"The underlying purpose of AI is to allow wealth to access skill while removing from the skilled the ability to access wealth". (comment on the discussion above: https://news.ycombinator.com/item?id=48334069)
It used to be only a few steps removed from actual value. Like an empty lot is worth an amount because when you plant trees it's worth an amount because when it has fully grown trees it's worth an amount because when you cut down the trees the timber is worth an amount. That's an acceptable function of finance, it allocates the value of a logging plantation between several different people who may own it at different times, so that one doesn't have to hold it for the full 20-year growth cycle to realize any value from it.
We needed to develop long-term value instruments to incentivize long-term planning. Bonds are necessary for large public projects, for instance. There are many complexities in deriving instruments from that basic concept. What makes you think a particular one shouldn't be acceptable and why?
>> something is valuable because some people think other people will pay more for it in the future, and not because it does useful things > This has been the definition of finance for hundreds of years. I don't know why it comes across here like this is a new phenomenon.
I don't think so. However, you can convince me by providing a reference to that definition in a textbook used by top schools, I'm honestly curious to see something like this.
There used to be much more of a belief that the reason this stuff has an appreciable value to speak of is because it's either creating a channel for society's productivity or it's creating a position to skim pennies off of society's productivity. But less and less does it feel like productivity is even in the equation anymore. What the quoted statement describes is much closer to a Ponzi scheme than what finance is actually supposed to be about, but it's getting harder and harder to tell the difference.
Also if the cutoff point for raising concerns about this was hundreds of years ago that really sucks for everyone alive today.
Is there a particular instrument or transaction you feel creates a scheme? What specifically would you disallow, and what would be the downstream impact you imagine of doing so?
Finance has always run on both: an asset that produces something has a floor. An asset that produces nothing does not. Between the two lies human nature. One way to get rich is to focus on fundamentals. One way to get rich or poor faster is to bet on human nature.
Just because something is doesn’t mean it ought to be. We’ve settled on this system because it seems to be generally the most effective way of valuing things. In times of extreme changes in valuation it comes off as more egregious than normal.
Well if it's eating the economy to an unprecedented extent of morbid obesity, it might lead to the most massive stroke and complete or partial collapse.
I just don't get it. How do you go from writing the kinds of future visions he has to staring at the singularity practically hitting you in the face and calling it "the world's money-losingest technology"?
Is it because he isn't actually using the technology for work on a day-to-day basis like a lot of us?
I just cant get over this term, do you honestly believe in this? I use AI daily and while it is super useful I see too many limitations for it to “recursively improve and cause an intelligence explosion”.
1. Clearly, the people selling AI with this idea benefit greatly with this promise of infinite upside. Can you can trust them?
2. Singularity essentially requires to handwave away a lot of baked-in issues with LLMs or rely on unrealized innovation.
What about the tools makes you think we've hit the singularity? My experience with them is that they've memorized a lot of stuff, but can't make anything fundamentally new. Most of the useful things LLMs do amounts to semantic search.
They are new proofs and for sure useful but as far as I’ve understood mainly interpolative. E.g. an LLM can “create” a poem about a purple chicken as it has datapoints for “purple” and “chicken”, so it can create something plausible inbetween.
Similarly, in my mind it can interpolate proofs by interpolating between data points for technique A and technique B. This is novel and brute-forcing proofs this way is useful. It is analogus to how sometimes it can generate programs that pass unit tests, I think.
However, creating fundamentally new concepts outside of the interpolated datapoints is not something I am convinced of. Maybe it can extrapolate some things, if correct add it as a data point, continue. Essentially a search, and it would be amazing if this works and maybe we can get some recursive improvement this way. But the “ideas” it will use to conduct this search are a function of the input data points as well, and thus in my view fundamentally limited in novelty. I am not discounting the usefulness, but I am not convinced you can just keep doing this indefinitely scaling intelligence exponentially.
Of course nobody can know yet really and I am just speculating just like you. But I also think the “experts” Sam and Dario also don’t know, and given their incentives I am not really convinced by them.
So, what your saying is that there is a perhaps linear, perhaps exponential increase, and that you are projecting that increase forward indefinitely. Let me know if this is unfair.
Counter argument: does anything else work this way? E.g. Moores law had an end too right? I would argue that the core tech breakthrough (Transformer-based LLM) has been improved, but no fundamental further innovation seems to have been made. The current architecture fundamentally hallucinates, even Fabel even on trivial problems. I.e. as number tokens increase error likelihood goes to infinity. How then, can this scale recursively to infinity?
I believe that AI is a singularity. Just like a Black hole, no information can escape and you're slowly (by subjective experience) dragged in and spaghettified.
I think having a technical argument is important, as we live in a time with lots of hype merchants who stand to benefit from record breaking IPOs. Propaganda can affect us all, how do you know you aren’t being sold to?
