FizzyOrange @ FizzyOrange @programming.dev

Posts

1

Comments

464

Joined

2 yr. ago

LLMs can’t think - only generate statistically plausible patterns

Ah still rolling out the old "stochastic parrot" nonsense I see.

Anyway on to the actual article... I was hoping it wouldn't make these basic mistakes:

[Typescript] looks more like an “enterprise” programming language for large institutions, but we honestly don’t have any evidence that it’s genuinely more suitable for those circumstances than the regular JavaScript.

Yes we do. Frankly if you've used it it's so obviously better than regular JavaScript you probably don't need more evidence (it's like looking for "evidence" that film stars are more attractive than average people). But anyway we do have great papers like this one.

Anyway that's slightly beside the point. I think the article is right that smart people are not invulnerable to manipulation or falling for "obviously" stupid ideas. I know plenty of very smart religious people for example.

However I think using this to dismiss LLMs is dumb, in the same way that his dismissal of Typescript is. LLMs aren't homeopathy or religion.

I have used LLMs to get some work done and... guess what, it did the work! Do I trust it to do everything? Obviously not. But sometimes I don't need perfect code. For example recently I asked it to create an example SystemVerilog file for me utilising as many syntax features as possible (testing an auto-formatter). It did a pretty good job. Saved some time. What psychological hazard have I fallen for exactly?

Overall, B-. Interesting ideas but flawed logic.

Because I can read? Lol ok.

This is fine precisely because it is a blog post. If it was a scientific paper... sure maybe they shouldn't say that. But the meaning is abundantly clear from the context. There is no ambiguity.

There isn't. This is the colloquial use of "exponentially" which is very obvious from the context.

I don't think this is a very interesting article. We already know AI suggests nonsense a lot of the time. That in no way demonstrates that it is net-negative. In my experience it's a net positive even accounting for the times it gets things wrong.

Yes you do have to review its code closely. News at 10.

It is kind of funny that they picked an example where it made an obvious mistake for their hero image though.

No, the filter is correct even for UTF-8. Any ASCII character is exactly unchanged in UTF-8 (part of the reason it is popular). Since this code only filters out ASCII characters it works fine with ASCII or UTF-8.

This is 14 years old by now - hardly "modern". RAII itself is even older.

Presumably because Forgejo didn't have CI support until extremely recently. And because Jenkins is trash.

Depends what it's for. I think a simple CLI text editor like this probably shouldn't have any smarts. Obviously an IDE like VSCode or IntelliJ should.

I agree. The right way to integrate AI into this process is to pre-fill the "release notes" box with an AI suggestion, that you then edit.

I don't know if it's necessary a bad thing. Presumably these people were enjoying the book until they read this. It's kind of like the invention of the printing press. Sure, the content may not be artistically crafted any more, and there may be waaaay more slop. But I bet we will end up getting way more high quality content too.

You can't automate these tasks without AI because anything that was capable of generalising to automating such a diverse set of tasks is AI.

Unless you're never doing new development you can't automate them. The kinds of tasks I've used this for:

• Making an HTML visualisation of some complex function inputs. One-off project. I could totally do this but it would take me way longer.
• Formatting a complex and very long SystemVerilog file. There aren't any existing SystemVerilog formatters (and certainly none that would handle the insane level of ifdefs in this file).
• Writing a script to delete all Git branches at a particular commit. I only used this once.
• Writing an Asciidoctor custom annotation. I don't know Ruby so...

You can't automate any of those.

I think you'd have to be a very limited kind of developer - only working in some tiny niche - to make AI completely useless for you. Most programmers occasionally have to do tedious but simple throw-away tasks, or tasks in systems they aren't familiar with. AI absolutely can save you time in these cases.

Even accounting for that (at least in countries with national healthcare), they're definitely more expensive than regular employees.

The games programming industry is high stress, but apart from that it isn't. I don't think it's known for burnouts any more than any other industry.

AI can absolutely save you time, if you use it right. Don't expect it to magically be as good as a real programmer... but for instance I made an HTML visualisation of some stuff using Claude, and while it got it a bit wrong, fixing it took me maybe 20 minutes, while writing it from scratch would have taken me at least a couple of hours.

Yeah obviously. Whenever a company says "we can't get enough X workers" they implicitly mean "at the price we want to pay".

But that doesn't mean they were wrong. Programming is still an amazingly well paying and low stress career. Being replaced by AI is a little worrying, but I think by the time AI is good enough to really replace programmers, it will also be able to replace most white collar jobs - HR, finance, etc. - and society will have bigger problems.

Since when are contractors lower pay? Companies waste fortunes on them.

I find the need to have an account in order to contribute to projects a deal breaker. It causes too much friction for no real gain. Email based workflows will always reign supreme. It’s the OG of code contributions.

This is dumb. I have followed the simple 12 step process to set up git send-email and it was so much more hassle than creating an account on GitHub or Codeberg or whatever, and in the end the UX is much worse.