Hacker News Reader

I feel like too many people have a binary vision of the world, i.e. you're either "pro AI" or "anti AI".

Ed Zitron doesn't say that AI doesn't work, that its impact on the world will not increase. He's basically saying that (1) the current data center investments are based on unrealistic revenue projections and (2) hyperscalers are using accounting tricks to move around "money" in a circular way to make it look like more money is already flowing to AI.

You can believe all of that and still believe that AI (as in "LLM based services") work, are useful and will probably see their usefulness grow even more. Just not in the magnitude necessary to make the current investments make sense.

After looking at some of his recent interviews, he's acknowledged his previous predictions that 1. He was naive enough to think the investment won't keep coming despite lack of profitability and 2. He acknowledges that AI in it's current form is a billion dollar industry, not a trillion dollar one, but also not zero.

Definitely less extreme than he used to be.

If you ever listen to more than a couple of Zitron interviews, it's clear that he's not the "reasonable centrist" viewpoint on AI. He's the doom-and-gloom guy.

Maybe there is someone out there positive on AI itself and calling out reasonable objections to some of the extreme things. But that's not Zitron.

The center is not always the reasonable viewpoint. And sometimes the doom-and-gloom guy is correct.

Right, but as the article linked lays out in detail, Zitron has been repeatedly and consistently objectively incorrect.

There are plenty of doom-and-gloom people to go around. Even if Ed Zitron is wrong about a lot of things, the doom-and-gloom part may still be accurate.

Personally I am way to the left of Ed Zitron, I actually think he isn’t doom and gloom enough.

My personal prediction is that whatever happens the people with all the money will find a way to spin the narrative in their favor and history will judge people like me (and Ed Zitron) as having been wrong.

Not this time.

As someone who's strongly anti-AI, i find Zitron to be absolutely too much pro-AI for my taste. He defends AI that's not LLMs. He barely touches on ecological impacts of AI, and barely mentions the enshittification of things. He's only really against the LLM craze, OpenAI/Anthropic over-evaluation (because Deepseek and local models will do the same for nearly free), and the crazy speculative bubble around datacenters.

If you find him very negative about AI, and i find him very positive about AI, he might just be this reasonable centrist.

Out of curiosity, what would you like to hear him saying about non-LLM AI?

There's plenty of arguments to be had. I guess there could be some useful applications of non-LLM AI (eg. translation tools), but quality output is still a concern. Environmental impact is also still a concern. If output is of sufficient quality, who reaps the benefits is a concern ; if not, who pays the price of shittier products is a concern.

Whether LLM or non-LLM tech, the question of surveillance capitalism and how models need to be fed swaths of privacy-invading data (eg. self-driving cars filming everything all the time) is also a concern.

Overall, i think the ecological concern is still the biggest one. We are at the start of a major climate crisis, and any self-preserving measures were already hard fought by industry. I fear that any form of AI investment will lead (because it's already leading) to destroying any hope of saving an inhabitable planet.

To play devil's advocate - unless you were to take the position of declaring bankruptcy on the possibility that a complex society of competing actors can agree on climate change - and therefore this being an unsolvable problem that you need AI to solve, as humans can't handle the complexity.

I find a lot of the debate on either climate change or AI collapses if you point out that "we shouldn't do this, we should ALL just do this" is an extremely unrealistic position in an international complex ecosystem of competing actors.

> "we shouldn't do this, we should ALL just do this" is an extremely unrealistic position

No it's not. It's only if people like you are so defeatist about it. Look what happened with CFC gazes, leaded paints and asbestos. Sure we still have more problems to face but some problems are so serious that we can in fact gather consensus.

> an unsolvable problem that you need AI to solve

That is completely wrong on so many level. I understand you're playing the devil's advocate, but do you understand this is a messianic-level delusion that's made up by religious psychopaths so people buy into their products without thinking twice? If you don't have hope, the only rational solution is not to give in to AI, it's to Luigi Mangione or Unabomber your way to a better world. Personally, i choose hope: we can solve this problem (and every other, who strangely have more or less the same culprits).

