I've been enjoying the industrial revolution parallels. Have the economics of bespoke products changed? Not really. But is it now easier to get a crude, but functional, approximation of an idea? Yes. The main issue appears to be users conflating B for A (and not realising their AI creation is fundamentally a mess).

I find the industrial revolution parallels interesting too. I figure we're in the steam age and this is a bit like the railway bubble. Or maybe a bit before that - I'm not sure current AI is as useful as railways.

Just wait for the economics to catch up and eat the marketing. No effort required.

Though, the larger the bubble becomes the more damaging the pop will be

That's my real fear, that I have massive indirect AI bubble exposure while having zero direct exposure. Similar to how relatively few Americans had liar loans, but we all got burned by the subprime crisis. And then there were those bad actors that made the crisis work for them, and we got burned again during the uneven recovery.

With everybody investing on index funds, directly or indirectly through their 401k, does economics ever have a time to catch up? Being out of the market during the crazy hype means losing out on "easy" gains, so basically it keeps going up and up, these companies get even more hyped, and I wonder what's the end game.

I keep hearing that retail investors are just small fry and have no power to affect any stock's valuation, but certainly the hype spreads to big money as well.

Even if I'm way off the mark, I would love to read about what happens to the economy in the mathematically plausible case that people keep buying and no one ever sells enough to mitigate the hype.

The endgame is owning equity assets which represent ownership in companies that make profits? I hold these valuable assets and then sell them in the future. What else could it be?

The assets are valuable because they represent ownership of the most productive companies on the planet?

But what about the damage along the way? Misallocation of capital? Too many data centers and not enough jobs? CO2? Expensive RAM? Endless articles on HN about AI saying there are too many articles on HN about AI?

The “reverse centaur” is a natural product of capitalism wanting workers to be as replaceable as possible in order to drive down wages. An ordinary “centaur” is counter to companies’ goals, they don’t want workers empowered.

If that is true, they are incredibly stupid. Empowered workers create better, more effective companies.

But they also create workers that can hop off the ship. If they're on a sinking one, they have no choice but to work hard to get back to shore.

> But they also create workers that can hop off the ship.

Sounds to me like a win for the workers, a win for the employer(in that they get better companies) and a win for society at large.

I don't think CEO class want better, more effective companies. They have nothing to gain by that. They want monopoly situation for themselves, locked customers and helpless dependent employees.

It depends but obviously they want both, both control and value.

Enough control will create value for them anyway, it just might be at the expense of everyone else.

As a business owner I assure you I am not overly obsessed with labor costs. I obsess on all costs, and all revenues. VC funded and public companies may or may not take a similarly holistic view.

AI spending is all what keeping american economy afloat or else a big recession

"If that’s the case, AI will let them wire the toy steering wheel directly into the drivetrain. So you can have an amazing idea as a corporate visionary, and you don’t have to have any ego-shattering confrontations with people who know how to do things, who tell you you’re actually an idiot"

Just beautiful.

It's kinda too big to fail in US now. If it bursts, US will decline quickly.

It's not, though. The banking system holds onto everyone's money in a fractional reserve system, if you let a run on the banks happen everyone's money is gone. That's "too big to fail".

If anthropic and openAI fail, the top 10% lose half their money in a stock sell-off. That's perfectly tolerable. Maybe congress bails them out anyways but they don't have to.

If they get bailed out with taxpayer dollers they better maje damn sure the taxpayers are getting immediate tangible benefits (as in cash or free mythos class AI every month at a minimum). Sama can go beg on the streets, we'll keep the tech.

So you’re going to be shorting AI stocks?

Bubbles do not have roots.

I agree the mixed metaphor and grammar are a bit awkward, but "its" refers to AI, innit?

> “The bubble doesn’t want cheap useful things,” Doctorow said. “It wants expensive ‘disruptive’ things

That about sums it up. I challenge anyone to name five things that have gotten cheaper or better in their life thanks to AI.

I can't think of anything that I consume that has gotten better, and the only thing that has gotten cheaper is the value of your skills to your employer, as it wants you to offload more of your work to a machine they own or rent. But perhaps someone here can find some tangible improvement.

Interesting challenge.

1 - I worked abroad and wasn't really familiar with the systems there. Gemini made me aware of a kind of pension account that I could withdraw from when I left the country netting me a few thousand dollars.

2 - Working as a tech contractor, charging by deliverable, Codex/Claude Code speed me up and it doesn't seem to have significantly dropped rates in the market.

3 - Also contractor related: I had Claude do a quick legal sanity check of my contracts, and it warned me of some clauses that I'd be better off removing/changing/refining. I was not aware of these nuances and would not have paid a lawyer for this as the contract was too small, but the changes were accepted by the client and reduced my risk exposure meaningfully.

