Live data from Hacker News

GCC steering committee announces AI policy

lwn.net

451–454 of 454 posts

Re: GCC steering committee announces AI policy

#451

Earlier quoted context omitted.

I find myself often trying to understand the madness of some of the code an LLM produces. Does that count? You're absolutely right! I made a mistake, and I'm sorry. . Then I end up questioning myself if it wouldn't have been easier to just do it myself. To be fair though, it's not limited to LLMs, I've felt that way other people's code too. Jokes aside, there's a difference to understanding the code and understanding…

I noticed something about LLM code generation and this is as good as place to post the finding as any. It's fairly obvious, but as we know LLMs choose the highest probability next token, so when generating code, it does so from left to right, without ever reorganizing, which unless it uses a harness, that's not how WE write code, that's property 1. Property 2 is that it will write out as many boilerplate that occurs…

I think I get your point. I can't say that I've observed it writing boilerplate just for the sake of it, but it is disproportionately/abundantly present in the training set. One of the first rules they teach in ML is to balance datasets, even if it's with fake/interpolated samples.

The linear L->R generation is definitely a thing, it's much more costly for an LLM to iterate edits, where a skilled vim coder will be jumping all over the place, trying to make all the LEGO pieces fit.

The skill therefore relies on just being able to one-shot entire chunks of code correctly, and it's amazingly good at this... But even the SOTA models still have a lot of unused imports and unused variable declarations. They just have to "guess" what they'll need and hope for the best. If they include a mass of numpy/scipi/react/icon imports that they might need, it opens the landscape for them later on when predicting relevant tokens, reaching a more ideal solution.

It doesn't hurt to add imports that might be helpful, rather than penalize the solution because you haven't got them. Although the last few SOTA models are more "harness/tool aware", they're starting to have the instinct to write the code anyway, and to be allowed to go and fix the imports later via tool calls.

For anyone who's seen the film Arrival (2016), their entire language is formed of complete concepts, not sequences of words and time. I keep thinking back to this.

Re: GCC steering committee announces AI policy

#452

Earlier quoted context omitted.

> arguably the most important invention in human history. Don't be ridiculous.

It's an invention that can invent things on its own, improve itself, do research, etc. The importance of such a thing should be obvious. The hockey stick that's coming for human progress overall is going to make the industrial revolution look like a flatline.

> It's an invention that can invent things on its own

No. It can't invent things on its own, it repeats what it's already seen.

People keep seeing LLM outputs without seeing the original source first, then exclaim: The LLM invented it.

Re: GCC steering committee announces AI policy

#453
post #390
post #370

Earlier quoted context omitted.

Need to replay it, or try the sequel. Is it any good?

It is on par with Portal 2, both the original and the sequel, imo. Some people think the sequel took a step back but I don’t really see why, if anything I liked 2 better than 1.

Wow, that is high praise. I really should get around to it sometime, too.

Re: GCC steering committee announces AI policy

#454

Earlier quoted context omitted.

No, it's far-fetched to believe GCC would get sued for incorporating LLM-generated code.

You can't bring a steak to a vegan potluck and say "Come on guys, we won't get sued over this!"

Exactly my point. They say they don't want steak due to legal concerns rather than just admitting they're vegans.
Post reply on HN