This is the internet, so I cannot tell at all whether you're being sarcastic or not. In my view, what LLMs should get us to reconsider isn't whether there is more to intelligence, but whether there is more to language. It's the latter which I underestimated.
Labor costs are orders of magnitude more than costs added by environmental or labor standards.
The average hourly rate in the US is $42.50. In Vietnam is is $6.30, in the Philippines it is $5.40[1]
These aren't people living in poverty - they live well.
OTOH environmental legislation costs are harder to come by, but [2] estimates around 10% in a number of case studies. To be specific, that would mean US cost would be $4 less - that's still around 6 times the cost of Vietnam
That paper discusses only the effect of increasing energy costs, and seems highly dubious as well. Their conclusion is that increases in energy prices would have a small effect that could even be positive. We got to see what happens after increasing energy prices since 2022. The effects have been neither modest nor positive. They studied only modest changes from the past, but they were speaking of energy costs shifts on a scale aiming to shift Earth's climate, which is not modest. The cost is going to be much more than 10%.
Also the US economy is increasingly polarized so averages mislead. The median hourly is a bit under $21. [1] You also need to account for what happened to the economy after we start shipping these affected jobs overseas. That's an entire segment of the economy which is going to influence wages, consumption, and other distributions. I agree wages are always going to be a major factor, but we can also afford a premium because shipping stuff half way around the world isn't free, the broader economic impact of having less control over your supply line is itself also a significant cost, and so on.
It worked for Australia (which has a fairly similar economy to Canada).
It has led to a "Dutch disease" reliance on the resources sector in Australia, but that isn't radically different to the Australian economy before (mostly proportionally more coal & iron ore exports, less wool and grain)
Accusations of Chinese dumping are cheap and easy to make, but the evidence is pretty mixed.
Most actual cases have been of raw materials (eg steel, yarn) and industrial chemicals.
I think this is a long way from the alleged "tsunami of Chinese goods".
It's also worth noting that many Chinese manufactured goods would mostly substitute for US manufactured goods (eg cars). That's actually the whole point.
The thing people tend to handwave about China's manufacturing prowess is that they have spent the last 25-ish years getting really good at 'making things to a spec' at a cost in-line with their labor market, being bolstered by their government's approach at Economy planning (i.e. subsidizing certain industries, required partnership agreements when foreign companies want to use that labor market to build things).
The quality of the product you get, like always, is about what people are willing to pay. The big difference is that the current cost of labor/material inputs there, is low enough that you can get the same or often better quality for less.
I wouldn't call that dumping. And to be clear, yeah they've done dumping in some cases.
But when we look at what's happened in the last 8 months or so around various AI SaaS-ish offerings, one could argue that plenty of US companies have done 'dumping with extra steps'. (And that's not even the first time necessarily, e.x. the DoorDash Pizza Arbitrage thought experiment...)
> It's also worth noting that many Chinese manufactured goods would mostly substitute for US manufactured goods (eg cars).
Heck, GM was (although they are moving it due to the Tariffs AFAIR) building the Buick Envision (Compact CUV) in China and shipping it to the US for sale. GM Announced in January that they'd move production to the US by 2028.
Similar for The Lincoln Nautilus (Midsize CUV), The current Gen is built in China and they only in the last month announced that they would move production back to the US... by 2030.
Garlic dumping. Those little socks of 4 or 5 bulbs sold for a dollar-something are from China (notice the tops are chopped off so they cant sprout.) I buy domestic garlic because it's fucking stupid to import something that's so easy to grow.
It's not "so bad in this case". The domestic version is called "predatory pricing" and is illegal. In practice, it's easier to raise anti-dumping complaints than predatory pricing cases, both because the producers don't vote in your country and because the complaint doesn't have to be proven in a court of law.
TBH I agree with you, even mentioned it in my other reply.
Modern China is not like the Soviet Union; it is much closer to a capitalistic society, but instead of private equity there is a (heavier than even the most 'socialist' of EU nations) government planning push.
In some ways, one could argue they're just taking lessons learned from all of the current systems and applying them aggressively in their centralized planning.
