LLMs create a new workflow wherever they are employed. Even if capable, that is not always a more desirable/efficient experience.

Internet, handheld computers, electric cars...The problem is the same dudes.

Putting beanie babies in with Quantum Computing and Nuclear Power completely ignores the potential life changing elements of some technologies, even if they don't work.

Oh, and smart glasses he put in there, so he'll be eating his words in 2 years.

75% of restaurant orders are delivery now due to widespread personal electric transportation. It already has fundamentally changed humanity.

https://youtu.be/KOSUEFqszK8

This only doesn't feel like substantiation if you reject the notion that these cases are analogous.

"You shouldn't eat that."

"Why not?"

"Everyone else who's eaten it has either died or gotten really sick."

"But I'm different! Why should I listen to your unsubstantiated claims?"

"(lists names of prior victims)"

"That doesn't mean anything. I'm different. You're just making vague and dismissive unsubstantiated claims."

The claim isn't "AI bad" the claim is more along the lines of "there's a lot of money changing hands and this has all the earmarks of a classic hype cycle; while attention/diffusion models may amount to something the claims of their societal impacts are almost certainly being exaggerated by people with a financial stake in keeping the bubble inflated as long as possible, to pull in as many suckers as possible."

If you want another example (which you won't find analogous if you've already drunk the koolaid):

https://theblundervault.substack.com/p/the-segway-delusion-w...

I mean, disillusionment is the least of my worries.

Squares are rectangles. The existence of rectangles that aren't squares doesn't negate that.

The stuff LLMs will democratize will be a lot more impactful than nice posters for car wash fundraisers though. So in that sense it will be different, but I don’t think it will crack the market for proficient experts in the field in the same way photoshop didn’t destroy graphic design and CAD didn’t destroy drafting. It may get rid of the market for a lot of the second-tier bootcamp grad talent though, so I wouldn’t be getting into that right now if I could help it.

The point is to take the hype with a grain of salt and knowledge that not all hyped technologies transformed the world as promised. Maybe AI is like the internet or electricity. But maybe the claims about AGI/ASI and full automation are just hype.

AI is different because the magic clearly is because of the tech. The fact that we get this emergent behavior out of (what essentially amounts to) polynomial fitting is pretty surprising even for the most skeptical of critics.

I spent most of Covid in VRChat and met my current live-in gf, so the metaverse was real for me too.

I also made decent money selling crypto, so that part was real for me too.

And AI coding, for as dumb as even the best models are, still enabled me to create things that I wanted to, but wouldn't have had time or gotten nearly as far without.

I dunno if the author realizes, but all the things they mentioned did materialize in one way or another, just not exactly how the hype described it.

Maybe if they could let go of some of the cynicism, they could find something to be optimistic about. Nothing ever goes exactly as planned, but that doesn't mean nothing is good.

I like technology. I made a decent living from it. But if I had chased every hyped fad that was promised as the next big thing, I doubt I'd be as happy as I am now.

It's different just like the steam engine was different, except technology moves much 100x faster now than it did then. It's different and the same.

Gen AI reached 39% adoption in two years (internet took 5, PCs took 12). Enterprise spend went from $1.7B to $37B since 2023. Hyperscalers are spending $650B this year on AI infra and are supply-constrained, not demand-constrained. There is no technology in history with these curves.

The real debate isn't whether AI is transformative. It's whether current investment levels are proportionate to the transformation. That's a much harder and more interesting question than reflexively citing a phrase that pattern-matches to past bubbles.

"AI will change everything!"

Few seem to understand that both of the above can be true. The parallel you draw to the internet revolution is apt; dot-coms were both a bubble and changed everything.

It's not a very legible situation for people outside of the profession, and a lot of them believe it's just another grift that will blow up in a few years.

Granted, it fixed the problem in the very next prompt.

I literally just went on Gemini, latest and best model and asked it "hey can you give me the best prices for 12TB hard drives available with the British retailer CeX?" and it went "sure, I just checked their live stock and here they are:". Every single one was made up. I pointed it out, it said sorry, I just checked again, here they are, definitely 100% correct now. Again, all of them were made up. This repeated a few times, I accused it of lying, then it went "you're right, I don't actually have the ability to check, so I just used products and values closest to what they should have in stock".

So yeah, hallucinations are still very much there and still very much feeding people garbage.

