RandomRanger
Just build nuclear plants!
No bio...
User ID: 317
RandomRanger
Just build nuclear plants!
1d ago
We need to distinguish between 'capital needed to achieve it drying up before it can be reached' and 'demand is so high that they have to ration resources'.
They kind of look the same but the underlying meaning is different. The former implies the Bubble is Popping whereas the latter implies It's Not a Bubble.
Firstly, I don't think the capital is drying up. Hyperscaler AI infrastructure spending rises year by year. Secondly, demand is huge. Anthropic ARR is now at $30 billion ARR (by their figures, though OpenAI says the real figures should be a few billion lower, depending on how you measure revenue shares). Whichever way you look at it, huge demand growth. $87M annualized run-rate in January 2024 → $1B by December 2024 → $9B by end of 2025 → $14B in February 2026 → $30B in April 2026 is pretty impressive, even if its juiced.
Clearly they're getting lots of demand. There are also issues with slow datacentre rollouts and delays due to the absolute state of Western electricity and construction sector. I think the phenomenon we're absorbing is rooted in high demand, not investors getting antsy and demanding higher returns.
RandomRanger
Just build nuclear plants!
1d ago
White men quite clearly have been naive and greedy and you guys exploited that masterfully. It was idiotic to invest so much in a hostile country because something something middle class democracy liberalization. China certainly isn't making the same mistake, they're not investing in outsourcing their manufacturing to India, they're trying to constrain Indian manufacturing.
This is our repayment for saving China from Japan, I suppose. The Hump, the Burma Road, all the Australians, Americans, Brits who bailed out China ended up fighting them in Korea, getting displaced economically later on.
Unfortunately for China, I think it's too late. The compute advantage is too great, singularity too close, Chinese fabs too far behind. Chip sanctions, albeit inadequately enforced, albeit undermined by Jensen's heroic lobbying/bullshitting efforts, will be sufficient. Mythos and its successors will overmatch Chinese human capital. The ultimate outcomes may well be bad for most of humanity but it'll surely be crushing for China.
I don't understand this mindset of sneering at the other side for being 'defeatist' if they pursue their strongest strategy in a straightforward competition, while simultaneously bragging about one's own power and the inevitability of their defeat. At least Jensen is trying to sell his chips, what are you trying to sell to a presumably mostly white, mostly American audience? It's not defeatist to take the most plausible path to success, jettison all the cope about magic dirt and wield some power.
RandomRanger
Just build nuclear plants!
1d ago
True, they sort of could innovate. Soviets were a capable technological opponent, the T-64 was far ahead of contemporary tanks and they made good use of what they had. But they were usually behind and only rarely ahead, there was no general trend of them creeping forward in all these domains, only occasional exceptions to the general rule. Soviet goods were also very uncompetitive on world markets, it was mostly just natural resources that they could export.
I was mostly thinking about electronics and chips where they had this excellent espionage system that secured all these chips and blueprints but never really got around to domestic R&D and quality production, usually they just copied and that kept them behind. Soviet innovation was not like Chinese innovation. China is not restricted to the most godawful cars and leaky refrigerators, televisions that occasionally explode...
RandomRanger
Just build nuclear plants!
1d ago
Yes, the UK and France send ships to help with US wars at times, UK aircraft help defend Israel, UK bases are used for US bombers. Australian AWACs planes are helping the US and gang out in the Middle East.
Or how the US instructs the Netherlands not to sell ASML chip equipment to China. They're squarely in the US camp.
Firstly, power is zero sum. With regards to conflict, power is the lens to use, not wealth or positive-sum dynamics.
It's idiotic for the US to sell China AI chips. China is already trying hard to make their own AI chips and chips generally. Any big power would want to secure such critical industrial, economic, military resources with domestic production. They're not going to stop if Nvidia sells a few billion more in GPUs. Nor is China going to accept a subordinate position in the US tech stack, they're not stupid and have been playing their well-honed playbook of imitation, innovation, espionage and absorption of foreign IP. Jensen Huang must think the audience are stupid with this rubbish:
The idea that you’re going to have an AI agent running around with nobody watching after it is kind of insane. We know very well that this ecosystem needs to thrive. It turns out this ecosystem needs open source. This ecosystem needs open models. They need open stacks so that all of these AI researchers and all these great computer scientists can go build AI systems that are as formidable and can keep AI safe. So one of the things that we need to make sure that we do is we keep the open source ecosystem vibrant. That can’t be ignored. A lot of that is coming out of China. We ought to not suffocate that.
