RandomRanger
Just build nuclear plants!
No bio...
User ID: 317
RandomRanger
Just build nuclear plants!
8mo ago
- Sexual harassment. There are all these newspaper headlines coming out of India about animals being raped to death, women who go there instantly regretting their decision. Or on university campuses at home women complain. We are the ones who invented #metoo and expended great effort getting Afghan girls into university, this does not go down well at all.
- A certain attitude. The Chinese probably do much more scamming than India but they're stealing turbine information, IP, software, schematics. They don't tend to go after grandma's savings saying they're from Microsoft. China comes off as threatening (nearly every day we have war propaganda in TV and newspapers against China). India is not threatening at all. But there's a certain kind of boasting/hypernationalism that you can see sometimes online, a certain level of entitlement to other people's money. Kitboga has done immense damage to India's reputation: https://youtube.com/watch?v=6m8Ln1yqeJE
RandomRanger
Just build nuclear plants!
8mo ago
From reading the wiki page, I'm having a hard time figuring out why anyone would mouth-froth over the idea of her having any position of power.
She says she was put under surveillance in the airport for some reason, agents follow her on planes:
https://www.newsweek.com/tulsi-gabbard-tsa-terrorist-watchlist-1985527
So there's clearly not much love lost between her and intelligence goons.
Relevant: Dominic Cummings complaints about the UK government and its clownish bureaucracy
For the first year of Gove’s time in the DfE (May 2010 – spring 2011), ministers were up until the early hours proofreading officials’ drafts of letters and rejecting about nine out of ten because of errors with basic facts, spelling, or grammar. When I got embroiled in rows about this in Q1 2011, some MPs had been sent no reply for six months. Despite several complaints to senior officials, nothing happened, shoulders were shrugged – ‘cuts, we need more resource, lack of core skills, all very difficult’ and so on.
The problem with Western governments isn't that they literally can't find people who know how to spell, or fix lifts, or avoid idiotic wars.
There are plenty of smart people in government and even more who are theoretically available. The whole institutional structure doesn't prioritize doing things correctly. As a collective, they pursue vibes of what they think might be popular amongst their peers (see Team Kamala's decision of why not to go on Rogan). They try to strengthen the power and control of their class and subdue any threats (this is their highest shared priority). They try to deflect all bad outcomes away from themselves. And they like to plot and play politics, diverting national resources for their own internal factional interests. That's how they rise up the ranks.
The key thing isn't scrapping programs or reducing spending but changing the whole incentive structure and culture so that stupid programs aren't initiated and wasteful spending doesn't emerge in the first place. Politicians and officials must not feel safe going 'let's invade this country for made-up reasons' and creating a mess. They must not feel safe wrecking national industries. In the private sector, if you wreck and blunder you end up sinking your company and getting removed from the leadership pool. Ideally you're sifted out through competition before you get into any high-ranking positions. You can't really wriggle out of that (though some manage it).
In the public sector, it's very difficult for even the most effective wreckers to completely destroy a country. Competition between states is quite limited in most places. Responsibility is so diffuse they can lay blame elsewhere. The culture gets more corrupt under the lesser competitive pressure.
To take a less contemporary example, consider Admiral Yi of Korea: https://en.wikipedia.org/wiki/Yi_Sun-sin
He was an incredible leader on the battlefield but he was constantly getting imprisoned, tortured and demoted by his jealous rivals and nervous superiors. The Korean governance culture was inferior, it squandered enormous amounts of talent. We can see a similar kind of suppression (albeit much less severe) on Musk under the old regime, despite him clearly being an incredible strength for America. Presumably his European equivalent got suffocated before he even got started.
That's what needs to be changed, the entire mindset. This is very hard to do, creating good institutions in the first place needed hundreds and hundreds of years of bloody wars in Europe. Maybe we could try introducing fearsome anti-corruption commissions and merit-based promotions like they have in China. But even then, there are problems with people gaming the rules: 'if the mayor is fired when a disaster kills 36 people, then all disasters will be reported as killing 35'. That example may not be specifically true but it gives a general impression.
