Pretty good analogy because health damage from cars don’t just come from accidents, which are terrible and tragically visible, but from pollution which is a lot less visible yet kills even more.

But your analogy also breaks down because once an AI has written the code it’s very much concrete, right? The AI may choose to modify that concreteness into something differently concrete, but it’s not like you are in any doubt about what the AI comes out with (if you bother to check it, which you should).

I use to written asset systems during the AAA gaming part of my career and the elusive “[my] perfect file format” always eluded me.

Over the years I spent many a showers and long drives thinking about what the “perfect file format” needs and what kind of architecture would support this. I started formulating these into long text files and over time things crystallised into “yes this would work”.

I never had the time to write the code fully. A couple of false starts, coupled with “I should do this in rust even though I’ve been in C++ my whole life” meant it lingered and lingered.

The architecture felt right though.

Enter AI; feeding the spec text files to an AI made me resolve a couple of “ah, yes, but in practice this bit of information isn’t available yet” gotchas but the code worked. I worked in architectural layers, carefully checking input and output of each before proceeding to higher layers.

It worked and it was real affirmation to see 20 years of pontification made real.

But then I started profiling it. And hot damn, an LLM knows nothing about cache nor does it think about ways memory allocators and virtual memory systems work.

There was 200-300% speed ups to gain from hand tuning it and iterating with a proper profiler.

And all of this happened at code level, with real concrete code.

So yeah, an AI can be a wonderful tool. And the code it produces and iterates on is concrete. It’s not like the AI starts from scratch every time you have a thought. So the very, very first pass I is not deterministic. But the thing you deliver forwards very much is.

A JIT compiler generates code depending on the hot path driven by user actions - which are non-deterministic. The same happens when a compiler is deterministic but in a different way than developers think. Developers think they know what code is generated, but they don’t. A compiler translates a double loop or a stream into just one SIMD instruction - they think they know what code the compiler generates, but they don’t - so their trust in the compiler is for the wrong reasons. And even if compilers would be fully deterministic, and people would know the code they generate, compilers are so complex that determinism is not useful for verification - just because ‘someone’ looks at compiler output does not mean it’s correct.

https://www.tabulamag.com/p/non-determinism-is-not-the-problem

I have it create things for me that I would otherwise never have. I wanted an app for dream tracking. They were either pay walled, didn’t have the features I wanted, or both. Ai created what I wanted in a weekend. That’s just one example. I’ve created many single use apps as well.

If I’m at a takeaway store I take a photo of the board and it makes me an interactive menu that totals the price of what everyone wants and let’s us browse without trying to see the board and get in people’s way.

I don’t expect anyone else to use these. I don’t expect anyone else to want to use them. Though they’re welcome to. But if I can have a thing create exactly what I want and use it why not?

“but what if it has a bug and you lose a year of journals”. Well it created me manual timestamped backups as well as automatic timestamped backups to oauth Dropbox in a dedicated folder without the app having access to the rest of my Dropbox. It could crash everything and I would have every single thing safe in 2 places up to the most recent entry. And they’re just text files. I could take them anywhere if AI disappeared and the app broke tomorrow. I could just continue writing in notepad in the worst case until I found something else. But the app looks nicer and does interesting things.

The alternative to this isn’t me learning to code and spending months creating them. The alternative is them never existing.

The same argument can be made about calculators.

You could say the same about a compiler

I think AI is a tool but very advanced tool. AI is great tool but very unpredictable in many cases. I think it should be used because it makes coding more effective but as you say it kill every fun in coding. As Linus said tool is not important if code is great. If you used AI and its good then why not. Sometimes you can write worse code then AI. But on other side yes it can be hard for repository owners when they must reject tons of shit code created with AI. So it have advantages and disadvantages and when do you want some fun (or maybe some security stuff) don’t use it.

but local models exist, they are a direct threat to anthropic/OpenAi/Xai thats why we must invest our time to make proprietary models irrelevant.

I promise I won’t submit anything to the Linux kernel that was generated by a compiler, AI, or my own achy hands.

corporate propaganda

You’re spreading RIAA’s corporate propaganda right now:

intellectual property theft

Copying is not theft

“Intellectual property” only benefits the bourgeoise

If I ask "why did the LLM output “for” as the XXXXth word of its utterance, can Linus say why?

If you have the model that was used, and the seed used for the generation, then yes you can trace the weights and figure out how everything was generated.

AI is deterministic. And compilers can be just as “nondeterministic” if you don’t know what compiler was used and what flags were enabled.

There was another article posted on Lemmy somewhere where recently other kernel developers had to clean up after Torvalds as of recent.

So yeah. I’d say his brain has rotted.