With respect to China, of course we want the United States to have as much computing as possible. We’re limited by energy, but we’ve got a lot of people working on that. We’ve got to not make energy a bottleneck for our country. But what we also want is to make sure that all the AI developers in the world are developing on the American tech stack, and making the contributions, the advancements of AI—especially when it’s open source—available to the American ecosystem. It would be extremely foolish to create two ecosystems: the open source ecosystem, and it only runs on a foreign tech stack, and a closed ecosystem that runs on the American tech stack. I think that would be a horrible outcome for the United States.
'We need open source', no, Nvidia needs open source to increase competition amongst its clients.
The amount of energy they have is incredible. Isn’t that right? AI is a parallel computing problem, isn’t it? Why can’t they just put 4x, 10x, as many chips together because energy’s free? They have so much energy. They have datacenters that are sitting completely empty, fully powered. You know they have ghost cities, they have ghost datacenters too. They have so much infrastructure capacity. If they wanted to, they just gang up more chips, even if they’re 7nm.
I’m telling you what it is. They have plenty of logic, and they have plenty of HBM2 memory.
And this is just bullshit. They don't have anywhere near '4x or 10x' as many chips because of export controls and a shortage of HBM too, because of export controls and sanctions. Amazon has more compute than all of China. Jensen is just nakedly grasping for any advantage, any line of reasoning that gets him where he wants to go, which is selling more.
First of all, the way to solve that problem is to have dialogues with the researchers and dialogues with China, and dialogues with all the countries to make sure that people don’t use technology in that way. That’s a dialogue that has to happen. Okay? Number one.
Maybe Jensen should head off and try to convince China to stop hacking every drop of IP they can get their hands on, see how that works out for him. It's just a profoundly unserious way to look at the world. Jensen's abilities have made me lots of money, so I'm not opposed to him totally. But there's this shameless inability to accept that Nvidia and US interests might not be the same. He weaves around good faith argumentation like Neo dodging bullets in the Matrix, just constantly attacking and pressing and cajoling and bullshitting. I'm not American either. But the nakedness of the duplicity is just staggering.
Even if the singularity were not imminent, it'd still be dumb to sell these chips from an American strategic point of view. AI is useful for chip development, Google has been using AI for chip design for many years now. AI would be helpful for squeezing out more yield on their inferior processes. Selling China chips accelerates their chip development. Selling China chips also accelerates lots of other research, AI cyberwarfare capabilities, military ISR and economic competitiveness generally... And he's just going on about how they need to be open source developers on the US stack, like that matters at all? It doesn't matter at all. The notion that huge companies are going to spend tens of billions of dollars training massive AI models and then open sourcing them so they won't make any money from them is absolutely retarded. Open source can only lose, in the long run. I like open source and use open source but I'm realistic about it. Nvidia doesn't open source anything if they can actually make money from it: they patent CUDA architecture, tensor cores, NVLink architecture. They're not retarded. Facebook has switched to proprietary with Muse. Even Facebook has cottoned on here.
In the Cold War the US was loathe to sell the Soviet Union advanced machine tools. It was obviously stupid to sell tools that could make stealthy submarine propellors, C&C tools with high tolerances for advanced engines. Fortunately the Soviets could mostly only copy and not innovate like China can.
Secondly, the US government and elite as a class does not accept the legitimacy of any non-liberal democracy. They might decide that it's too hard to overthrow any given power at a given point in time, that the costs aren't worth the gains at any given point. But if they have the power, they'll give it a shot. Iraq, Iran, Syria, Libya, North Korea and Russia are all to some extent considered targets. So is China. The US leadership class is not going to change its mind about this.