Not to mention I'm a part of multiple FB groups for car enthusiasts and the amount of AI misinformation that we have to correct daily is just staggering. I'm not talking political stuff - just people copy pasting responses from AI which confidently state that feature X exists or works in a certain way, where in reality it has never existed at all.

I encounter stuff like this every week, I don't know how you don't. I suppose a well-structured codebase in a statically typed language might not provide as much of a surface for hallucinations to present themselves? But like you say, logical problems of course still occur.

LLM's are not artificial general intelligence (i.e. not sci-fi AI). Why haven't they transitioned to being mere algorithms by now? Why is the public being told AI is finally arriving when it's really just another algorithm?

We have some truly slick and shady corporations involved in the bubble right now and they're marketing LLM's like tobacco. LLM's have been pushed out, at immense cost, to the public in a way that makes them more directly accessible to average people than any past algorithm. Young children can ask a LLM to do their homework for them. Middle managers can ask a LLM to create a (shitty) ad campaign for them. Corporations have gone to tremendous expense to make that widely available and, for the moment, mostly free. They seem to be following the Joe Camel school of marketing. Get them hooked while they're young so they come to you first when they're older! The only difference is that nobody is stepping in to stop the new Joe Camel from handing out free samples to kids.

Then there's the "go big" aspects of the bubble. The major competitors are trying to out-spend each other to dominance, but the suckers are so colossally big that their bubble is affecting global GPU, memory, and storage prices. This bubble is going to stress power grids wherever it operates and do considerable environmental harm. The financial games being played behind the bubble are absolutely stupid. The results, so far, are tantalizing for billionaires. LLM's offer the promise of being able to fire all their pesky and annoying human workers. It won't deliver on that, and none of these companies is ever going to make enough to pay their debts. There might be "too big to fail" government bailouts, but there are going to be some big bankruptcies too.

Useful algorithms will come out of all this, a lot of tears too, but not "AI".

What is meta-technology?

Umm, what? For the past 3 years, every year I've said something along the lines of "even if models stop improving now, we'll be working on this for years, finding new ways to use it and make cool stuff happen". The hype is already warranted. To have used these tools and not be hyped is simply denial at this point.

I was so expecting to find this wind-up aimed at those peddling the "AI is hype" laziness.

It's laziness because they have little CS fundamentals to base such claims on, and the deductions can be made, just not clearly to people who need to study a lot more.

It's like watching an invisible train (visible to those with strong CS) rolling down the tracks at a leisurely pace. Those sitting in their stalled car on the tracks are busy tweeting about "AI HPY PE TRAIN." Until it wrecks their car, the gimmick is free oxygen. It's a lot easier to write articles than it is to build GPUs and write programs.

The one you keep citing, here and in the article, Quibi, lives on in technology-form (the spirit of your article we must presume) as an 8 billion dollar business in China and is rapidly upending every Hollywood film studio.

So, arguments about substantiation or even 'this time' fall flat in the face of not even understanding your own message.

    x-clacks-overhead GNU Terry Pratchett

Non-coding work is thinking about the system architecture, thinking about how data should flow, thinking about the problem to be solved, talking with people who will use it, discovering what their objectives are.

Producing 40k lines a code per day simply means you're not doing any of that work: the work that ensures you're building something worth building.

Which is why the result is massive, pointless things that don't do the things people actually need, because you've not taken any time to actually identify the problems worth solving or how to solve them.

It's a form of mania that recalls Kafka's The Burrow, where an underground creature builds and builds an endless series of catacombs without much purpose or coherence. When building becomes so easy when it was so hard -- and when it becomes more fun to build and watch codex's streams of diffs fly by, than to plan -- we forget the purpose of building, and building becomes its own purpose, which is why we usually so little actual productive impact on the world from the "40k lines of code a day" cohort.

From the post, which is not a very long one: "All of the above technologies are still chugging along in some form or other (well, OK, not Quibi). Some are vaguely useful and others are propped up by weirdo cultists"

I mean you're just stating that sometimes tech doesn't meet it's hype. What's insightful about that? It's a given; cherry-picking examples doesn't prove your case.

Similarly to how titles that start with "how" usually have that word automatically removed.

It is physically and physiologically impossible for anyone to be reviewing "30-40K lines of nearly perfect code a day" to the extent needed to push it with confidence in a sensible development process.

  > 39% adoption in two years (internet took 5, PCs took 12).

Will that happen in the future, maybe. but I don't have enough insight into how AI is evolving in the labs to make a judgement on that.