4 - Learning a foreign language, I use it to check my draft emails and messages. It corrects them but also serves a tutoring role providing feedback, improving both the accuracy of my communication and my rate of language acquisition.

5 - Gemini Deep Research helped me narrow down tent models that met my fairly specific set of requirements. Very happy with the tent I ended up buying, from a brand that was not on my radar before.

I can name three:

- better speech to text, - better auto translation, - better image to text.

Other then that, I hate how AI is inserted into everything, how I cant tell anyone I did something without them telling me I could have use AI for it, the doom trolling , the AIG singularity bullshit as if it was new incoming god rather then a set of technologies.

It's allowed me to learn things more quickly. For example, I haven't written any C++ professionally for 20 years, and found myself in the position of needing to know it again. The existing material online for learning it is thick; sometimes I have a question that isn't answered directly or clearly in any way I can find with a standard search. I can grill the LLM much more quickly and pointedly.

However, this does not answer the usefulness-at-work question. Does anyone care if I know how the initializer list braces in constructor `Foo{1,2}` work? Today for hiring managers smitten with AI, seems like they don't.

Personally, I'm just trying to stay sharp on the off-chance that things crash out and people who know how to engineer *and* code become highly valuable. If not, I'll be doing something else, anyway.

While I agree with your sentiment and observatin by and large, ideally you could have added a few constraints as a lot of people here have had "Holy shit" moments with AI, but often it's most pronounced in a very specific personal context. Learning topics faster, fixing things without having to call a repair man, life-saving medical tips (somewhat more of a thing because of a handicapped health system, but that's another discussion).

Funnily enough though, the benefits don't seem to scale to the work environment as smoothly across the board. It's almost as if, to use pg's famous line, they built things that don't scale.

The moment you scale AI beyond the personal context, the noisiness and non-determinism of the technology starts wrecking havoc in ripples through the industry, damn the productivity. Managing the slop you've created, but more importantly (I would normally put an emdash here) managing the slop others created turns out to be a problem as big as the tech that created it.

What is the evidence of this? Isn't it strange no revolutionary product has yet been created from the coding output of this technology? More to the point, the only thing that comes close is the Claw. So what we have as the first breakaway product is something to help orchestrate the slop generators, instead of say, a new Phone OS that rivals Android and iOS, or a Metaverse that is actually worthy of its name. Nothing to see here, just more shovels to shovel the shit essentially.

We are still in the grasp of the electric and information revolution. Peter Thiel said more than a decade ago that if you compare life in the 70s and now the only major visible difference would simply be more screens in various forms nowadays. This he presented as evidence that no new revolutionary technology has been created in the interim. At present, the same is more true of AI. It has made access to information faster than before. However, unlike other processes of learning, it has made it more difficult to ascertain the correctness of the information presented. It is doing this by actively removing the direct connection between the information it presents and the primary sources responsible for that information (think Google replacing its search field with an AI field).

The non-LLM AI stuff in the medical field is what I find most promising. Not the banal AI doctors and nurses replacing real doctors and nurses, but more the research end of things, where hallucinating and pattern matching to create new compounds to test seems promising. Of course, any technology that is able to routinely aid discoveries in applied maths and physics will also be welcome.

If this half decade saw an explosion of new bespoke hardware being built and sold aided by LLMs (think a tech shop of 10 employees being able to build and sell robust smartphones from scratch to a niche market sustainably etc), or jumpstarts an era of reverse engineering on crack that pierces through the tech oligopoly then that would be evidence of disruption on a macro scale... But no, alas, we are still waiting.

says " Peter Thiel said more than a decade ago that if you compare life in the 70s and now the only major visible difference would simply be more screens in various forms nowadays. "

You could extend Thiel's statement back even to the 1960's: beyond address and phone lookup, everything added by the WWW has been of little use to a wise consumer IMO.

Sounds more like a simpleton than a simple man.

And yet you choose to enshitify the comments of this thread

The bubble is weirdly being kept alive by the promise it can replace software developers, which is ironic because all it has really achieved is a deluge of slop.

Social media has also gone utterly crazy. The last time I saw a gaslighting operation on this scale and volume (including accounts who "hate" AI "because its so scary good") was during the start of the Gaza war.

These accounts becoming easier to distinguish too, because whether they boost AI or feign criticism they all categorically refuse to use the s word.

I guess there is a few trillion riding on perpetuating this mass psychosis so it makes sense they'd try to use every trick to keep it going as long as possible.

Internet bubble was nothing like this. The scale is greater, the promises are more insane and the pop is going to be much more devastating.

> The bubble is weirdly being kept alive by the promise it can replace software developers

And lawyers, and doctors, and tax preparers, and financial managers...