Even if China was a liberal democracy it'd still be very threatening because of how big and rich China is. If you read Colby's book on strategy he lays this out. If China reaches a certain level of strength, China becomes the lynchpin of Asia and the rest of Asia falls into China's orbit. If Asia falls into China's orbit, Chinese standards, Chinese technology, Chinese markets, so too does much of the rest of the world. Then China can start interfering in the Americas and undermine US national interests. The US will not have sufficient power to resist this, since it's a fundamentally smaller country. Colby makes all the song and dance about how they're only really opposed to the Chinese government... but at the end of the day we know the Chinese government has the support of the people, they're nationalistic and quite reasonably so. They want the best for their country and that means expansion overseas, in one form or another.
It doesn't matter if current Chinese leaders say they have no such interests, or even if they're being sincere. Leaders change. The US used to be isolationist and changed. Power corrupts. Huge powers have global interests, especially today. They get drawn into conflicts, they feel immense pride in their power, glory in their victories. China is no different from any other country in potential for rapacity and aggression.
Furthermore, America does not have a chance in competing with China on an even playing field. Chinese people are smart and very hard working. The Chinese system is very capable. They don't have a construction sector that squanders billions building imaginary railways. They don't have a fent zombie infestation in their biggest cities. They don't have a political class behoven to Israel. They don't have people like Jensen running around undermining export controls with their lobbying, China's economic elites obey instructions. China's internal problems pale in comparison to America's.
America would have to be 4-5x more capable per person to compete with China just based on pure population. That's incredibly difficult! Without AI, without its ever-diminishing time-based advantages in technology, America cannot compete with China. The great hope seems to be that Chinese demography falls off ... in an age of automation where China already has a huge industrial and robotics base. In an age of AI. In an age where everyone has growing youth unemployment. The demographics cope is just cope. China is huge, has enormous cohorts of highly educated young people. Demographics will not avail America.
China's optimal strategy is 'lets be friends, win-win cooperation, surely you'd never attack a peaceful country like ours, lets trade and cooperate and especially share technology'. Then once they have the technology and a giant fleet, a military budget 2-3x America's, all-domain technological superiority, then comes revenge for past humiliations, real and imagined. That's the privilege of size and intellect.
America's optimal strategy is looking for some kind of force-multiplier, a wonder-weapon that can be used to lock in its advantages. AI fits perfectly. Frankly, America has been ridiculously, impossibly generous to China, as generous to China as it has been destructive in the Middle East. America opened up domestic and world markets to China, China systematically grabbed every strategic market it could with aggressive state-backed industrial expansion. And exported inordinate amounts of fentanyl precursors and toxic social media like tiktok to America. They're playing the game. The US can also play the game and take steps to cripple and constrain China.
RandomRanger
Just build nuclear plants!
2d ago
I don't accept that the war is over. I don't even think Trump accepts it:
https://x.com/RapidResponse47/status/2045352142253252818
"If you don't have a deal by Wednesday when the ceasefire ends, will you extend the ceasefire?"
@POTUS : "I don't know. Maybe I won't extend it — but the blockade is going to remain... Unfortunately, we'll have to start dropping bombs again."
Nor are the straits even open. Iran says the straits are open but only if you go through their route between their island, fill out their paperwork and aren't a hostile country and maybe pay a fee. US says the straits are open but not to Iran or anyone who pays a fee to Iran?
RandomRanger
Just build nuclear plants!
2d ago
Well since you aren't on the unemployment/UBI queue yet, it follows that AI can't do your job. AI is pretty bad at using interfaces designed for humans.
But the hard part has already been done. I can prompt out a shader-evolution program, it basically evolves, based on human ratings of who propagates, generations of graphical animations, patterns of energy that would make pretty screensavers, save the ones I like to a folder.
That's easy, a couple of prompts.
Or I can have it write out fun skits, like the Iran/Hormuz situation mashed up with the "Ship Front fell off" skit: https://youtube.com/watch?v=3m5qxZm_JqM
The raw intelligence is there, there are just certain fiddly parts and a level of precision that's missing. A skit doesn't necessarily need to be precise, if code doesn't work right you can just tell it what went wrong and have it improve. Whereas with sending payment documents to people, precision is quite important.
