectives:
Innovation is undoubtedly costly, and our past tendency to adopt existing technologies was tied to China’s earlier developmental stage. But today, China’s economic scale and the profits of giants like ByteDance and Tencent are globally significant. What we lack isn’t capital but confidence and the ability to organize high-caliber talent for effective innovation … I believe innovation is, first and foremost, a matter of belief. Why is Silicon Valley so innovative? Because they dare to try. When ChatGPT debuted, China lacked confidence in frontier research. From investors to major tech firms, many felt the gap was too wide and focused instead on applications.
NVIDIA’s dominance isn’t just its effort—it’s the result of Western tech ecosystems collaborating on roadmaps for next-gen tech. China needs similar ecosystems. Many domestic chips fail because they lack supportive tech communities and rely on secondhand insights. Someone must step onto the frontier.
We won’t go closed-source. We believe that establishing a robust technology ecosystem matters more.
No “inscrutable wizards” here—just fresh graduates from top universities, PhD candidates (even fourth- or fifth-year interns), and young talents with a few years of experience. … V2 was built entirely by domestic talent. The global top 50 might not be in China today, but we aim to cultivate our own.
BTW: I know @SecureSignals disagrees on the actual innovativeness of all this innovation. Well suffice to say the opinion in the industry is different. Their paper on Native Sparse Attention, pushed to arxiv (by Wenfeng personally – he is an active researcher and is known to have contributed to their core tech) just the day before Wenfeng went to meet Xi, looks more impressive than what we see coming from the likes of Google Deepmind, and it has a… unique cognitive style. They have their very distinct manner, as does R1. They had nowhere to copy that from.
Maybe all of it is not so sudden; the hockey-stick-like acceleration of Chinese progress is a matter of boring logistics, not some spiritual rebirth, much like the hokey stick of their EV or battery sales. For decades, they've been mainly a supplier of skilled labor to America, which masked systemic progress. All the while they have been building domestic schools to retain good educators, training new researchers and engineers without entrusting this to Microsoft Asia and Nvidia and top American schools, growing the economy and improving living conditions to increase retention and have businesses to employ top talent and give them interesting enough tasks… so at some point it was bound to happen that they begin graduating about as much talent as the rest of world combined, a giant chunk goes to their companies, and that's all she wrote for American incumbents in a largely fake, sluggish market. DeepSeek, or Wenfeng personally, is not so much a crown jewel of Chinese economy as a seed of crystallization of the new state of things, after all pieces have been set.
The boost of confidence is visible outside the AI sphere too. I find it remarkable that He Jankui is shitposting on Twitter all the time and threatening to liberate the humanity from the straitjacket of «Darwin's evolution». A decade earlier, one would expect his type to flee to the West and give lectures about the menace of authoritarianism. But after three years in Chinese prison, he's been made inaugural director of the Institute of Genetic Medicine at Wuchang University and conspicuously sports a hammer-and-sickle flag on his desk. The martyr of free market, Jack Ma, also has been rehabilitated, with Xi giving him a very public handshake (alongside Wenfeng, Unitree's Wang Xingxing, Xiaomi's Lei Jun and other entrepreneurs).
…but this is all fragile, because China remains a nation led by the CCP, which remains led by one boomer of unclear sentience and a very clear obsession with maximizing his control and reducing risk to himself. In that, Wenfeng is similar – he's bafflingly refusing all investment, from both private and state entities, because it always has strings attached, I suppose.
“We pulled top-level government connections and only got to sit down with someone from their finance department, who said ‘sorry we are not raising’,” said one investor at a multibillion-dollar Chinese tech fund. “They clearly are not interested in scaling up right now. It’s a rare situation where the founder is wealthy and committed enough to keep it lean in a Navy Seal-style for his pursuit of AGI.”
But you can't just refuse the CCP forever. Reports that he's been told not to interact with the press seem credible; perhaps the story about passports will come true too, as DeepSeek's perceived value grows. In that moment, China will largely abandon its claim to ascendancy, vindicating American theory that Freedom always wins hearts and minds. People, even in China, do not acquire world-class skills to be treated like serfs.
…If not, though? If China does not just shoot itself in the foot, with heavy-handed securitization, with premature military aggression (see them flexing their blue water navy they supposedly don't have in Australian waters, see their bizarre landing ships designed for Taiwan Operation, see their 6th generation aircraft…), with some hare-brained economic scheme – where does this leave us?
I've been thinking lately: what exactly is the American theory of victory? And by victory I mean retaining hegemony, as the biggest strongest etc. etc. nation on the planet, and ideally removing all pesky wannabe alternative poles like Russia, China and Iran. Russia and Iran are not much to write home about, but what to do with China?
