Anamon

How do you see a possibility of avoiding model collapse? It seems to me to be an unavoidable consequence of two facts I consider pretty much irrefutable:

1) LLM content can at best be as good as the source material it was trained on. That's the upper bound. "Out of distribution" output of LLMs is mostly unusable.

2) LLM-generated content increasingly drowns out original content, online and elsewhere.

This spells "monotonically decreasing content quality" to me.

I find Kinsella's argumentation extremely weak, and his defense of it (e.g. his AMA) entirely unconvincing. Unless you susbscribe to a number of very daring assumptions which, in my view, do not at all track with our reality, his ideas just don't hold up.

That being said, in this scenario of yours, how much of that additional work do you figure would be actually original? If someone wants to make a living with music, why bother learning an instrument and writing songs if nobody can stop them from just recording someone else and selling that?

That's really the heart of this lawsuit, isn't it? If we give freeloaders the right to legally monetise other people's work, we will drown in bland, derivative, stolen garbage and drive out the actual creators, in the process also breaking the "business model" of the freeloaders. A real lose-lose-lose situation which I see as much more realistic and believable than Kinsella's libertarian dream world.

Not much need for extrapolation, either. We can just watch it happen right now. "Content production" on the web is definitely exploding, as per your point. But even in the mainstream, most people will agree that this flood of new, LLM-generated "works" is worth less than the originals they're sloppy pastiches of.

And it gets less likely the longer each use is.

I also doubt this. When I observe people out in town, most of them stare at their phones non-stop. To get to pick them up that often, they'd first have to put them down every now and then.

Not all of us are past that. Touchscreens have a variety of downsides that I think most people simply don't think about too much and put up with, and physical keyboards a number of benefits that people aren't aware of because they've been out of fashion for so long.

Less on the device and more on the keyboard. On Android, you can install third-party keyboards like any other app, and there's a wide range in quality. The default Android one is decent, but not among the best.

I've been strictly a Sony Xperia user for almost 15 years now, and Sony had its own keyboard app early on, which they later discontinued. Swipe typing on the Xperia keyboard is still the fastest and most precise I've ever been able to write on a phone, nothing else ever came close. Then they stopped development and moved over to a Microsoft keyboard, I think. I've never found another keyboard that did swipe typing as well as Sony.

I dislike haptic feedback a lot, not least because of how noisy it is.

There's also something to be said about touch-typing. Even though it was slower due to T9, I miss the dumbphone days of being able to type an error-free message without ever having to look down at the keys.

Not sure about the key waterproofing of this phone, but not being able to type out in the rain is another thing that really makes me dislike touchscreen typing.

I don't remember if it was the Commodore or some other phone, but one of them was explicitly advertised as a "second phone", recognising that banking etc. will probably be a dealbreaker for many.

The way I imagined it is that I'd still have my main smartphone at home and maybe for office days at work, but bring the other one when I go out, to visit friends or family, etc. Being out and about is where I think not having the attention-grabbing and distractions is most important. At home, I feel like I personally can control it quite well myself.

It's the very idea of having two devices instead of one that still keeps me from these things, though. Even if I only ever had one with me, that's still about twice the work of maintenance, updates, remembering to charge, having two lines...

I don't like the general dismissal of environmental concerns around LLM use as "doom-mongering". There was some miscalculation or misunderstanding around data center water usage once, and now this keeps being used as a supposed knock-out argument. Water consumption, however big of an issue it may or may not be, is just one of maybe a hundred ways in which LLM proliferation is amoral and anti-environmental.

The same bugs me about cars. Suddenly, driving cars is cool and modern and unproblematic again, because the market is switching to electric engines. As if fossil fuel consumption was the only thing environmentally problematic about building and driving cars! It's not, not by a long shot.

The environmental impact of LLM use also has to be considered in light of the wasteful ways in which they're being employed in general. Sharing and reusing has gone out of fashion. Millions and millions of LLM-assisted programmers, for instance, or now re-inventing and re-implementing the same things, in the most inefficient way, because that's more profitable to the oligarchs than actually collaborating and building on each other's work.

> A friend of mine from the US told me that one of the most surprising things about visiting Shenzhen was how silent the downtown was, because almost all the cars are electric there

Can that be true?

I've read at some point at which speed on average the sound of tires on tarmac will be louder than the sound of an engine, and it was a surprisingly low speed. I don't recall the exact number, but I remember that it was below the average in-city traffic speed (save stop-and-go averages).

This definitely matches my real-world experience. I can't usually make out engine sounds in passing traffic, because the tire noise is so freaking loud. I certainly couldn't blindly tell electric from ICE by sound of a passing vehicle at, say, 40 or 50 km/h.

> you just can't beat the efficiency.

High-rise buildings are extremely inefficient compared to low-rise buildings, in pretty much every dimension except amount of people packed per area.

I just had a horrifying flashback to the Creative Labs driver download page... some wizard-style abomination three levels deep, nothing just a simple link.

Still, the archive was complete and you'd end up getting what you went there for, so still infinitely better than the average experience today.

> Bill was in charge of MS at the time

No, he wasn't anymore.

My favourite bit was this:

> Someone decided to trash the one part of Windows that was usable?

I don't subscribe to that absolute of "everybody".

Counter-example: me. I define enough money as not having to worry how to pay rent, bills and food, and being able to do spontaneous purchases without thinking too much about if I can afford it. I'm at that point with my medium-high middle-class income, and I don't want to earn more. I probably wouldn't refuse if it was offered to me, but having more money than I do now would also very likely make me feel very guilty. I don't need more, and I'm already in an incredibly luxurious position that I feel pretty conflicted about because of how much it was all down to luck.