Whenever I read Cory Doctorow, I feel like someone took the complement of Paul Graham's writing and posted it. I personally find both of them vapid and annoying.
Edit: the article that the author is commenting on is IMO much better than the linked commentary. There's not much to it
> That's the logic of the whole market today. AI – the world's money-losingest technology – attracts investment at the expense of everything else.
I expect Cory to have skepticism about technology that can be exploited for dystopian purposes, but calling AI "the world's money-losingest technology" is out of touch. If AI can support/replace some intellectual work, it'll be revolutionary, and that's what the investment bet is about.
I get that the blog post is making a separate point about Musk's companies but it's dissapointing to see mistakes like this in Cory's thinking
I share your feeling that LLM-based AI is a high-potential technology.
The issue is the objective dollars and cents financials of the situation. It’s literally the technology that is the money-losingest at this time.
The commercial utility of the technology can’t become viable just by being really useful.
There’s a good accounting argument to be made for AI IPOs happening out of a serious need for capital.
I wouldn’t bet money at a casino on this, but if OpenAI went completely out of business or was absorbed into irrelevancy within a calendar year, nobody with a finance background would be surprised. They objectively cannot exist in ~18 months without massive spending cuts or additional cash infusion. And they can’t make their models better and serve more tokens to build that future potential that justify their present valuation without additional capital, which becomes decreasingly efficient as data center build costs skyrocket.
AI has wonderful potential but no amazing product is guaranteed commercial viability. If Uber spends $1500 on tokens per employee they might as well spend $0 on AI and hire more real people to compensate.
I think about how the railroad barons went through a somewhat similar process. By the end of the American railroad buildout, numerous lines became financially unviable within a few short years or decades, some not even really making it into the automobile era. The only railroad business that ended up with any sort of long term profit viability was freight.
The economy gets bigger every year. Therefore, any similarly proportioned large "thing" becomes the largest of all time. Uber lost tons of money for years, I remember the complaints about them. Investors made lots of money because they believe in future value. You're not being forced to make that bet - "money losing" is great for us if we aren't invested!
"money-losingest" -> AI already brings in billions and many AI companies could become profitable in little time if they'd stop R&D and simply keep selling what they already have.
yes, it's risky and investment-heavy but it's not a bottomless pit with no path to break even. there are many other recent technologies - NFTs? data centers in space? - that would be a better fit for this label.
> If AI can support/replace some intellectual work, it'll be revolutionary, and that's what the investment bet is about.
That's one big "If". From personal experience AI just tends to burn money. Time will tell if the investment pays off but I disagree that, at this time, it is out of touch to say AI is a money sink.
> I expect Cory to have skepticism about technology that can be exploited for dystopian purposes, but calling AI "the world's money-losingest technology" is out of touch.
Since when stating facts is "out of touch"? Currently AI is losing many billions of dollars, it's a fact.
> If AI can support/replace some intellectual work, it'll be revolutionary, and that's what the investment bet is about.
"replace some intellectual work" says nothing about the cost of doing it, the losses keep piling up and "revolutionary" pies in the skies come and go without real economic or financial improvement - on the contrary, the economic situation is getting worse mostly due to AI because of the fake urgency of the "AI race". A slower pace of development would cost a lot less and might be economically viable but that's not what we have.
> but it's dissapointing to see mistakes like this in Cory's thinking.
Again, what mistakes? He's stating facts, you're hyping expectations based on marketing propaganda - who and what is mistaken here?
Obviously yes. On evidence alone the path to profit doesn't exist for most of the massive capital sinks. A small number of players At best MAY return on investment, but in the cycle time capital needs a return, most are functionally incapable
AGI isn't happening. So, it's incremental improvements on LLM and Generative methods. Any advance which requires more tech inputs demands more capital. Any advance which requires less tech makes all the existing capex look stupid.
There was a recent discussion:
The Dead Economy Theory: https://news.ycombinator.com/item?id=48324712
"The underlying purpose of AI is to allow wealth to access skill while removing from the skilled the ability to access wealth". (comment on the discussion above: https://news.ycombinator.com/item?id=48334069)
> something is valuable because some people think other people will pay more for it in the future, and not because it does useful things
This has been the definition of finance for hundreds of years. I don't know why it comes across here like this is a new phenomenon.
It used to be only a few steps removed from actual value. Like an empty lot is worth an amount because when you plant trees it's worth an amount because when it has fully grown trees it's worth an amount because when you cut down the trees the timber is worth an amount. That's an acceptable function of finance, it allocates the value of a logging plantation between several different people who may own it at different times, so that one doesn't have to hold it for the full 20-year growth cycle to realize any value from it.
We needed to develop long-term value instruments to incentivize long-term planning. Bonds are necessary for large public projects, for instance. There are many complexities in deriving instruments from that basic concept. What makes you think a particular one shouldn't be acceptable and why?