Only a clear and inescapable need for true performance can really get it done. I don't know how to achieve this, apart from warfare or international races to achieve a certain goal.
RandomRanger
Just build nuclear plants!
8mo ago
I think the most palatable change would be something akin to banning those under age 16 from having social media accounts.
They're trying to legislate this in Australia right now.
Does everyone need to show their ID to get a social media account up? Do I need to show ID for this website? Who is storing my information and where is it going? What about VPNs (every second youtuber is shilling them, kids could easily set one up) to bypass the ban and log in from a more permissive jurisdiction - ban them? What about 4chan or its derivatives - they can't be bothered to do age-verification (and don't have the resources) - ban them? Xbox and Playstation have online chat, thousands and thousands of games have online chat. Are they all social media? In a stroke online privacy is greatly diminished, along with all small web forums.
At least in crypto you have marginally higher chances of making money and it's not inherently rigged against you. Down with sports betting, up with Shiba and Doge.
Societies' restrictive energies should remain focused on drugs, they cause much more harm than gambling does.
RandomRanger
Just build nuclear plants!
8mo ago
There are beginners playthroughs on youtube: https://youtube.com/watch?v=CgBnpbaQFo4 or https://youtube.com/watch?v=_f-pwq6cKwk?list=PLs3acGYgI1-vw-A3LHOb_BDQxKNtv1tze
There's a text guide here (this would be the best IMO for getting started, in terms of efficient reading): https://forums.civfanatics.com/threads/beginner-help-the-basics.648469/
There's a slightly more advanced tactic/strategy guide here: https://forums.civfanatics.com/threads/sisiutils-strategy-guide-for-beginners.165632/
The game manual is here: https://forums.civfanatics.com/resources/civ-4-manual.12753/
The map is pretty straightforward. It's all about getting three resources - food, commerce and production. There's a little button you can press on to show the per-tile yields, another one that highlights special resources.
You get the most value in making cities near food resources so they can quickly grow and get pops working other tiles: hills, mineral resources and forests for production or luxury resources/coast/rivers for commerce. Commerce is wealth, culture, espionage and most of all research, you control where exactly it goes with sliders.
RandomRanger
Just build nuclear plants!
8mo ago
https://www.pcworld.com/article/1364477/nvidia-rtx-4090-vs-nvidia-rtx-3090.html
The GeForce RTX 4090’s $1,599 MSRP is significantly less than the $1,999 whopper of a price that the RTX 3090 Ti launched with. It’s also $100 more than the original RTX 3090’s debut $1499 price. Good news, however – the RTX 3090 Ti has dropped to a much lower $1099 for the Founders Edition, and sometimes can be found for less. The 3090 can often be found for under $900, and even closer to $700 if you’re OK with a used graphics card.
The 3090 price fell precisely because of the 4090, due to market forces. Today, the 4090 seems to be back up to around 2000 USD due to the AI boom and sanctions/sanctionsbusting. But anyone would rather have a 4090 for $2000 than a 3090 TI for $2000. In theory, you could get a 4090 for $1600 compared to a 3090 TI for $2000, which is a very good deal. Progress continues.
When the 5000 series emerges, the 4090 will fall to the $1000-1500 range too.
Secular falls in GPU prices (and heightened price/performance) are being suppressed by high demand but they're still observable.
If Nvidia's products have the same or worse price/performance ratio as 5 years ago then why are they the biggest company in the world today and a minnow five years ago? Shouldn't it be the other way around?
For some tasks, there's no difference. My favourite game Civ IV can run on 20-year old hardware. It runs a little faster on a modern CPU but that's about it.
The 4090 is not really for gaming, it's for mucking around with advanced image-generation, AI and training consumer-level LORAs. For some things the 4090 really is the cheapest way to run it, there is 0 performance per $ for anything below 24 GB of VRAM. Just like how there is 0 performance per $ on the Geforce 3 for most tasks. It doesn't even run modern OSes, you'd be better off with whatever comes with your CPU.