As someone who has lived in/around the heart of the 'Arsenal of Democracy' for my whole life, gotta say they are taking better lessons learned than the US has.
I'm sorry our educational system has not made the differences between bad faith companies and dictatorship clearer.
Let me put it this way: American companies sell you a bad car and lay you off to give a CEO a bonus.
Chinese dictatorships put you into reeducation camps if you're a Muslim dissident or unapproved member of a religion. China tells the Catholic Church who to appoint as bishops and the Pope spends his time criticizing...anyone but China.
China pumps metrics tons of drug precursors onto the black market which accidentally find their way into Mexican drug labs and...are imported into Detroit.
China backs our national enemies because they are friends to anyone who opposes us. China funds Russia. China buys Iranian oil.
You just described it yourself. American companies will build a Cybertruck, blame their customers, cancel a $25,000 EV, and then lay off employees for Elon's ketamine/AIPAC shill fund.
China has their own Abu Ghraib moments swept under the carpet, but their workers accept it for the same reason America ignored Seymour Hersh. Their middle class is growing, the Chinese economy has vast surpluses and market leverage that only America has ever enjoyed. Their asymmetric warfare stealing IP, from Texas Instruments to ARM to Lockheed Martin went unpunished. The Chinese intelligence agency, through Salt Typhoon, probably has more blackmail on American politicians than the Mossad does. From a Chinese perspective, the war is won. They're supporting the winning side in Ukraine since both US administrations rebuffed full NATO intervention. They're supporting the winning side in Iran after the IRGC proved they can disable planes, drones and ships with Chinese hardware.
When the United States calls Canada a "51st state" or renames border territories to manufacture leverage, it's not helping China lose. They're undermining the trade agreements that all of North America relies on, and strengthening China's case as a reliable alternative to America's fickle export restrictions. In that sense, humanitarian outrage is almost an advantage; why would Pakistan, Iran or Russia worry about ethical reprisal from China if they're equally amoral?
> a large language model trained on chess commentary as well as being trained on millions of chess games would outperform stockfish which is only trained on millions of chess games
Not really, if anything it's closer to the opposite. The Bitter Lesson essay literally has this as an example:
> These researchers wanted methods based on human input to win and were disappointed when they did not.[1]
and
> Enormous initial efforts went into avoiding search by taking advantage of human knowledge, or of the special features of the game, but all those efforts proved irrelevant, or worse, once search was applied effectively at scale[1]
The actual bitter lesson is this:
> breakthrough progress eventually arrives by an opposing approach based on scaling computation by search and learning. The eventual success is tinged with bitterness, and often incompletely digested, because it is success over a favored, human-centric approach.[1]
Applying to the "LLMs-for-chess" example the bitter lesson approach would be to put many, many more games into the LLM.
> These researchers wanted methods based on human input to win and were disappointed when they did not.[1]
This was/is basically a strawman though. Like maybe "human input winning" was desirable for chess masters but for computer science wonks? Not the point or the disappoint. It's always neats and scruffies fighting about using some kind of recognizable method (logic) instead of magic (ML).
> breakthrough progress eventually arrives by an opposing approach based on scaling computation by search and learning.
More to OP's point I think: nowadays when someone wants to beat you over the head with the bitter lesson, they aren't as careful to include learning and search. They want to say learning leads to intuition (magic) whereby we can avoid work (logic/search), and maybe argue or assume from there that neats and scruffies is settled. TBF, something like reasoning in latent space does resemble intuition!
But the real lesson is confirmed every time we bother to check, and not very bitter for anyone. Search/learning/logic are ALL always necessary on any sufficiently difficult problems, and hybrids that interleave always outperform everything else. Stockfish being the example in this thread that different camps of absolutists would like to claim, but also all the MCTS examples, evolving examples, and new hybrids all the time. My favorite lately: https://arxiv.org/pdf/2511.08983
There's two different goals to AI research - one was to get results - a chess engine thst wins, etc. But the other goal (which seems to have been abandoned in the deep learning era) was to use AI to help understand how human minds work. A chess engine modeled after human grandmasters is much more interesting in that regard than either a min-max algorithm like beat Kasparov or modern deep learning engines.