I don't know - examples of people coordinating where there's cost involved are stunningly rare in the history of society. Coordinating when there's cost and information asymmetry - even more so.

I am probably on the pessimistic end of the spectrum, and personally I don't believe AI can solve it, but I can see that line of reasoning if what you believe the culprit is - is the sheer complexity of the coordination needed.

Really only large language models require huge swaths of training data and extremely expensive training/inference. Other “AI”, like basic neural networks or something like path finding, is basically free. It’s just an algorithm, a small one that runs on computers.

The term AI is very overloaded. When people say AI, they really mean LLMs. AI like old Siri has basically no cost. It’s not environmentally damaging, as it runs on device. It’s not privacy gore, because it runs on device.

Or take, for example, lane keep assist systems on vehicles. This doesn’t require a huge data center or anything, it’s basic computer vision and a small on-vehicle model.

Yes AI has a “cost”, but for virtually all non-LLM AI the ecological cost is the same as any other application or algorithm. Siri doesn’t burn more energy than instagram.

> This doesn’t require a huge data center or anything, it’s basic computer vision and a small on-vehicle model.

Yes, but it still has a strong ecological impact, and for what benefit? I only ever hear people complaining about electronics/assist in cars, and recalls due to electronic faults is about 50% of car recalls nowadays.

The problem is not that this compute is overall so much compute, it's that it's used to replace "no compute", and that in the ecological lifecycle of electronics production is the biggest problem (not sustained use). A modern car has hundreds of microcontrollers and probably as many sensors… Although some of these features are technically not machine learning, i think it's part of the same problem.

I wish we could have better/safer/cheaper cars over time, that can be more easily maintained. Instead we have shitter and more expensive cars that are less rapairable than ever before.

Well that’s what I’m saying - a basic machine learning system is no more ecological cost than an algorithm. Because it just is an algorithm.

Computerization of cars is orthogonal, it’s not an AI problem. Things like engine control for fuel efficiency isn’t AI, it’s just an algorithm. Cruise control is just an algorithm. The cost is the same as any other software.

Now LLMs DO have a new cost. You really require super computers to run LLMs at scale, that much is true. But that’s very distinct from machine learning, it’s a different beast.

Nobody is complaining about, like, monster AI in The Binding of Isaac and how much ecological impact it has. Because it doesn’t. Well… that’s AI, and 10 years ago that’s the only AI people really talked about. Now AI really means LLMs. The ecological impact is from LLMs.

> Ed Zitron doesn't say that AI doesn't work, that its impact on the world will not increase

He's said variants of both of these things before! That's the point OP makes too, people say "He doesn't say that" or "But what about this" when he's said a million, sometimes contradictory things.

Ed's whole cornerstone thesis is that "AI doesn't work." Like, if I had to sum him up in a short sentence, that's what it'd be

that or "Angry but doesn't know much"

I think when cornered, he does say it works for coding and some other tasks. But to be honest it also makes a lot of bugs when writing code.

I don't agree with everything he thinks, but I think I agree with him more than I do Sam Altman or Jensen.

I’d be interested in a real accounting of what he said, rather than what people say they’ve heard.

If he says AI is changing work in some places but not-at-all in most… that’s defensible. Or at least we can look at that. Did he say it about the llm offerings from one or two providers or about (broad gesture) all AI efforts, everywhere? Etc.

Certainly it’s changing sw development work. It’s less clear, most everywhere else.

No, the cornerstones of his thesis are:

- LLM’s potential and capability are dramatically overstated

- Business leaders are either driven by the Emperor’s New Clothes effect, or the legitimate fear of stock price or valuation drops, to admit it.

- The amount of money being dumped into this industry will never pay off.

He’s snide and condescending, to the detriment of his message. I can completely understand why people aren’t interested in listening to him. However, straw-manning his stances by oversimplifying or flat-out making uninformed assumptions about them isn’t a useful response to what he’s saying.