NVIDIA boasts 25x energy efficiency gains over the last generation for its flagship AI processors. OK, that's advertising - round that down to 10x or even 5x... That's still a huge improvement.
https://www.nvidia.com/en-au/data-center/gb200-nvl72/?ncid=no-ncid
RandomRanger
Just build nuclear plants!
8mo ago
UPRO doesn't seem to have performed that well.
https://www.marketwatch.com/investing/fund/upro
On 1/1/2020, it was at $36, now it's at $95. Tesla went from $36 to $335. Bitcoin did even better, going from about $7,000 to $88,000 today. Even Apple went from $74 to $224, it did better than UPRO (and pays dividends). Microsoft did similarly well.
It's not like Microsoft or Apple were unheard of back in early 2020, they're basically blue-chips.
ETFs are generally mediocre investments and have management fees, better to just pick out stocks or crypto specifically. If we look at just the 1 year, Bitcoin is up 140%, UPRO is up a measly 100%. UPRO might be a decent investment but it's not a great one.
A decent amount of volatility is good. You want to get in before the institutional investors, not after they've pumped the market up to high heaven. They're already all over ETFs.
RandomRanger
Just build nuclear plants!
8mo ago
Moore's law has slowed down but his point is about a general trend of compute becoming cheaper per dollar, not a specific trend of transistors miniaturizing. He traces out a similar pattern of acceleration back to the dawn of life on Earth, long epochs of tiny creatures, followed by larger and more complex life. And then bang! It's the Anthropocene, goodbye to all the land mammal biomass that isn't us or ours. Even before transistors there was acceleration in compute capacity through electromechanical computing. Presumably acceleration will continue into photonics or some other method, perhaps with a delay period or sudden acceleration. You could argue that it's still accelerating, if you include the software and architectural improvements in the newest GPUs their effective compute/$ for AI tasks is rising faster than before.
I think his AI predictions turned out quite well - his original prediction was AGI by 2029 which looks conservative, if anything. Many today give a date of 2027, assuming all goes according to schedule. Singularity by 2045 is even more conservative. He was saying this back in the early 2000s, so clearly his reading of trend-lines has some merit to it.
His health practices however will probably not stand the test of time.
RandomRanger
Just build nuclear plants!
8mo ago
He's saying that we'll have to scale up something else - inference time is the most obvious choice. Synthetic data is also widely used. Ilya wouldn't be founding his own 'making superintelligence' company if he thought it wasn't possible.
RandomRanger
Just build nuclear plants!
8mo ago
The new Qwen 2.5 32B dropped, people are saying it's roughly as good as the newest Sonnet for coding. I don't know how easy it is to get access to Qwen but it is Chinese and should be cheap, it is open source...
Might be good to try out as a comparison, see if it's really that good or if they've just been benchmark hacking? But Sonnet is the most obvious pick IMO.
RandomRanger
Just build nuclear plants!
8mo ago
If you jumped onboard 14 hours ago, you'd already have made 7-8% profit, which is what SPY might make in a year. It's at 87K now, rising to 88 as I write this post.
Untold riches for nothing is a very high standard that we've only ever seen with bitcoin and ETH (which was originally distributed to BTC addresses). There used to be BTC faucets where people gave them away, evangelizing to new users.
RandomRanger
Just build nuclear plants!
8mo ago
Bitcoin is at $81,000 right now...
Over 3 years (from the last ATH in November '21) it's roughly even with SPY, maybe a little behind. Over 2 years it crushes SPY. Over 4 years (and any further) it crushes SPY.
RandomRanger
Just build nuclear plants!
8mo ago
A big problem is how you measure intelligence.
Any human can count the number of r's in strawberry but most would be hard-pressed to translate Chinese to English or do anything at all in python or other programming languages. Intelligence is multi-domain, possibly the most complicated thing we can try to measure.
Even within domains, how do you rate intelligence? Sometimes it would be worth spending 10x more to get a marginally better programming AI because 'marginally better' is like a 0 to 1 increase in that it provides genuinely useful input that a human can use to get a good answer and speed up their work.