> This was/is basically a strawman though. Like maybe "human input winning" was desirable for chess masters but for computer science wonks?
Oh no!
The whole field was full of people whose entire career was built around the idea of developing smart priors.
To quote Wikipedia:
> For computer vision in particular, much progress came from manual feature engineering, such as SIFT features, SURF features, HoG features, bags of visual words, etc. It was a minority position in computer vision that features can be learned directly from data
This undersells the change though! David Lowe's reputation as the best image researcher in the world was based on his SIFT patent[1]
This approach worked until 30 September 2012.
That was a bitter day for many, many computer science researchers.
Is being wrong/ignorant about whether/how something can be automated the same as having a preference for doing it manually? Maybe so if it's your patent, your thesis I guess..
But as it relates to more/less magic, maybe the more modern lens on this is e.g. https://arxiv.org/html/2505.11581v1 . Is manual feature-engineering more like what you'd evolve, or more like what you'd get from SGD ? Feasibility and performance is always a question, there are others like what is robust, stable, adaptable, predictable, explainable. Maybe the manual-features people were interested in something besides the manual part? Maybe the story isn't so simple, and maybe it's not finished yet.
I checked this, and Claude says "it's a real term of art from Michael Feathers' Working Effectively with Legacy Code (2004), where he defines it as a place you can alter a program's behavior without editing in that place."
While I agree the model doesn't have insight into how it was trained I do think the history of the term itself is interesting.
"shipped" was a pretty common term before AI though. It does show as 17x more common on github in this dataset but it was used a lot more in product management than in PRs previously.
To be clear, I am referring to software and shipping as a developer who has finished, more or less, their app. Not that I am against its use, I just never really heard that word being used before, and now I hear it all the time. Thus, what the real issue I have with it is, is because my brain has associated its use with a torrent of crap, ie. AI slop. I hear the word used, and I already have bad feeling about it. And dialed in even further, my issue is not with AI just because, it's simply the association I have of this technology taking something that many of us spent real blood, sweat, and tears earning the right to represent, to it now being automated, and represented (badly) by novices with no respect for the spirit of it all. But, this is not a new or unique sentiment nor are my grievances with it. All hope is not lost, though. I know I need to adapt, and I know that like the world, I need to change too. But no one ever said I have to like it :-)
Context aside, I still gotta stand by the 17x indicator or see something that deflates that finding, and if I do see that, I guess all of the results are in question then? Based on the specific use case mentioned at the beginning there, coupled with personal experience (I totally could just be more sheltered than I give myself credit for), and the 17x, I gotta disagree with you guys and say that the finding of the link above is accurate in my experience being the *
imo its a change in definition. Previously to "ship" was a one-time thing that you did for big, highly verified and checked stuff. Physically shipping your software to your user tends to be done after its a product.
Nobody was "shipping" a PR, or a jira ticket previously
I can get on board with this...so you are suggesting the usage isn't new, but the frequency changed can be attributed to the scope change, so the frequency is associated with all of the new cases that the new scope encompasses, which in hind sight I'm surprised that didn't occur to me already, it's pretty obvious. I'm probably a bit hampered by the preexisting dislike for the aforementioned association though. Is there a reliable way to determine the difference in frequency versus the difference in usage, or is the frequency multiplier already based on usage? Or maybe I am trying to ask if there is any way to get a sense of what the 17x is in relation to, ie. what is it 17x more than...if someone said the word "example" was used 17x more than it was six weeks ago, I would know that baseline by experience, but for something like "shipped" it's not as straight forward.
ultracode is different to ultrathink (because of course it is...)
ultrathink was "throw more tokens" (it was a Claude feature that has gone away[1])
ultracode is "throw more tokens AND write code that creates and orchastrates subagents": https://code.claude.com/docs/en/workflows
[1] https://claudelog.com/faqs/what-is-ultrathink/ claims it has returned, but that might not be accurate
reply