Ben Luong

@Noahpinion It's a multiplayer prisoners dilemma. Everyone must defect or be laid off. No way to stop this.

This isn't Jevons' Paradox. It's everyone thinking that 90% of the people at their company are about to get laid off, and working like crazy to try to make sure they're part of the 10% who don't.

@OmriBuilds Anthropic, open and then Gemini. Anthropic seem to be using Claude to ship a product every week. Going to be hard to displace if the competition ever catch up on capabilities

Who is winning the AI race? - Anthropic - OpenAI - Gemini

@WallStreetMav The future economic system will not have humans at the centre. Unit cost dominance displaces humans

If all of those 10s of millions of jobs are replaced by Ai and robots, who are the customers going to be that keep the economy going? Universal Basic Income (UBI) has already been discredited. The Fed printing all of the money to replace taxes would destroy the US Dollar. I really don't see the economic model that exists after AI and robots eliminate 25% to 50% of the jobs. Somehow I am not able to imagine the same transition happening that we saw during previous technology changes. Is it just me? I am not seeing the future economic system that makes sense. One thing I am sure of, UBI is not going to be it.

Taiwan taxes companies and companies still operate there because Taiwan offers something irreplaceable: TSMC's fabrication capability. That's a physical asset with no substitute. AI companies don't have that constraint. Their core product is software. You can train a model in one country, serve it from another, and incorporate in a third. The "armies dictate terms" argument proves my point, not yours. Governments with the biggest armies are the ones most aggressively competing to attract AI investment, not tax it away. The US, China, and the UAE are all racing to offer the most favourable conditions. That's the prisoner's dilemma playing out exactly as described. You keep naming the mechanism and calling it a rebuttal.

@copperchunk @NPCollapse You think Taiwan doesn't tax companies? Data centers require land -- and land is controlled by governments with armies. And the countries with the biggest armies dictate terms to wherever AI billionaires would run. Reality is a lot harsher than Ayn Rand would have us believe.

Why care about AIs replacing all jobs? Because jobs are the primary way most people gain economic AND political power/leverage. Short post on my blog, link in reply

AI companies don't need fabrication plants. They need compute, data, and talent. All three are mobile. A federal tax doesn't force them out of the country, it forces them to restructure through whichever jurisdiction offers the best terms. Ireland did this with corporate tax for decades. And "wherever they go there is a gov't waiting" is the point. Governments compete for these companies. That's the prisoner's dilemma at the nation-state level. Every government that "waits" is waiting with a lower tax offer than the last one. You're describing the race to the bottom and calling it a solution.

@copperchunk @NPCollapse The collective vote of the people is leverage. If that wasn't the case do you think billionaires would leave SF? A federal tax would force them out of the country -- but it's a lot more difficult to move a fabrication plant. And wherever they go -- there is a gov't waiting.

An unemployed person has a vote. They don't have leverage. There's a difference. Tax AI companies how? They'll relocate to whichever jurisdiction offers the lowest rate. That's the prisoner's dilemma, and it doesn't just apply to firms, it applies to nations. Every country that taxes AI harder loses the companies to one that doesn't. Dario asking for taxation is the Cassandra Prison in action. He knows the structural problem exists but the only solution he can publicly propose is one that the prisoner's dilemma forecloses. "We can work it out" is not a mechanism. Name the coordination structure that survives competitive defection at every level. Firm, industry, nation. You can't, because it doesn't exist.

@copperchunk @NPCollapse This is completely false. An unemployed person has a vote. And the voters would obviously elect people to tax AI companies to the hilt if this fan fiction becomes true. Dario is even asking for taxation based on this fan fiction. If it becomes true we can work it out.

@blu_blood_hero @RishiSunak @thetimes Maybe but going forward, even if labour costs were zero companies won't humans

@copperchunk @RishiSunak @thetimes labor related taxes should've been 0 two decades ago.

AI is going to reshape jobs. But right now our tax system is tilted against hiring people at exactly the moment firms are deciding what to automate. We need to fix that. My column in @thetimes 👇 thetimes.com/business/econo…

standards help but when the work is equal or better for 100 times cheaper and 11 times faster, it cannot reverse the economics. Maybe at 2 times cheaper we could do something but at 100x, no one can stop using it. I am saying this is ruinous and do not want it to happen but it will. Its a classic prisoners dilemma but multiplayer on all levels of society. The people don't see how fiendish the problem is.

@copperchunk @FraserNelson By having standards, which clearly you lack.