1. http://lucacardelli.name/Papers/Binary.pdf

2. https://www.researchgate.net/publication/221321423_Parasitic...

Second, asymmetric multimethods give something up: symmetry is a desirable property -- it's more faithful to mathematics, for instance. There's a priori no reason to preference the first argument over the second.

So why do I think they are promising?

1. You're not giving up that much. These are still real multimethods. The papers above show how these can still easily express things like multiplication of a band diagonal matrix with a sparse matrix. The first paper (which focuses purely on binary operators) points out it can handle set membership for arbitrary elements and sets.

2. Fidelity to mathematics is a fine thing, but it behooves us to remember we are designing a programming language. Programmers are already familiar with the notion that the receiver is special -- we even have a nice notation, UFCS, which makes this idea clear. (My language will certainly have UFCS.) So you're not asking the programmer to make a big conceptual leap to understand the mechanics of asymmetric multimethods.

3. The type checking of asymmetric multimethods is vastly simpler than symmetric multimethods. You're algorithm is essentially a sort among the various candidate multimethod instances. For symmetric multimethods, choosing which candidate multimethod "wins" requires PhD level techniques, and the algorithms can explode exponentially with the arity of the function. Not so with asymmetric multimethods: a "winner" can be determined argument by argument, from left to right. It's literally a lexicographical sort, with each step being totally trivial -- which multimethod has a more specific argument at that position (having eliminated all the candidates given the prior argument position). So type checking now has two desirable properties. First, it design principle espoused by Bjarne Stroustroup (my personal language designer "hero"): the compiler implementation should use well-known, straightforward techniques. (This is listed as a reason for choosing a nominal type system in Design And Evolution of C++ -- an excellent and depressing book to read. [Because anything you thought of, Bjarne already thought of in the 80s and 90s.]) Second, this algorithm has no polynomial or exponential explosion: it's fast as hell.

4. Aside from being faster and easier to implement, the asymmetry also "settles" ambiguities which would exist if you adopted symmetric multimethods. This is a real problem in languages, like Julia, with symmetric multimethods. The implementers of that language resort to heuristics, both to avoid undesired ambiguities, and explosions in compile times. I anticipate that library implementers will be able to leverage this facility for disambiguation in a manner similar to (but not quite the same) as C++ distinguishes between forward and random access iterators using empty marker types as the last argument. So while technically being a disadvantage, I think it will actually be a useful device -- precisely because the type checking mechanism is so predictable.

5. This predictability also makes the job of the programmer easier: they can form an intuition of which candidate method will be selected much more readily in the case of asymmetric multimethods than symmetric ones. You already know the compiler is using: it's just double-dispatch, the trick used for "hit tests" of shapes against each other. Only here, it can be extended to more than two arguments, and of course, the compiler writes the overloads for you. (And it won't actually write overloads, it will do what I said above: form a lexicographical sort over the set of multimethods, and lower this into a set of tables which can be traversed dynamically, or when the types are concrete, the compiler can leverage monomorphize -- the series of "if arg1 extends Tk" etc. is done in the compiler instead of at runtime. (But it's the same data structure.)

6. It's basically impossible to do separate compilation using symmetric multimethods. With asymmetric multimethods, it's trivial. To form an intuition, simply remember that double-dispatch can easily be done using separate compilation. Separate compilation is mentioned as a feature in both the cited papers. This is, in my view, a huge advantage. I admit, this I haven't quite figured out generics will fit into this -- at least if you follow C++'s approach, you'll have to give up some aspects of separate compilation. My bet is that this won't matter so much; the type checking ought to be so much faster that even when a template needs to be instantiated at a callsite, the faster and simpler algorithm will mean the user experience will still be very good -- certainly faster than C++ (which uses a symmetric algorithm for type checking of function overloads).

To go a bit more into my "vision" -- the papers were written during a time when object-orientation was the dominant paradigm. I'd like to relax this somewhat: instead of classes, there will only be structs. And there won't be instance methods, everything will be a multimethods. I'll adopt the Python approach where everything is public, so you need to worry about accessibility. Together with UFCS, this means there is no "privileging" of the writer of a library. It's not like in C++ or Java, where only the writer of the library can leverage the succinct dot notation to access frequently used methods. An extension can import a library, write a multimethod providing new functionality, and that can be used -- using the exact same notation as the methods of the library itself. (I always sigh when I read languages, having made the mistake of distinguishing between free functions and instance methods, "fix" the problem that you can only extend a library from the outside using free functions -- which have a less convenient syntax -- by adding yet another type of function, an "extension function. In my language, there are only structs and functions -- it has the same simplicity as Zig and C in this sense, only my functions are multimethods.)