Frankly, this industry squirts out enough bullshit to smother an active volcano every single day. If you think Ed isn’t credible, I sure hope you dismiss the industry leaders just as readily.

Zitron’s superpower is to get reasonable people like you to turn his bullshit into prose.

He’s a scam artist. So is Altman. The correct answer is to not listen to either one, at least not on a regular basis and certainly not as a source of truth.

> straw-manning his stances by oversimplifying

He literally lies about numbers. The spreadsheet errors the article describes are either gross negligence or simple evidence of lying.

> No, the cornerstones of his thesis are: [...] The amount of money being dumped into this industry will never pay off.

That has only been a cornerstone of his thesis for the last 10 months or so, once Claude Code and Codex began making gobs of revenue. Before that the cornerstone of his thesis was "nobody will ever pay for AI" and then he silently pivoted, without acknowledging any error, once it was clear that was not true.

This is what bugs people about Zitron and why people aren't listening to him. He's has repeatedly been proven wrong, and every time it happens he just silently moves the goalposts and never admits any mistakes. It's fine-- actually, correct-- to write someone off once it's clear that every time they get disproven they'll just move to a new thesis instead of reflecting on what went wrong and what assumptions might be fundamentally incorrect about their thinking.

Do you know anyone whose understanding of the LLM business has remained static for the past 10 months? Would it be a positive trait?

if you amass 115k subscribers by basically being wrong along the way just about everything and the shift as to "not remain static" and then continue to be full of shit, that is one thing - another things is going from "hm, what a great toy this is" to "holy shit, I can take on 4 contracts at the same time and run 25 terminals of CC is parallel" :)

In less than 3 years we’ve completely redefined how we interact with computers, and Zitron’s saying that there’s no way that investment in that pays off. He moves the goalposts every time something new proves him wrong.

To me, that’s not only not a credible stance, it’s immensely stupid.

The only way anyone could logically take that stance in the face of all the evidence otherwise is if their entire livelihood depended on it.

Oh… wait.

You could make the monetary motivation argument against anyone involved— much moreso Altman, et. al.

What specifically about the criticism do you find stupid? That’s a very general dismissal of some pretty specific criticism.

To be fair he is nothing more than a talking head who historically has had some pretty negative AI pieces. Sure nothing is binary but more than ever it feels like there are talking heads on both sides that have some pretty extreme views.

> You can believe all of that

You can believe all of that (I do) and see that Zitron is a liar and an influencer.

> Ed Zitron doesn't say that AI doesn't work

Ive seen him say it does work pretty consistently? He also talks about other things that are more interesting, but he still says it doesnt work

If you watch his latest interviews you can see him talking about how he finds it useful for simple cases like in the Bloomberg terminal for generating code for queries, but it still has fundamental limitations for doing anything autonomously.

As someone who has used LLMs heavily at work and at home in various serious experiments, I agree. It still requires heavy babysitting and a lot of its limitations wrt context length are fundamental, not something that’s going to be easy to overcome.

I also agree but the autonomous part comes from coding. We build autonomous systems with code, so it AI is really good at writing code (it is), we increase automation. Not with the AI itself, but with what it produces.

Which is good, for many reasons:

- environmental costs go down. Algorithms are cheap, LLM inference and training is not.

- code is deterministic

- code can be read and understood by professionals

- code can be deployed trivially, whereas deploying an LLM costs huge infrastructure

So using an LLM to help write code is the best of both worlds. You get velocity increase from the LLM, with none of the hallucination or environmental downsides at deployment time.

From anything i read/watched of him, the shorthand is "it doesn't work". When taking more than a few words to explain, it means it's not replacing humans altogether, because the output is low-quality and needs human quality control. In the end, he's not arguing LLMs are not working (as in returning 500 in your browser), but that the promise of autonomous agents replacing humans for increased productivity is a lie and doesn't work.

I overall agree with his interpretation, but i'm afraid he might be wrong in the conclusion that noone is going to replace human labor with shittier agent slop. As long as we humans don't really have regulations and choice for proper quality of service, companies might just in fact enshittify everything with slop and make more profits while we suffer from lower-quality products and support.

re: substituting with cheaper models. I think some people here are oblivious to how much of Anthropic and OpenAI’s usage is artificial.