A new brutally hard question set dropped a few days ago. I have no idea what any of this gibberish means, it's well beyond me.
https://epochai.org/frontiermath
The two toned-down o1 models get 1%, along with GPT-4. Claude 3.5 and Gemini Pro get 2%. Does it follow that Claude 3.5 new is smarter than I am since I would get 0 and that therefore AGI has been achieved? Probably not, Sonnet 3.5 makes all kinds of order of magnitude mistakes that I can eyeball as wrong. But it is pretty damn smart and noticeably smarter than its older incarnation, it has certain new tendencies in writing that qualitatively improve it.
GPT-2 would get 0 on nearly all benchmarks because it just babbles. GPT-3 would also get 0 because it just babbles (albeit more interestingly), remember these are the base models that might just answer your question with more questions. GPT-3.5 was the inflection point where AI became useful for a bunch of things, for consumer use. The old GPT-4 (as of March 2023) opened up coding. Opus 3 was thought to be the first really creative writer, it can maintain an engaging twitter persona. Sonnet 3.5 is on a whole new level in coding, opening up Cursor. The newest Sonnet, o1 and Gemini can start to barely grapple with these advanced mathematical questions.
From the perspective of 'can it answer Frontier Math', there have been no advances in AI before the last 6 months or so. Intelligence is so complex, what looks like a slowdown in one domain can just be the start of something new in another domain.
RandomRanger
Just build nuclear plants!
8mo ago
True, on reflection there's a lot of flexibility with these things. The US used to only focus on the Americas as its sphere of uncontested influence - that changed into a global crusade.
China used to be principally concerned with mainland Asia and its immediate neighbours, acting in Korea, Vietnam and India. But even in the Maoist era they had a global foreign policy, propping up Albania against the USSR. Today they're still most interested in immediate neighbours but they do have global interests in resources, investments, infrastructure and so on. Australia is competing to out-influence China in the Solomon islands, well beyond the Nine Dash Line.
They're a big power and I think they have big ambitions. They're feeling the same seductive rush of power that saw America head out into the world all those years ago.
RandomRanger
Just build nuclear plants!
8mo ago
China wants to work peacefully with us
They've been saying that for ages, they have this holier-than-thou attitude where they go 'unlike the US, we think the world is big enough for America and China to be big powers - also stop making provocations in the South China Sea and encouraging separatism, you're stirring up trouble and spreading a Cold War mindset'.
The Chinese version of 'working peacefully with us' is just the same as the US version of 'being held accountable to the international community', it's a polite way of saying 'we are the good guys, we set the fundamental rules on what's acceptable, you can retain some sovereignty but not where it crosses our red lines'.
Lots of people in Iran are furious with Trump for reneging on the nuclear deal, Maximum Pressure and killing a high-ranking general of theirs. The Iranian government released a fan-made trailer for an assassination attempt on Trump on an official website: https://youtube.com/watch?v=wqcrK-3207g
That's not to say it couldn't be misinformation or some neocon plot. But Trump was Grand Marshal of the Salute to Israel, he's bemoaned how Israel's control over the US Congress was lost in recent years, there are many good reasons for Iran to hate this man. He took a lot of money from Adelson who was basically a single-issue Israel partisan. Adelson's dead today but his wife still gave Trump about 100 million this election, she likely shares her husband's priorities. The Adelson's keep a much lower profile than Musk but one imagines they'll get what they pay for.
RandomRanger
Just build nuclear plants!
8mo ago
This is tangential to my main point. Nate's beliefs about the world, if expressed, would've cost him a lot of money. There are probably large numbers of people who trusted Nate's modelling and lost money thinking 'oh well if Nate says it's 50/50 then I can profit from Polymarket being 70/30'.
I think Nate is trying to claw back lost prestige. It reeks of cope to say 'my model exactly predicted the most likely election map' when his model gave it a 6% chance of happening. He struggles to even say what he wanted to happen on election day from the perspective of 'what makes my model look good'. If you're not producing a useful, falsifiable prediction then what is the point?