>> something is valuable because some people think other people will pay more for it in the future, and not because it does useful things > This has been the definition of finance for hundreds of years. I don't know why it comes across here like this is a new phenomenon.
I don't think so. However, you can convince me by providing a reference to that definition in a textbook used by top schools, I'm honestly curious to see something like this.
There used to be much more of a belief that the reason this stuff has an appreciable value to speak of is because it's either creating a channel for society's productivity or it's creating a position to skim pennies off of society's productivity. But less and less does it feel like productivity is even in the equation anymore. What the quoted statement describes is much closer to a Ponzi scheme than what finance is actually supposed to be about, but it's getting harder and harder to tell the difference.
Also if the cutoff point for raising concerns about this was hundreds of years ago that really sucks for everyone alive today.
Is there a particular instrument or transaction you feel creates a scheme? What specifically would you disallow, and what would be the downstream impact you imagine of doing so?
Finance has always run on both: an asset that produces something has a floor. An asset that produces nothing does not. Between the two lies human nature. One way to get rich is to focus on fundamentals. One way to get rich or poor faster is to bet on human nature.
Just because something is doesn’t mean it ought to be. We’ve settled on this system because it seems to be generally the most effective way of valuing things. In times of extreme changes in valuation it comes off as more egregious than normal.
Well if it's eating the economy to an unprecedented extent of morbid obesity, it might lead to the most massive stroke and complete or partial collapse.
I just don't get it. How do you go from writing the kinds of future visions he has to staring at the singularity practically hitting you in the face and calling it "the world's money-losingest technology"?
Is it because he isn't actually using the technology for work on a day-to-day basis like a lot of us?
> singularity
I just cant get over this term, do you honestly believe in this? I use AI daily and while it is super useful I see too many limitations for it to “recursively improve and cause an intelligence explosion”.
1. Clearly, the people selling AI with this idea benefit greatly with this promise of infinite upside. Can you can trust them?
2. Singularity essentially requires to handwave away a lot of baked-in issues with LLMs or rely on unrealized innovation.
I do. Just based on personal experience of using these tools for the last several years and how they’ve progressed.
What about the tools makes you think we've hit the singularity? My experience with them is that they've memorized a lot of stuff, but can't make anything fundamentally new. Most of the useful things LLMs do amounts to semantic search.
They’re discovering previously unknown mathematical theorems.
How is that not new?
They are new proofs and for sure useful but as far as I’ve understood mainly interpolative. E.g. an LLM can “create” a poem about a purple chicken as it has datapoints for “purple” and “chicken”, so it can create something plausible inbetween.
Similarly, in my mind it can interpolate proofs by interpolating between data points for technique A and technique B. This is novel and brute-forcing proofs this way is useful. It is analogus to how sometimes it can generate programs that pass unit tests, I think.
However, creating fundamentally new concepts outside of the interpolated datapoints is not something I am convinced of. Maybe it can extrapolate some things, if correct add it as a data point, continue. Essentially a search, and it would be amazing if this works and maybe we can get some recursive improvement this way. But the “ideas” it will use to conduct this search are a function of the input data points as well, and thus in my view fundamentally limited in novelty. I am not discounting the usefulness, but I am not convinced you can just keep doing this indefinitely scaling intelligence exponentially.
Of course nobody can know yet really and I am just speculating just like you. But I also think the “experts” Sam and Dario also don’t know, and given their incentives I am not really convinced by them.
So, what your saying is that there is a perhaps linear, perhaps exponential increase, and that you are projecting that increase forward indefinitely. Let me know if this is unfair.
Counter argument: does anything else work this way? E.g. Moores law had an end too right? I would argue that the core tech breakthrough (Transformer-based LLM) has been improved, but no fundamental further innovation seems to have been made. The current architecture fundamentally hallucinates, even Fabel even on trivial problems. I.e. as number tokens increase error likelihood goes to infinity. How then, can this scale recursively to infinity?
Not indefinitely but at least to the point where they’re smarter than humans.
I believe that AI is a singularity. Just like a Black hole, no information can escape and you're slowly (by subjective experience) dragged in and spaghettified.
Sounds great, but what is this belief based on?
I think having a technical argument is important, as we live in a time with lots of hype merchants who stand to benefit from record breaking IPOs. Propaganda can affect us all, how do you know you aren’t being sold to?
I think you are confusing Charles Stross with Cory Doctorow
I think you’re right. And yet still.
Ideological capture is an often socially enforced prison
For the people around them, yes. They themselves don't see it that way.
'Dead' economy theory, you say? I guess the economy must have been alive at some point, then. In my short life, though, I don't think it ever was....
Whenever I read Cory Doctorow, I feel like someone took the complement of Paul Graham's writing and posted it. I personally find both of them vapid and annoying.