Take this one example of a user with 15 Codex subscriptions ($3k) generating $60k in API-equivalent usage per month: https://hraness.com/writing/my-girlfriend-asked-me-why-i-hav...

“there’s a once in a lifetime discount happening at the OpenAI Intelligence Depot, and I brought 15 shopping carts.”

We have to understand current usage with this behavior in mind, there are tens of thousands of people just like this author who are intentionally generating as much usage as possible on their subsidized plans because they feel compelled by some need to get free intelligence.

The only reason these expensive models are generating so much usage is because they are so heavily subsidized. Pragmatic users will shift to cheaper models which will hurt per token revenue, yes, but huge volumes of usage is going to just disappear because there isn’t the demand when it isn’t being subsidized. Less revenue per token and less tokens.

https://tokscale.ai/leaderboard a small sample of just 2k users have generated over $100m of API-equivalent usage while paying closer to just $1m.

https://aicharts.io/gpt-subsidy

It goes a layer up, too. Harvey (legal AI) recently tweeted that a single user query cost that firm 26k USD. This, while they have thousands of users and reportedly whole law firms paying zero. That difference is being subsidized by billions of VC dollars.

A bit of a collective action problem, but if everyone leverages this services as hard as they can without cost concern, better capital decisions will eventually be made. Like the Federal Reserve draining the M2 money supply, AI users must drain the capital supply of subsidized tokens. This pulls forward the future of "Does this tech have value at actual unsubsidized costs?" If it does, tremendous, if it doesn't, also a reasonable outcome to the grand experiment. The current pain comes from the valley of uncertainty we find ourselves in at the moment.

> That difference is being subsidized by billions of VC dollars.

"History never repeats itself, but it rhymes." -- Twain

Doordash and Pizza Arbitrage - https://news.ycombinator.com/item?id=23216852 - May 2020 (514 comments)

> I cut this deal with my neighborhood Italian restaurant! I texted the owner about being miffed they hadn’t told me they were on DoorDash. He replied. They aren’t. We compared pricing, and found the prices advertised are way off from what the restaurant charges. So I placed a $5,000 order to the neighbourhood homeless shelter. DoorDash paid him over $20,000, and I get free pasta for the rest of the year. (My neighbours have also partaken.) Glad to know it’s scaling. SoftBank has assembled a unique concentration of stupidity for itself.

Yes, there are good examples of this working, anti-competitive as it may be.

What if, in the DoorDash example, the pizza shop was itself venture backed and selling pizza at a loss to win customers and using this 20k revenue as a basis to raise money?

What if the same VCs were backing both and gearing up for an IPO?

Once circular funding scales, VCs become the Fed.

Why do everyone assume they are subsidized? When we seemingly have no idea what it costs? Maybe average subscription is breaking even and token spend is pretty much pure profit?

Case in point, claude code seems hell bent on increasing usage at all cost. Which makes sense in the growing phase (get people hooked) but it does not make sense given the hardware shortage. So, which is it?

Anthropic admitted last year to losing money on inference, it had negative margins. The margins have improved and are now positive but there’s still a significant cost. If plans aren’t being subsidized it would mean that the margin on inference is ~99%+ which would mean OpenAI and Anthropic should be wildly profitable but both are still losing money. So, it’s mathematically impossible that they’re not subsidizing plans.

The most widely accepted estimates (though I disagree with them) are that Anthropic’s margin on API inference is ~70% from which people extrapolate what their token usage would cost via the API and compare that to what their plan costs.

https://newsletter.semianalysis.com/p/anthropic-3q26-profit-...

(edit: better link https://newsletter.semianalysis.com/p/anthropic-growth-and-b...)

re: increasing usage with resets, it’s because they’ve overblown usage and need to show that usage is growing ahead of the IPO. They’re increasing usage on fixed price plans without increasing the cost, the only plausible explanation is they have unused capacity. If they were capacity constrained then the last thing they would do is give away more usage for free.