The main narrative I see is something something AGI Race: the US builds a God-level AI first, then… uh, maybe grows its economy 100% a year, maybe disables China with cyberattacks or nanobots. I used to buy it when the lead time was about 2 years. It's measured in months now: research-wise, they have fully caught up, releases after V3 and R1 show that the West has no fundamental moat at all, and it's all just compute.
In terms of compute, it's very significant to my eyes that TSMC has been caught supplying Huawei with over 2 millions of Ascend chip dies. This could not have been obfuscated with any amount of shell companies – TSMC, and accordingly Taipei, knew they are violating American decree. Seeing Trump's predatory attitude towards TSMC (them being forced to invest into manufacturing on American soil and now to fix Intel's mess with a de facto technology transfer… as an aside, Intel's new CEO is a former director of SMIC, so literally all American chip companies are now headed by Chinese or Taiwanese people), I interpret this as hedging rather than mere corruption – they suspect they will not be able to deter an invasion or convince the US to do so, and are currying favor with Beijing. By the way, Ascend 910c is close to the performance of Nvidia H800. R1 was trained on 2048 H800s; So just from this one transaction, China will have around 500 times more compute, and by the end of the year they will be able to produce another couple million dies domestically. So, it is baked in that China will have AGI and ASI shortly after the US at worst, assuming no first strike from the latter.
In terms of cyberattacks for first strike, AIs are already good enough to meaningfully accelerate vulnerability search; coupled with the vast advantage in computer-literate labor force (and to be honest, actual state-backed hackers), China will be able to harden their infrastructure in short order, and there's no amount of cleverness that gets past provably hardened code. So this is a very uncertain bet.
In terms of economic growth, this is usually tied to automation. China seems to be on par in robotics research (at least), controls almost the entire supply chain, and has an incomparably bigger installed automated manufacturing base (see their EV factories, which are now also producing robots). They will have OOMs more humanoids and probably faster compounding growth. This more than covers for their workforce aging, too.
Then I hear something about Malacca strait blockade. Suffice to say this seemed more convincing when they really didn't have a «blue water navy», which they now clearly have, contra Peter Zeihan. They're also making great progress in weaning their civilian economy off oil (high speed rail instead of planes, normal rail for freight, EVs again, nuclear and renewable buildouts…) and have stockpiled giant reserves so oil cutoff won't really deter them. They are not quite food-secure but likely won't starve without imports. So blockade is no solution.
Lastly, I've seen this theory that Starship (once it's ready for prime time) provides the US with insurmountable advantage in mass to orbit, thus all the old Star Wars plans are back in action and Chinese nuclear deterrence is neutralized. This doesn't seem feasible because they're working on their own economical reusable rockets – across multiple companies as usual – and are very close to success, and there are signs that this project has very favorable scalability, to the point the US will lose its mass to orbit lead in under three years, or at least it will be diminished. (Personally I think Zhuque-3 is a more sensible design than Musk's monstrosity, though it's just a tasteful interpolation between Falcon and Starship. Learning from mistakes of others is a common late mover advantage).
Sector by sector and attack vector by attack vector, it's all like that.
So… what is left?
As far as I can tell, at this trajectory only China can defeat China – the hidebound, unironic Communists in control, fulfilling the mawkish Western prophecy they try to avoid, bear-hugging to death the young civilization that grew around their mandate and is now realizing its destiny. Confiscating passports, banning open source that widens the talent funnel, cracking down on «speculative investments», dragging them back into the 20th century at the brink of the long-coveted «national rejuvenation».
…Parallels to the US are probably clear enough.
I'd say it's another bit of evidence for Google upgrading their product strategy, but nothing unexpected capabilities-wise. Shame they did not release the weights, instead shipping only Gemma 3 with image-in text-out. «Safety» reasoning is obvious enough.