I use it for work too, mildly superhuman proofreading and some other tedious document processing tasks, some of which need a certain level of judgement and wisdom. Precision is pretty good. But for these I have to use my custom scripts for that specific hours-long task, scripts that I've tried and tested over some time. And they really need to be more 'document input, document output' tasks rather than 'messing around with portals and niche software'.
But that's now, they're working on making agents reliable and cost effective. Eventually they'll get there. I can't understand a scenario where it's possible to create fairly complicated pieces of custom code with AI but logging into outlook, sending emails, logging into whatever govt portal, getting a 2FA code... is some domain reserved for humans alone. In the long term either AI has to be dumb as a rock/incapable of anything but specialized tasks like chess/go or it can be general purpose, doing anything a human can do.
the UK as its current society no longer exists in 50 years. Maybe reality has to be the escape valve that forces women's beliefs to become moored to reality again.
Well does the society of UK circa 1976 exist today? Coal miners. British Steel. The Mid Atlantic accent. Roger Moore. Rumpole of the Bailey. That Britain is dead and gone, it may as well be another world.
This isn't quite a 'change happens so you should accept bad change' post. It's certainly bad if high-quality peoples like the British are dying off. But I don't think we have to care about sustainability into the long term future like we used to. 50 years is too far away. Reduce TFR to 0, why not? AI will certainly be good enough to compensate within 10 years, let alone 50.
How does TFR matter? We invest in a society that produces children today. In 18 years the kid is qualified to be an entry level worker, after 100s of thousands more in investment? +4 years for university? What is the estimated return of this? I say zero. That kid is not going to be militarily, economically, politically relevant. The AI of 2044, when he is 18, will be overwhelmingly more capable than he is. Society is a machine for producing and distributing resources, so that groups can enjoy wealth and military power. That's why we have culture and capitalism and politics.
It would of course be ideal if we had healthy, confident, advancing civilizations rooted in correct values rather than delusions. But the fact that we don't may not be fatal. It's harmful mostly in the delusions damaging or derailing the transition in this final stage of human life.
But the distinction is that I don't think we need a real cure, only a quick fix. Better a band-aid than deep cultural medicine if it takes 20 years to work. Better to burn the bond markets, state finances, TFR, climate and social trust as long as the right people control the datacentres and AI models.
I know this is kind of boring as a post, feels like it trivializes and robs the meaning out of every single issue. I think it's fine to talk about non-AI things and not have it injected with AI but I think long-term thinking at this point is a trap.
RandomRanger
Just build nuclear plants!
5d ago
In what sense are they required?
Well the system we have has this kind of 'harmony' in that it evolves to meet the needs of the economy and the economy evolves to meet the needs of the system. Managerial state, managerial culture, niceness and rule-abiding, cancellation of dissidents because some middle-aged woman decided you shouldn't have a bank account or legal representation or whatever. And the economy gets altered by the system, so we get lots of pensions and healthcare spending and no nuclear energy. There sure are a lot of tensions and conflicts going on, 'contradictions' in Marxist terms, but the basic system remains intact.
My point is that rugpulling the economic base will rugpull the ideology too.
RandomRanger
Just build nuclear plants!
6d ago
America is presumed to be acting in all manner of irrational and stupid ways because
they are acting in all manner of irrational stupid ways.
Firstly they started this war thinking that the Iranian people would rise up and overthrow the regime, which was stupid. Then they go back and forth unsanctioning Iranian oil to lower fuel prices, proposing dual control of the straits, now blockading Iran, threatening to destroy Iranian civilization, then constantly rolling back threats.
It's a pathetic display of weakness and stupidity as some very mediocre intellects thrash around trying to escape a self-inflicted blunder. It's exceeded only by cheerleaders dressing up the flailing as 4D chess.
RandomRanger
Just build nuclear plants!
6d ago
Annihilating Pyongyang with what? North Korea could shell Seoul intensively with their artillery, nevermind nuclear attacks. A few measly ballistic and cruise missiles are no match for conventional artillery in just wrecking whole areas.