Together with my ideas for how the parser will work, I think this language will offer -- much like Julia -- attractive opportunities to extend libraries -- and compose libraries that weren't designed to work "together".

And yeah, Claude Code and Gemini are going to implement it. Probably in Python first, just for initial testing, and then they'll port it to C++ (or possibly self-host).

Plenty a visual programming language has tried to toot their own horns as being the next transformative change in everything, and they are mostly just obscure DSLs at this point.

The other issue is nobody knows what the future will actually look like and they'll often be wrong with their predictions. For example, with the rise of robotics, plenty of 1950s scifi thought it was just logical that androids and smart mechanic arms would be developed next year. I mean, you can find cartoons where people envisioned smart hands giving people a clean shave. (Sounds like the making of a scifi horror novel :D Sweeney Todd scifi redux)

I think AI is here to stay. At very least it seems to have practical value in software development. That won't be erased anytime soon. Claims beyond that, though, need a lot more evidence to support them. Right now it feels like people just shoving AI into 1000 places hoping that they can find an new industry like software dev.

Source?

So far, life goes on roughly the same as it did five years ago. This can feel 'underwhelming' in contrast to the onslaught of public discussion about, and huge investments in, AI.

Most of us here on HN are programmers, and we all know how radically LLMs have changed our code projects. Even so, the change to our everyday lives (aside from our work or hobby project) is not, just yet, glaringly obvious. This year, it's mainly... every website shoves an AI box on its site that nobody seems to want!

I don't hear people saying "nothing is going to change", but I do hear questions about the timeline and if the current levels of investment match returns. Branding these people as stuck in some sort of negative identity is bullshit.

Most of Mag-7 are planning to spend over 500B on capex this year alone on building out datacenters for AI pipelines that have yet to prove that it can generate a sustainable profit. Yes, AI is useful in some environments, but the current pricing is heavily subsidized. So my point stand, the hype is not warranted.

Well, no, the ratio is most definitely not 1-to-1.

I also found the "it's almost always dudes" line a bit strange, because I've seen plenty of women doing marketing for startups running on hype.

MRNA vaccines. Where are the countless breathless articles about these literal life saving tech? A few, maybe, but very few dudes pumping out asinine "white papers" and trying to ride the hype train.

Solar and battery. Again, lots of real world impact but remarkably few unhinged blowhards writing endless newsletters about how this changes everything.

I'm struggling to think of a tech from the last 20 years which has lived up to its hype.

Not everything is written to be insightful. Some things are just written to get them out of my head.

But if they don't and if I have to think twice about how much every request's going to cost, the cost-benefit analysis will look differently fast.

Otherwise his entire team must collectively groan when a Slack message appears: "Got a new PR ready for review everybody!"

So, what CS fundamentals do you need to evaluate if AI is the real thing, or will disappoint in the future? Until a few months ago, coding agents were met with skepticism, until Anthropic introduced their new model and, with it, a hype train that cannot be rationally justified. Look, SOTA LLMs, and coding agents in particular, are impressive. However, current predictions about the future of software development (and the world in general) are speculative. There is little to no data showing whether AI can deliver on its promises. How could there be in this short time frame? No one knows what the future will hold, no one knows how coding agents will be integrated into our work life and everyday life in the long run, or what hard limitations they will reveal. No one can tell you how professions will change in the coming years; every prediction is purely speculative, and anyone making prophecies is either trying to cope with the uncertainty themselves or has some stakes in the AI bet. It would be nice if people were actually humble enough to admit that they have no idea what will happen in the future, instead of writing the hundredth doom and gloom post.

I think you're right but for the wrong reasons wrt sustainable profit.

Specifically, overcounting how much it will cost in 5 years to run AI because you're extrapolating current high prices, and at the same time undercounting how the demand will drive efficiency gains.