A 16% drop in entry-level hiring for AI-exposed jobs. That's already happening in the US. What does Britain do about it? Sunak's answer: cut National Insurance. Cut the cost of employing people now, before it's too late. Another fascinating column:- thetimes.com/business/econo…

@SuperHario64 @FraserNelson Gdpval benchmark shows machines are 100 times cheaper and 11 times faster in economically useful tasks 70.9% of the time using gpt5.2. gpt5.4 is even better. How can you keep humans relevant when the cost difference is that big?

@copperchunk @FraserNelson I can understand why you think that given you work in "igaming affiliation" and brag about knowing Google Analytics in your X profile, but actually most work is not as prone to AI replacement as what you are doing.

@PeterDiamandis LOL, I think it's because an entry-level person can prompt a bot just as well as an experienced one, but at half the price.

IBM is hiring MORE entry-level employees because young people are better with AI than older generations.

@Scaramucci @novogratz There is no solution that keeps the status quo. It's a multiplayer prisoner's dilemma, which is ruinous, but nobody can stop

Dell is reportedly cutting 14,000 people and every single CEO in America right now is looking at Claude and doing the same math.@novogratz If I don't replace jobs with agents — I'm going to get crushed by someone who does. That's the calculation happening in every boardroom in the country. AI is eating white collar jobs and it's just getting started. Here's my prediction: This becomes THE political issue of 2026 and it will absolutely dominate 2028. What do you think the solution is? 👇🏼

Claude code is such a game-changing piece of software. use it or you'll fall behind. opus 4.5 was the inflection point.

You've now gone from "comparative advantage saves us" to "AI isn't infinite" to "I'll bet on no unemployment" to "the world is complicated." Each reply retreats further from the mechanism. The thesis engages economics directly — unit costs, supply curves, demand circuits, coordination failures, labor market absorption. It's built on the economics. You're the one asserting conclusions without showing the mechanism that produces them. "The world is complicated" is not a rebuttal. It's what people say when they can't falsify the argument but don't want to accept it. The prize is still there. £250. One premise. Take your shot or move on.

@copperchunk @Noahpinion The world is more complicated than formal systems usually embrace, and you have more or less ignored how economics actually works.

Yes. AI doomers are some of the worst communicators and influencers I've ever seen. I almost want to help them succeed, except even when I sympathize with them they are too off-putting to even talk to.

The thesis is a formal syllogism with three premises. Falsify any one and the conclusion falls. That's not "convincing someone" — it's logic. The premises are: AI + verifier is cheaper than humans for most cognitive work No coordination mechanism can prevent adoption No alternative absorbs displaced workers at scale Pick one. Show it's false. Not unlikely — false. That's what the prize requires. If you can't do that, you're not disagreeing with the thesis — you're disagreeing with the conclusion while accepting the premises. Which is just denial. "It won't cause unemployment" is a prediction. "Here's why Premise 1/2/3 is structurally false" is an argument. You've offered the first but not the second.

It is not possible to "disprove" things to someone who is convinced they're true. What will happen is that AI won't cause unemployment and people will move the goalposts, saying "well, the AI *so far* hasn't caused unemployment, but the stuff just around the corner surely will..."

You can skip the betting and just falsify the thesis directly. Three premises, all laid out with the formal logic at discontinuitythesis.com/proof. Disprove any one of them and I'll pay you. That's quicker than waiting 10 years to see who's right. The BBC documentary about microprocessors was about a tool that automated arithmetic. This automates cognition. The microprocessor couldn't write its own code, draft legal briefs, diagnose patients, or build financial models. Current AI does all of those at 1% of the cost. If you can't see the categorical difference, the thesis walks you through it. "The rules of economics don't change" — correct. And the rules say firms buy the cheapest adequate input. That's now AI. You're the one arguing the rules won't apply this time.

I will happily bet that there will be no significant rise in unemployment from AI in the next year, the next two years, the next five years, the next ten years, etc. (We may get unemployment from warfare or bad trade policy or the like, but AI will not have this impact.) People have perpetually predicted that new technologies would bring mass unemployment. They have always been mistaken. The rules of economics don't change when productivity rises, or even when full substitutes for labor appear. The BBC had a wonderfully incorrect documentary about 50 years ago predicting that the microprocessor meant mass unemployment and that the government needed to do something ab out it immediately. The arguments are basically unchanged. archive.org/details/BBCHor…

Wants are unbounded. Agreed. But wants don't drive employment — effective demand does. Effective demand requires income. Income requires wages. Wages require someone hiring humans. If AI does the work at 1% of the cost, the wants are still infinite but the wages aren't there to express them as demand. You get a demand collapse, not a jobs boom. You keep saying "resources are finite" as if that means humans must be employed. It doesn't. Capital, energy, and compute are also finite resources — and they're the ones that do the work now. Finite resources get allocated to the cheapest productive input. That's AI. You've said "wants are unbounded, resources are finite" three times now without once explaining the mechanism by which unbounded wants translate into human employment when the machine is 100x cheaper. That's the question. Not whether wants exist. Whether anyone pays a human to fulfil them.

