Hacker News Reader

Comment by bluGill | Hacker News Reader

A week ago there was an artical about Donald Knuth asking an ai to prove something then unproved and it found the proof. I suppose it is possible that the great Knuth didn't know how to find this existing truth - but there is a reason we all doubted it (including me when I mentioned it there)

i have never written a c compiler yet I would bet money if you paid me to write one (it would take a few years at least) it wouldn't have any innovations as the space is already well covered. Where I'm different from other compilers is more likely a case of I did something stupid that someone who knows how to write a compiler wouldn't.

View on HN

So I would like to know how it found the proof. Because it’s much more likely to have been plucked from an obscure record where the author didn’t realize this was special than to have been estimated on the fly.

This makes LLMs incredibly powerful research tools, which can create the illusion of emergent capabilities.

Here's the PDF: https://www-cs-faculty.stanford.edu/~knuth/papers/claude-cyc...

It wasn't Knuth who used Claude, but his friend. Nevertheless, Knuth was quite impressed.

If you read https://www-cs-faculty.stanford.edu/~knuth/papers/claude-cyc... it was more of a guided effort to write a program to find examples that helped with moving the proof along

> as the space is already well covered

The US patent commissioner in 1899 wanted to shutdown the patent office because "everything that can be invented has been invented." And yet, human ingenuity keeps proving otherwise.

There are lots of small innovations left. Only a few patents have ever been for revolutions. Small innovaions add up to big things.

This is apocraphyl :(

I'd bet if you read the Dragon book (yes, I'm dating myself) you'd have something working in less than three months. More importantly, you would understand every bit of it.

Probably. I know what book you mean and never tried to read it. As I noted elsewhere I could probably brute force something in a week without reading the book. However the ai tried to be better than just a basic translator and that takes more time and exberience than I have.

You could probably do it in a few days, C is not that hard to compile

Right, and that was a design goal of C language... to be close to the machine.

Yes, and I was responding to

> it would take a few years at least

Claude built an optimizer as well. (Not a great one) that takes a lot more. Yes I could lively brute force a C compiler that works much faster.

And a few seconds more to write a stub `stdio.h` that will allow to compile at least helloworld.

Writing a compiler that can compile the Linux kernel is a bit more involved.

Famous last words.

C# back in 2000-2007 had a bunch of innovations. I expect we will have more.

I don't see any reason to doubt that plausible-next-token-guessing could sometimes plausibly-next-guess a sequence that happens to decode to the answer to some question we'd not yet solved.

... it'd be even more likely if, as other have suggested in this thread, we actually had recorded the answer in writing but nobody had noticed it yet, say, but even without that I don't see why it couldn't happen.

Reply on news.ycombinator.com