Edit: the article that the author is commenting on is IMO much better than the linked commentary. There's not much to it
https://crookedtimber.org/2026/06/15/one-big-grift/
[dead]
most interesting sentence is the first one
> That's the logic of the whole market today. AI – the world's money-losingest technology – attracts investment at the expense of everything else.
I expect Cory to have skepticism about technology that can be exploited for dystopian purposes, but calling AI "the world's money-losingest technology" is out of touch. If AI can support/replace some intellectual work, it'll be revolutionary, and that's what the investment bet is about.
I get that the blog post is making a separate point about Musk's companies but it's dissapointing to see mistakes like this in Cory's thinking
I share your feeling that LLM-based AI is a high-potential technology.
The issue is the objective dollars and cents financials of the situation. It’s literally the technology that is the money-losingest at this time.
The commercial utility of the technology can’t become viable just by being really useful.
There’s a good accounting argument to be made for AI IPOs happening out of a serious need for capital.
I wouldn’t bet money at a casino on this, but if OpenAI went completely out of business or was absorbed into irrelevancy within a calendar year, nobody with a finance background would be surprised. They objectively cannot exist in ~18 months without massive spending cuts or additional cash infusion. And they can’t make their models better and serve more tokens to build that future potential that justify their present valuation without additional capital, which becomes decreasingly efficient as data center build costs skyrocket.
AI has wonderful potential but no amazing product is guaranteed commercial viability. If Uber spends $1500 on tokens per employee they might as well spend $0 on AI and hire more real people to compensate.
I think about how the railroad barons went through a somewhat similar process. By the end of the American railroad buildout, numerous lines became financially unviable within a few short years or decades, some not even really making it into the automobile era. The only railroad business that ended up with any sort of long term profit viability was freight.
The economy gets bigger every year. Therefore, any similarly proportioned large "thing" becomes the largest of all time. Uber lost tons of money for years, I remember the complaints about them. Investors made lots of money because they believe in future value. You're not being forced to make that bet - "money losing" is great for us if we aren't invested!
> calling AI "the world's money-losingest technology" is out of touch
In what way is it out of touch or wrong? It is objectively correct today.
It may very well not be correct 2 years from now, but his statement was about the present, not the future.
"money-losingest" -> AI already brings in billions and many AI companies could become profitable in little time if they'd stop R&D and simply keep selling what they already have.
yes, it's risky and investment-heavy but it's not a bottomless pit with no path to break even. there are many other recent technologies - NFTs? data centers in space? - that would be a better fit for this label.
Is that true? If Anthropic stopped last November at versions 4.5, would people keep using it enough to recoup the investment?
> If AI can support/replace some intellectual work, it'll be revolutionary, and that's what the investment bet is about.
That's one big "If". From personal experience AI just tends to burn money. Time will tell if the investment pays off but I disagree that, at this time, it is out of touch to say AI is a money sink.
> I expect Cory to have skepticism about technology that can be exploited for dystopian purposes, but calling AI "the world's money-losingest technology" is out of touch.
Since when stating facts is "out of touch"? Currently AI is losing many billions of dollars, it's a fact.
> If AI can support/replace some intellectual work, it'll be revolutionary, and that's what the investment bet is about.
"replace some intellectual work" says nothing about the cost of doing it, the losses keep piling up and "revolutionary" pies in the skies come and go without real economic or financial improvement - on the contrary, the economic situation is getting worse mostly due to AI because of the fake urgency of the "AI race". A slower pace of development would cost a lot less and might be economically viable but that's not what we have.
> but it's dissapointing to see mistakes like this in Cory's thinking.
Again, what mistakes? He's stating facts, you're hyping expectations based on marketing propaganda - who and what is mistaken here?
“Losing” is a loaded word choice. If I buy something I’m really happy to have, I probably don’t describe it as losing the money.
Obviously the investment expense has been extremely high, which is what the replies are quibbling about.
AI is currently a massive money sink. Yes or No?
Not for me, it means I don't have to hire an engineer right now.
Obviously no...? The implications far outweigh the "money sink" notion...
Obviously yes. On evidence alone the path to profit doesn't exist for most of the massive capital sinks. A small number of players At best MAY return on investment, but in the cycle time capital needs a return, most are functionally incapable
AGI isn't happening. So, it's incremental improvements on LLM and Generative methods. Any advance which requires more tech inputs demands more capital. Any advance which requires less tech makes all the existing capex look stupid.
I have noticed that those that are the most optimistic about AI almost always talk in a future-tence.
It WILL do this, it COULD achieve that etc.
I think Ed Zitron challenged us all to talk to an AI booster without letting them use future tense.
It can't. It can only provide the appearance of doing so. It's just the Eliza effect on steroids.
if
[dead]
[flagged]