RandomRanger
Just build nuclear plants!
6d ago
He consistently states he wants the benefit of the technology his company is developing to be widely distributed
But Altman lies constantly about everything. He started off saying he was running a non-profit, then weaselled out of it. Ilya hates him and accused him of lying. Dario hates him and accused him of lying. Helen Toner accused Altman of failing to inform the board about the ChatGPT launch and hiding his financial interest in the OpenAI Startup Fund.
Even in that very trial it brings out their plans to basically run off with Elon's money and make a for profit.
He might just be saying that because it sounds a lot better than 'this is my path to universal domination, you dumb, dumb fucks'. Deceptive people shouldn't be trusted.
RandomRanger
Just build nuclear plants!
6d ago
There's a political ideology that's good at rallying moralistic middle-aged women, teachers, doctors, officials, NGOdom... They do the caring for the homeless.
But is that ideology strong with tech billionaires or high-ranking military/intelligence spooks? Billionaires and spooks are the ones to worry about with AI takeover, they have their hands on the buttons.
An economic system that produces and requires a bunch of middle-aged women in office jobs, a bunch of teachers and bureaucrats and officials and journalists, that seems to produce leftism we recognize today, just like a system that requires a bunch of professional heavy cavalry leans towards martial valour and manly vigour...
But take away that economic system and replace it with robots, we'd see something quite different surely?
RandomRanger
Just build nuclear plants!
6d ago
This would prevent Iran nuclear efforts
Why might Iran want nuclear weapons? In large part because the US goes around attacking countries that lack nuclear weapons. Kim Jong Un sleeps at night because he has nuclear weapons. Saddam Hussein, Qaddafi and Assad are dead or in exile because they lacked nuclear weapons.
The best way to prevent Iran acquiring nuclear weapons is to negotiate sincerely, flush every piece of hysterical Israeli 'intelligence' straight down the toilet and get on with pursuing national interests, in contrast to slavishly serving Israel.
That's what Europe's general stance has been. Diplomacy, trade, win-win. The US has wrecked this and done its absolute best to make a strong case for Iranian nuclearization. Quite clearly Iran's conventional forces have failed to deter attacks on the country.
The US launched this retarded war, without any clear plan for victory, without even the necessary capabilities for victory. The US army can't be deployed quickly enough to bases that are being bombed out and would probably bog down in the first couple of major cities before even reaching Tehran. The US air force can post all these epic videos of explosions (real or decoys, nobody knows) but cannot seem to prevent Iran firing off waves of missiles and drones, cannot defeat Iran's plan to inflict economic pain until the US gives up.
Trump has played into the Iranian plan masterfully, obsessing about market manipulation. He shows weakness every day of the week, while constantly backflipping and making fresh ultimatums, then extending them. One day he unsanctions Iranian oil to lower prices, the next he blockades Iran. Inconstant and incoherent. An ideal target for a prolonged economic campaign.
The US navy can blockade Iran but cannot undo Iran's blockade of the straits. They cannot defeat Iran's plan.
European forces don't have the capability to do what the US can't, there is no reason for them to charge into the valley of death. It's egregious for the US to pussy out, slinking away from the straits with the 'world's most powerful navy' and demand their allies charge in and die, for the sake of a war that directly harms their interests.
Model collapse is not really a major concern. The original researchers in that paper trained small models on only AI outputs (of the previous model). Them being small models, they made mistakes and the mistakes compounded over time. It's more like a Chinese whispers experiment.
Big companies make great use of synthetic data and autonomous training, in addition to human originated data. For example, consider Deepseek R1-Zero, which was just trained on reinforcement learning, verified signals and not human reasoning patterns. It was kind of weird and switched languages a lot but it did work and got smarter over the course of training. In fact, all modern models are trained in this way. When Claude occasionally slips into Chinese for a single word it's not because any human ever does that in the training corpus, it's because during the training process they have them autonomously bootstrap and get smarter over time and that's just how it goes. AIs are omnilingual by nature it seems.
More options
Context Copy link