People seem to think it's all smoke and mirrors. IDK. My employer, in an industry as far removed from Silicon Valley as you can probably get, makes more and better use of it all the time. There's enormous amounts of work done every day in corporate America that amounts to "I need X, but X involves some data from legacy system Y and legacy system Z, and that's going to take me an hour to glue together because our entire enterprise runs on a system cobbled together over the past 50 years". You know what Codex/Cowork/etc can do really stunningly good these days? Take the files you provide, listen to what you want from them, ask clarifying questions as necessary, then write a script that programmatically does exactly what you ask for, checks its own work and then gives you the result.

We also realized that a project we thought was going to involve thousands of man-hours of very expensive, senior draftsperson labor to backport a feature into decades of CAD files we could, after years of procrastination, just forget! AI has got to the point where it can see and count what we need for us with greater accuracy, in our tests, than our very best humans, so we can just make the change going forward and let AI read the old files as needed.

There's incremental commercial adoption of that nature that I'm sure is happening slowly across the corporate spectrum, and that kind of thing is durable demand, not a bubble.

Note though that I'm talking about real revenue, etc. Not stock market bubbles. The feds been inflating that junk since the housing bust. We'll all pay for that eventually.

>There's enormous amounts of work done every day in corporate America that amounts to "I need X, but X involves some data from legacy system Y and legacy system Z, and that's going to take me an hour to glue together because our entire enterprise runs on a system cobbled together over the past 50 years

I did a bunch of that type of work pre covid. IME it's usually an hour to build and about 3-5 weeks to handle bureaucracy, data cleaning and detective work to account for the lack of docs. AI would probably handle that 1 hour's worth of work OK but using it for the rest of that work would be a shortcut to hallucination town, especially when the context is something you have to dig up by actually talking to people.

Also it's the only bit of that job that isnt mind numbing.

There's an enormous number of projects all over the economy these days which are being presented as an enormous "AI win" where the benefits were dubious at best, because it's become a clear route to career advancement.

I'm equally certain that it has its niches but ~80% of where I'm seeing it used it's either smoke and mirrors or creating more problems than it is solving.

My default position is: if an LLM is good at it, we probably don’t actually need it.

> refuse to use the s word

strike?

Maybe slop?

slop

While that's a bit cruel, I do have to wonder how exactly Doctorow got in the position he has -- he isn't somebody who has done great things and is now a tech critic like someone like Geoffrey Hinton. He seems a pleasant enough fellow, but basically he got fame from writing SF books that he gave away for free. How does this translate into being an expert that journalists fawn over?

He's also been a vocal advocate about copyright reform and against DRM, and his arguments are coherent and generally well-regarded among the tech crowd. In context, "giving away books for free" is a particularly strong form of putting his money where his mouth is.

Journalists fawn over many people who write for public consumption in some way (book authors, academics, etc.), and who already agree with the point of view the journalists want to promulgate. I don't think Doctorow has particularly more insight about AI than anyone else does, but I also don't think that "insight about AI" is the main reason Ars Technica chose to publish an interview with him - they did this because Ars Technica has an anti-AI editorial position and so they find it useful to promote Doctorow's anti-AI-book. A different group of journalists who did not have an anti-AI editorial position would not have interviewed and published Doctorow, or at least not done so under as friendly circumstances as Ars Technica did.

Is this supposed to be surprising?

Hinton is a classic case of an academic making increasingly grandiose statements as they age.

Has Doctorow implemented Transfomer from scratch as Lesswrong wants us to atleast pass the smell test?

Usually when talking about bubbles bursting its about a stock market bubble. P/E ratios now are approaching/passing the P/E ratios during the dot com boom/bust. Another reference point with high P/E ratios was around 1929-1930 (there are others too).

The dot com boom can be thought of as ecommerce which enabled investing in a lot of silly ideas at the time. A key technology that underpinned it all and won out was search. Every ecommerce site felt like they needed search, there were search engine companies, lots of competition across google, yahoo, etc.

An interesting lens to put on AI and the current stock market is that the software will be commoditized, its an eventuality. Its trending towards being able to run LLMs locally and get decent output. Decent is subjective but output similar to Q4 of 2025 models is when we started seeing more consistently usable output.

I believe that will a potential the inflection point for a bubble bursting the stock market: local or DIY LLMs producing "good enough" output and companies publicly backing away from enterprise contracts, lowering their AI spend if they can find cheaper ways to do it.

Ironically I had a job figuring out what he meant by strike at its roots so I asked an LLM. Gemini seems to think he means labour bargaining/strikes, attacking the stock bubble with better accounting regulations and open source.

I'm not sure any of those will make much difference beyond what's already happening?

Didn’t this man just release enshittification? He’s killing it from north of the border

AI is necessary evil.