I don't buy the story that the music industry is scared. As you mention, Spotify *is* the music industry. They control the conduit. There wasn't exactly a shortage of real music on existing catalogues before. They still managed to fuck over real artists, first by not paying them, then by replacing them with cheaply purchased library music wherever they could get away with it. They have no reason to be afraid. Non-industry people threatening their grift with GenAI slop? They have the means to produce a million times as much slop, and the power to make the non-industry slop disappear. The industry loves GenAI because it is just the next tool with which they can get closer to their vision of killing human creativity forever, replace it with their own zombie versions of it, and never pay anyone for anything ever again.

And exactly what do you think the point is of listening to music? Who sits down to listen to a GenAI album? Literally, what's the point? If there is no human behind it, no artistic intent, nobody expressing anything, why would I want to waste even one second of my lifetime listening to it? Do you think music is purely functional?

Except that authenticity really does matter. I have as much interest in listening to GenAI slop music as I do in reading made-up GenAI "news" stories. That I initially found it an interesting story doesn't change the fact that I'll deem it a garbage waste of time if I only later find out that it was all fake.

The problem is that people *don't* want to listen to it. The fact that they sometimes don't know what it is without being told, doesn't really factor into it.

I don't want to eat beef jacked up on hormones and antibiotics, even if my taste buds can't tell the difference. Declarations for informed decisions matter.

The Canadian one was meant to be CKUT.

There are ways to find it. Manually curated blogs, local events for unsigned artists... and even some great radio stations if you can find them. They're usually declared "freeform radio" because the hosts freely decide what to play, not algorithms or label payola. Some great ones include Nova in France, KCRW or KEXP in the US, KCRW in Canada, and some shows on BBC 6 Music (my information may be outdated). Also, some have dedicated programmes to promote young upstart artists (like BBC Introducing).

It matters a whole lot to a lot of people, including me. I listen to music because I'm curious what ideas and thoughts an artist might want to convey. I'm sure some AI music could fool me, but whether I'm aware of it or not, I'm 0% interested in regurgitated slop with no artistic intent behind it.

Compare it to news. Maybe an AI could generate news for me that would feel especially relevant and interesting to me -- if I don't realise it's all made up. I still have no interest in that drivel because what I was interested in was factual news.

At the risk of sounding elitist, I don't think anyone wilfully listening to GenAI music could rightfully claim to be interested in music. Those who just want some background noise were never music fans, even before GenAI.

But that usually works the other way around: the fake games shown in the adverts look a lot more involved, bigger, and diverse, while the actual game is another match-3 clone that shouldn't take more effort to develop than the 1-person shareware games of the 90s.

Refusing to do it would just mean they're getting fired earlier, no?

Have you been?

How would you scale up that ingest of validated, high-quality information, though? That stuff is extremely expensive and laborious to produce. The required amount of training data is massive, and returns are diminishing. And then there's the increasing awareness of copyright and compensation issues.

I see this working out for highly focussed specialist areas. But I'm highly doubtful about something like coding for a variety of reasons, such as the rapid pace at which technologies change, how protective companies are with their trade secrets -- and maybe the fact that I don't think the quality level of the average codebase is one we should deem acceptable :)

Do those revisions a few more times, then have a look at your entire codebase and what LLM modifications has turned it into. It won't be pretty. And the reason this happens is fundamental to the basic principles of LLMs. It won't be fixed with better prompting or improved models.

Universities need to prepare their graduates for a career of 4-5 decades, they need to take a longer view than the current fad MBAs and consultants are hyped about. In this concrete case, they need to educate graduates to be ready for the inevitable, upcoming "oh, shit!" moment of all the companies who went all-in on GenAI.

I admire your optimism, but I see the alternative path of an accelerating race to the bottom. Nobody bothering to improve anything anymore, because the others are just as bad. There are strong forces pushing the narrative that unstable and insecure software is just a natural law, and that it's absolutely economically impossible to engineer software well.

And those were control systems you were actually well-versed in, and you checked the resulting code?

I'm sorry, but I don't believe you. Likewise, I've heard *your* position a thousand times over the past years, and so far, there was never a wizard behind the curtain.

That's just wishful thinking and gaslighting by those who are trying to sell you their LLM services.

Have you actually seen and tried to use LLM-generated code? Worked with people who use it? Horseshit is putting it kindly. At least the previous ten iterations of "coding is dead" over the past 40+ years, were deterministic and somewhat forced people to still realise when they messed up.

LLM makes coding the opposite of dead because at this lap on the merry-go-round, underqualified users believe their horseshit code actually works, and they push it. Until it needs the first change, or someone actually checks the results, or a hacker takes half a second of a look at it. And because nobody hired juniors for a few years, few people will be around to fix it.

I'm not afraid of coding jobs going away, I'm afraid that a majority of the time they'll be about fixing the insufferable horseshit LLM code ldiots vibe-coded into the world.