But there's a lot of things playing out to our advantage. Vast swathes of useful and publicly available training data. The rigorous precision of said data. Vast swathes of data we can feed it as input to our queries from our own codebases. While we never attained the perfect ideal we dreamed of, we have vast quantities of documentation at differing levels of abstraction that the training can compare to the code bases. We've already been arguing in our community about how design patterns were just level of abstraction our coding couldn't capture and AI has access now to all sorts of design patterns we wouldn't have even called design patterns because they still take lots of code to produce, but now for example, if I have a process that I need to parallelize it can pretty much just do it in any of several ways depending on what I need at that point.

It is easy to get too overexcited about what it can do and I suspect we're going to see an absolute flood of "We let AI into our code base and it has absolutely shredded it and now even the most expensive AI can't do anything with it anymore" in, oh, 3 to 6 months. Not that everyone is going to have that experience, but I think we're going to see it. Right now we're still at the phase where people call you crazy for that and insist it must have been you using the tool wrong. But it is clearly an amazing tool for all sorts of uses.

Nevertheless, despite my own experiences, I persist in believing there is an AI bubble, because while AI may replace vast swathes of the work force in 5-20 years, for quite a lot of the workforce, it is not ready to do it right this very instant like the pricing on Wall Street is assuming. They don't have gigabytes of high-quality training data to pour in to their system. They don't have rigorous syntax rules to incorporate into the training data. They don't have any equivalent of being guided by tests to keep things on the rails. They don't have large piles of professionally developed documentation that can be cross-checked directly against the implementation. It's going to be a slower, longer process. As with the dot-com bubble, it isn't that it isn't going to change the world, it is simply that it isn't going to change the world quite that fast.

Hype is often early, in 10-20 years we'll start seeing the value as the rest of the world catches up

https://www.sfgate.com/food/article/rise-fall-bay-area-start...

Still don't understand what's the end goal here. Assuming they don't deliver, then there are billions of investments that will go bust. Assuming they deliver, millions lose their jobs and there's going to be a bloodbath on the streets.

Do feel AI is overall just hype? When did you last try AI tools and what about their use made you conclude they will likely be forgotten or ignored by the mainstream?

> Not everything is written to be insightful. Some things are just written to get them out of my head.

I like that, going to use it as the motivation to get some things out of my own head.

Or maybe judicious use of an LLM here could be helpful. Replace the auto-edits with a prompt? Ask an LLM to judge whether the auto-edited title still retains its original meaning? Run the old and new titles through an embedding model and make sure they still point in roughly the same direction?

It's amazing to me how those willing to seize on the speculative nature of any ANY uncertainty cannot recognize the inherent uncertainty of the inverse.

> what CS fundamentals do you need

1. Tarski's undefinability theorem 2. Gödel's incompleteness theorems 3. Curry Howard correspondence

And a lot of exposure to deductive reasoning, vague ideas of automated theorem proving and formalization.

I won't pretend its easy, but let's be clear, a small fraction of people who know things are being forced to entertain the hysteria of a vast majority who are unwilling to know things and just go around beating their chests and will continue doing so until the train hits them.

There are 2-3 minor architectural changes in between now and what I would identify as a completely unbounded AGI with clearly discernible dynamic, self-defined objective functions and self-defined procedures for training and inference. It can be done in megabytes. Oh god. Get me out of this forum. I wish to return to my code editor.

It's high time to stop accumulating debt while providing free picture of pelicycles, just charge the full cost for them - enough to generate profits and pay back debt.

What we see now is literally burning money and energy to generate hype. The only true measures of success are financial and macroeconomic. If the hype is real, there should be no problem for the mighty AI to generate debt-free profits for its providers while the overall price level in the US goes down.

We observe the exact opposite which makes the AI hype act only as market manipulation for capital misallocation.

There is a third outcome that combines both of these.

LLMs can massively displace the workforce (and cause widespread social instability) AND the companies pouring hundreds of billions into them right now could, at the same time, fail to capture significant amounts of the labor savings value as late-mover alternatives run the race drafting their progress without the massive spend.

I'd honestly be surprised if this double-whammy isn't the outcome at this point. AI is going to have a massive impact on everything, but there is still no moat in sight.

It was an hour of pasting in error messages and getting back "Aha! Here's the final change you need to make!"

Underwhelming doesn't even begin to describe it.

But, even if I'm wrong, we were told that COBOL would make programming redundant. Then UML was going to accelerate development. Visual programming would mean no more mistakes.

All of them are in the coding mix somewhere, and I suspect LLMs will be.

> usage is copy pasting code back and forth with gemini

the jokes write themselves

As I said, maybe I'm wrong. I hope you have fun using them.