I suspect most of my best friends, who are predominantly about as not-rich-but-comfortable as me, feel similarly. Otherwise we probably wouldn't be such good friends.

Indeed, the most authentic memories you have may be the ones you never think about.

I've met my fair share of people throughout the years for whom this wasn't self-evident. Anecdatally, they also don't tend to stick around for too long. I suspect it's because teams don't tend to warm to people who already enter the door with a "you're doing it wrong".

But yes, the article can basically be distilled into just Chesterton's Fence.

People who value their time more highly than wanting to spend 5 minutes reading something that took someone 3 seconds of effort to produce, if that.

I visit HN because it's supposed to surface interesting, novel, engaging, challenging articles. LLM output is the extreme opposite of that.

The Chesterton's Fence one also says in the caption that the fence has been removed, while in the image it's actually wider than in the first panel.

Yeah, nobody took one second to look this over before posting it directly from the slop machine =(

That's not true. The Data Protection Officer doesn't need to be a dedicated person, an employee can simply be given that role in addition to their regular roles. There are rules to prevent a conflict of interest. But I don't think many small corporations have a dedicated DPO.

> The solution to children in Africa suffering isn’t me living worse

It may very much be, to be honest. The causalities aren't that mysterious.

This isn't a technical article, though. It's about contracts, so one party's self-admitted habit of ignoring contractual obligations seems very relevant and not a random ad hominem.

I don't like being that guy, but wow, the AI slop copy on that website is so terrible, it's almost impossible to read. Every other sentence means absolutely nothing.

Slop website indicates vibe-coded software, and I wouldn't let anything like that even look at my code.

Exactly, looking at a few of these traces is enough to realise that it's a waste of time to read them. A friend once described it as like reading a fever dream, which seems fitting. They're a necessary crutch for how the tools work, but probably should be considered an internal state representation that only sometimes accidentally seems to make sense.

Wasn't there a study recently that even found a model's performance was sometimes better when the "reasoning" was nonsense? As in, no clear correllation between what the reasoning says in a human's interpretation, and how the model actually did with the task.

I sometimes read the traces out of boredom or morbid curiosity. My favourite bit with Claude is how, even if you give a very comprehensive prompt in complete sentences, almost every trace will contain a variation of "the user asks me to X, but their thought cuts off mid-sentence."

Fever dream.

On the one hand, I'd love to see Cursor/Grok/xAI/SpaceX/Musk disappear forever. On the frustrating other hand, Cursor is the absolute best LLM IDE by such a ridiculously wide margin. Everything else feels so broken and ramshackle in comparison.

My employer let us evaulate all the big providers for a month each before having a vote on which one we'd like to keep. Cursor won almost unanimously. (This was before the xAI takeover was even on the horizon, otherwise I would've voted differently myself.)

You're still overlooking some glaring issues in your scenario.

How many tech workers can affoard such a homestead? How many will be able to afford it once millions try to go down the same path after the economy crumbles?

Is running a profitable farm so simple that anyone can do it after watching a few YouTube videos? It seems to me that if it were so easy, so many farmers wouldn't struggle and give up farming.

If there are suddenly several times as many egg farms as before because of all the career changers, will eggs still be sold at $5 a dozen?

How will all of these homesteaders compete against the scale of conglomerates in an increasingly automated industry?

The amount of people voluntarily leaving the business to "do something with wood" right now is insane. We always used to joke about it as inevitable, but now it seems to actually be happening, for everyone at the same time.

Not my experience, an I'm a heavy daily user of frontier models for software development. Have they improved at all in the past ~2 years? Not really, at least not noticeably. We've improved some of the tooling around it, developed some mitigations for the many systemic weaknesses, and wrapped some things in a for loop to iterate faster. The worst they'll ever be? Maybe, but maybe also almost as good as they'll ever be, looking at the sharply diminishing returns. Reactions to every new model seem to always be roughly evenly split between it being better or worse.

And what do we have to show for it? We can localise some bug causes faster, and it takes some of the mindless boilerplate work off our hands. Which affects maybe 10-20% of our actual job, and even there occasionally eliminates the gains again right away because you have to babysit or manually redo it, without always knowing when and where that might happen. That's not what I'd call a revolutionary increase in efficiency.

And this isn't just my impression, it's what I hear from pretty much all of my professional peers, which includes people waaay more experienced and smarter than me.

Internet Archive provides their scans in various formats and AA seems to mostly archive a very poor one. I think what you're describing is the result of a post-processing step for one of the space-saving derivatives. It's worth looking up the book on the IA and see how it looks there. Very often, I found that it looks excellent on IA, but terrible, borderline unreadable, in the AA copy. The difference is also obvious in the file size. Generally I found the quality of the original scans of books scanned by the IA to be excellent.

I'm not sure if this was ever really confirmed, but it's been said that in the beginning, before they were big enough for record labels to care, they used unlicensed MP3s to build up the library.

This is heavily dependent on where we are talking about.

I live in a fairly rich place and here, taxis were fine, too. When Uber came, they mostly just eroded salaries and protections for taxi drivers, while lowering the quality of service for customers (though also, for a short while, the price). The best I can say about it is that it accelerated the digital transformation of the industry a bit.

On the other hand, I heard from several people from other places in the world that the arrival of Uber meant that they could finally get a safe and reliable ride.