Contra @SkoomaDentist I think it's not fair to describe this as «The LLM is still talking to the image generator», ie that the main LLM is basically just the encoder for some diffusion model or another separate module. The semantic fidelity and surgical precision of successive edits suggest nothing like that, and point instead to a unified architecture with a single context where each token, be it textual or visual, is embedded in its network of relationships with all others (well, that's what these models are – literally, hypotheses about the shape of the training data manifold). Back when OpenAI announced their image-out capabilities with 4o, the teaser generation said «suppose we directly model P(text, image, sounds) with one big autoregressive transformer». Shortly after, Meta (or really Armen Aghajanyan, who has since departed largely in protest over Chameleon's safety-informed nerfing, and his team) published their Chameleon, a parallel work in identical spirit:
This early-fusion approach, where all modalities are projected into a shared representational space from the start, allows for seamless reasoning and generation across modalities. … Chameleon represents images, in addition to text, as a series of discrete tokens and takes advantage of the scaling properties of auto-regressive Transformers … We train a new BPE tokenizer (Sennrich et al., 2016) over a subset of the training data outlined below with a vocabulary size of 65,536, which includes the 8192 image codebook tokens …
Later, DeepSeek, who are probably the best team in the business (if not for resource limits), have been working on Janus, which is also a unified model of a potentially superior design:
Specifically, we introduce two independent visual encoding pathways: one for multimodal understanding and one for multimodal generation, unified by the same transformer architecture … Autoregressive models, influenced by the success in language processing, leverage transformers to predict sequences of discrete visual tokens (codebook IDs) [24, 65, 75]. These models tokenize visual data and employ a prediction approach similar to GPT-style [64] techniques. … Chameleon [77] adopts a VQ Tokenizer to encode images for both multimodal understanding and generation. However, this practice may lead to suboptimal outcomes, as the vision encoder might face a trade-off between the demands of understanding and generation. In contrast, our Janus can explicitly decouple the visual representations for understanding and generation, recognizing that different tasks may require varying levels of information. … for text understanding, we use the built-in tokenizer of the LLM to convert the text into discrete IDs and obtain the feature representations corresponding to each ID. For multimodal understanding, we use the SigLIP [92] encoder to extract high-dimensional semantic features from images. These features are flattened from a 2-D grid into a 1-D sequence, and an understanding adaptor is used to map these image features into the input space of the LLM. For visual generation tasks, we use the VQ tokenizer from [73] to convert images into discrete IDs. After the ID sequence is flattened into 1-D, we use a generation adaptor to map the codebook embeddings corresponding to each ID into the input space of the LLM. We then concatenate these feature sequences to form a multimodal feature sequence, which is subsequently fed into the LLM for processing. The built-in prediction head of the LLM is utilized for text predictions in both the pure text understanding and multimodal understanding tasks, while a randomly initialized prediction head is used for image predictions in the visual generation task.
I expect DeepSeek's next generation large model to be based on some mature form of Janus.
I think Gemini is similar. This may be the first time we get to evaluate the power of modality transfer in a well-trained model – usually you run into the bottleneck of the projection layer, as @self_made_human describes. But here, it can clearly copy an image (up to the effective "resolution" of its codebook and tokenizer) and make isolated transformations, precisely the way transformers can do to a text string. Hopefully this means its pure verbalized understanding of the visual modality (eg spatial relations, say… anatomy…) is upgraded. Gooners from 4chan ought to be reaching the conclusion as I type this.
In the next iteration video and probably 3d meshes are getting similar treatment.
P.S. SkoomaDentist being bizarrely aggressive and insistent that this is whatsoever like inpainting is being very funny. Inpaint this. No, no, these are not vulgar tricks, and I don't see why one could be invested in bitterly arguing against that.
DaseindustriesLtd
late version of a small language model
1yr ago
Manus is a generic thin wrapper over a strong Western model (Sonnet 3.7), if a bit better executed than most, and I am quite unhappy about this squandering of DeepSeek's cultural victory. The developers are not deeply technical and have instead invested a lot into hype, with invites to influencers and creating a secondary invite market, cherrypicked demos aimed at low value add SEO-style content creation (eg “write a course on how to grow your audience on X”) and pretty UX. Its performance on GAIA is already almost replicated by this opensource repo. This is the China we know and don't very much love, the non-DeepSeek baseline: tacky self-promotion, jumping on trends, rent-seeking, mystification. In my tests is hallucinates a lot – even in tasks where naked Sonnet can catch those same hallucinations.
The real analogy to DeepSeek is that, like R1 was the first time laymen used to 4o-mini or 3.5-turbo level slop got a glimpse of a SoTA reasoner in a free app, this is the first time laymen have been exposed to a strong-ish agent system, integrating all features that make sense at this stage – web browsing, pdf parsing, code sandbox, multi-document editing… but ultimately it's just a wrapper bringing out some lower bound of the underlying LLM's latent capability. Accordingly it has no moat, and does not benefit China particularly.
Ah well. R2 will wash all that away.
It is interesting, however, that people seemed to have no clue just how good and useful LLMs already are, probably due to lack of imagination. They are not really chatbot machines, they can execute sophisticated operations on any token sequences, if you just give them the chance to do so.
DaseindustriesLtd
late version of a small language model
1yr ago
My radical thesis is that both are shitholes but not really militarily inept ones in the way people might imagine.
I do know this and I wonder how that coexists with the common East Asian respect for the Hebrews. Have they considered playing one great tribe against another? Or learning the Manchu ways to beat them at their own game, like Koreans try with Talmud? I should ask.
More options
Context Copy link