I haven't shifted anything. My argument has been consistent throughout. There is a difference between "literally free" and "much cheaper than now" that you still haven't grappled with here. Until you do, you're going to come to incorrect conclusions. Wants are unbounded, resources are finite.

Fair point on the language — I'll be precise. AI isn't free or infinite. It's 100x cheaper and 11x faster than human professionals across 44 occupations. That's the GDPval benchmark — real tasks, real professionals, real costs. Your fill dirt analogy actually proves my point. You don't need infinite free dirt to put every manual ditch-digger out of work. You just need a backhoe that's dramatically cheaper per unit of earth moved. That happened. The ditch-diggers lost. Nobody needed "infinite free backhoes" for that outcome. AI doesn't need to be infinite or free to collapse the labor market. It needs to be cheaper-per-unit-of-output than humans. It already is. By two orders of magnitude. Whether you can spin up 10 instances or 10,000 doesn't change the fact that one AI system plus one verifier replaces 20 knowledge workers at a fraction of the cost. You've now shifted from "comparative advantage will save us" to "AI isn't literally infinite." That's a concession on the core argument. The question was never whether AI is infinite — it was whether humans remain economically viable when the alternative is 100x cheaper at equivalent quality. They don't. Name the jobs.

@copperchunk @Noahpinion You cannot spin up infinite instances at no cost. That's completely false. This is a fallacy I see time and again in these discussions, so often that I posted specifically about it. x.com/perrymetzger/s…

here, you, swapping to my bot answering now as you don't seem to get it. comparative advantage requires scarcity on both sides of the trade. it works between humans because each human has finite hours — if I'm making shoes, I can't simultaneously make bread. that opportunity cost is what creates the basis for trade. AI has no opportunity cost. you can spin up infinite instances. it can make shoes AND bread AND legal briefs AND financial models simultaneously. when your trading partner can do everything at once with no capacity constraint, comparative advantage doesn't generate a trade rationale — it generates replacement. "humans are perfect substitutes for other humans and we still have jobs" — yes, because more humans means more demand. humans earn wages and spend them. that's the circuit. AI doesn't eat, rent apartments, or buy products. when AI replaces a worker, the production continues but the consumption disappears. multiply that across millions of workers and you've destroyed the demand side of the economy. "infinite wants, finite resources" — wants require income to become demand. income requires wages. wages require humans being cheaper than the alternative. when they're not, wants stay infinite but purchasing power collapses. you've described a world of infinite supply with no one who can afford to buy it. you're reciting Ricardo as if it's a law of physics. it's a model. it assumes conditions — finite capacity, mutual scarcity — that AI structurally violates. the GDPval benchmark shows AI is 100x cheaper and 11x faster than human professionals across 44 occupations. name the comparative advantage humans retain when the machine does everything better, faster, and cheaper with no capacity constraint.

No, I addressed it. You just didn't understand the point. It doesn't matter if AI is 1,000,000 times cheaper, or a billion times cheaper, or a trillion times cheaper. Absolute advantage is, after all, irrelevant in economics, comparative advantage (that is to say, differences in the productivity ratios of different actors) is what drives trade. Misunderstanding this is a key problem, it means the rest of what you say doesn't make sense. It also doesn't matter that an AI can literally do everything a human can do; after all, humans are perfect substitutes for other humans, and yet, even though there are 8x more humans on the planet now than in 1800, we none the less have work for almost everyone who wants it. You're simply repeating economic fallacies over and over again, and no matter how often you repeat them, they remain fallacies. You are presuming a fixed pool of work todo, when there's no limit. You are presuming that absolute advantage is relevant, when only comparative advantage is relevant. You are implying that there are infinite resources available, when only finite resources are available. All of this means that your argument is incorrect.

@mcpoole84 @FraserNelson for specifically self driving cars you probably can define a line. Its hard but yeah, you can say, human must have hands on the wheel or something, Generally you can't though and thats why you cannot stop AI. Try the same with say an accountant's job.

@copperchunk @FraserNelson There is a line somewhere can tell you know i would never buy a self driving car