The important thing is getting it right. I want models that give alpha. If I'm going to pay for a substack, I'd rather pay for someone who got it right - Dominic Cummings who said confidently that Kamala would lose, based on his special polling methods. He actually predicted a Trump victory vs Kamala 311-227 in 2023. He foresaw that Biden would likely become obviously senile, that the Dems needed a better candidate.
RandomRanger
Just build nuclear plants!
8mo ago
He says so and that Nate later refused to sign.
RandomRanger
Just build nuclear plants!
8mo ago
"My model produces unhelpful outputs because it has bad inputs" is still only an excuse at the end of the day. Nate is a pretty influential guy, famous, respected by many. Why doesn't he have six figures to spend on his own poll and make his model better? Do none of his rich friends trust him enough to invest in him?
RandomRanger
Just build nuclear plants!
8mo ago
If Nate put his money where his mouth was, he'd have lost $100,000. He talks the talk (after it's decided) but doesn't walk the walk when it actually means anything.
RandomRanger
Just build nuclear plants!
8mo ago
If you're measuring the Russian economy in current USD, then sure there was a GDP per capita decline in 2023. But the Russian economy also apparently shrank about 40% from 2013 to 2016, an economic apocalypse comparable only to the 1990s. Exchange-rate games don't really matter for this.
https://data.worldbank.org/indicator/NY.GDP.PCAP.KD.ZG?locations=AU
I see 0.6% for 2023 and that's not even real growth, that's just nominal growth. We've been in per-capita recession for some time now.
https://www.macrobusiness.com.au/2024/05/australias-per-capita-recession-worsens/
RandomRanger
Just build nuclear plants!
8mo ago
This makes me think of someone stuck on a very sticky wicket, trying to justify an argument that was fundamentally wrong. Of course there are facets of any sophisticated but wrong argument that are right. You can highlight the correct facets and minimize the wrong facets. You can pre-prepare reasons for why you might be wrong to conserve credibility.
Nate has the rhetorical skills to pull it off. But it still feels very slimy. The 90 IQ twitter pleb mocking him with '60,000 simulations and all you conclude is that it's a coin flip?' may not be that numerically literate. But he has hit on a certain kind of wisdom. The election wasn't a 50/50 or a dice-roll. It was one way or another. With superior knowledge you could've called it in Trump's favour. Maybe only Bezos and various Lords of the Algorithms, French Gamblers and Masters of Unseen Powers knew or suspected - but there was knowledge to be had.
I prefer prediction models that make money before the outcome is decided, not ones that have to be justified retroactively. Nate wasn't heralding before the election that this 6% was the modal outcome, it wasn't really useful information.
RandomRanger
Just build nuclear plants!
8mo ago
despite signs that the Russian economy could well collapse within a year.
What signs?
According to the World Bank, Russia is now a high-income country. Real GDP per capita growth was at 3.6%! If an Australian politician could deliver that kind of growth, they'd be heralded as a living god and probably get Putin-level approval ratings (as opposed to negative approval ratings).
https://carnegieendowment.org/russia-eurasia/politika/2024/05/russia-war-income?lang=en
Even the Carnegie Endowment is struggling to find much bad to say about Russian wages growth. If Biden had delivered positive real wages growth over his term, I think he would still be in office today. Just look at the chart on page 25. Apparently the crushing impact of Western sanctions in 2022 was less harmful to the Russian worker than whatever was going on in America (or the UK, Germany, Australia...) with inflation. And in 2023 Russia left the US in the dust in real wages.
China's struggling, failing economy was massively outperforming the vibrant, dynamic US economy in 2022 and 2023, presumably it's still doing so. Real wages, real GDP per capita are rising much faster in China and Russia. They're rising from a lower basis level but are rising fast nonetheless. Yet all we see in newspapers and television is stories of disaster, stagnation and decline over there.
I dunno about the monitor lizard but there sure does seem to be a lot of it: https://www.indiatoday.in/india/story/almost-5-lakh-animals-became-victims-of-crimes-in-last-10-years-in-india-report-1770190-2021-02-17
Pakistan certainly doesn't cover itself in glory either, on the rape front.
More options
Context Copy link