The last I saw with Claude was that the plans are subsidized somewhere around 10x. The usage of Claude and similar products (for people not paying the actual API token costs) would be lot less if they were paying 10x more per account/seat.

A lot of people are using Claude and ChatGPT for all kinds of minor things at work, and they probably wouldn't be if they were paying the true cost of the product. And this is all the while their work product is suffering because AI is not a great fit for a lot of use cases.

>Anthropic admitted last year to losing money on inference, it had negative margins.

No they did not. Dario has repeatedly stated if they stopped training new models, they would be very profitable.

When?

“The majority of the cost is inference, not necessarily the training of the model.”

https://youtu.be/7xij6SoCClI

> Why do everyone assume they are subsidized?

It's not an assumption when the companies hosting the models publicly announce that they're subsidizing costs.

Depends on whether the costs they have are inflated ten-fold due to hardware shortages.

Ignoring current state of shortages, the interesting part (for me at least) is how much a SOTA token costs on current hardware if you exclude the costs for paying for current TSMC shortage and without factoring in the cost of new datacenters being rushed and renting hardware from your competitors (who are also supply limited).

The "actual" cost is of course also relevant and interesting but it is another thing entirely.

Have they actually announced that they're charging less than the marginal cost of inference, or that the revenue they're taking in doesn't make up for the cost of training each newer and better model?

The former does not pass the smell test when there are random providers selling tokens for competitive open weight models.

You can compare the subscription cost to the API per token cost for similar queries.

That tells you that subscription is cheaper than tokens. Which ought to be a given.

But it does not tell you what and how much either is subsidized...

Given the cost of openweight frontier models like Kimi I’m fairly sure the token prices actually make some money.

theyre not being subsidized, and no ones using tokens just because.

Yeah personally I think he is right about a lot of the current state of affairs. I think he has a decent grasp of the situation despite being incredibly biased.

His predictions though? I don’t think I’ve ever read anything from him to get his read on how things will play out.

Just because one makes the wrong predictions does not mean that the data used to make them was wrong. A lot of people seem to dismiss his reporting because of his takeaways

> I am sure that what they're using it for can be substituted in large part by way cheaper models.

I saw a commit in the repo at work that edited a few config files and a dictionary.txt file. "co-authored by Fable"

I have colleagues who have replaced ctrl-f on a (long) document they already have open with prompting an LLM to find things for them.

thats me

It never ceases to amaze the lengths people will go to avoid using grep

None of this actually matters if this is a nuclear weapons-style arms race.

The data center build out will become a matter of national security, and will be backstopped by governments.

LLMs are a national security issue? Why is China releasing their models then?

The question is not whether you should have datacenters, or the most advanced chips, or the ability to build the most capacity. But do we need all this now? Will there be enough demand? People want to make profits form this thing, and what's being pointed out is that maybe there won't be enough demand to generate profits for all this investment.

I would guess there is probably a capability ceiling beyond which China will not open source their weights. Maybe there is a scenario where they do so or threaten to do so if the US is so far ahead or something as a deterrent.

>But do we need all this now?

Did people want to make profits from the nuclear arms race? I'm sure some did, but it was not really a requirement.

The bottom line is there is a theoretical capability of AI which is effectively a super weapon that can be deployed against your competitor. If you think that is a possibility, even with a low probability, you should probably have a giant AI-related build if you are the US or China.

This reads to me as incredible cope. Maybe not both of these companies but certainly one of them will be enormously valuable, and there will always value in the frontier models even if much of the practical usage can be done locally

Why is it “certainly” the case that one will be enormously valuable?

Neither appears to be on track to long-term profitability specifically once you take into account depreciation on CAPEX.

On top of this the “productivity gains” from AI across the board seem to be a very mixed bag. The pitch from these companies has been huge gains in productivity and automation and although there is anecdotal evidence some people are able to do that, the broader studies seem to show marginal gains in most cases.