Hacker News new | ask | show | jobs
by daniel-cussen 1521 days ago
Yeah so I'm peddling the engine for this shit, the matrix multiplication, https://www.fgemm.com, and you know it's just a Markov chain. What can I say.

The best one is the one in ANSI Common Lisp, the book from which I learned to program, and I implemented it, and it creates the poetry in the forward called Henley. That's a good bot right there, and 1994 tech to boot.

The more you spice the algorithms up the more intimidating at first blush, but if you dig deep the more fancy the model the more embarrassing the fuckups, rarer though they may be.

Contrary to how I started this post, there is a way to make it a bit better, so here I'm going to make an almost shameless plug, showed some shame at the start. More matrix multiplication means fewer embarrassing machine learning fuckups. Not none, never none, but fewer. So that's expensive, so you want something cheaper, and that's full of error, you want precision. I intend to charge half the going rate, with naive error, which I can do because I have a better algorithm for multiplying matrices.

Stay posted, there's nothing on that site yet, and in fact I'm not interested in tracking visitors anyway, gauging interest by feel instead. Reply to this.

1 comments

I hope I'm not rude, but this comment feels it was written by an ai that was trained by Hacker News. The paranoia is a bit high on this thread.
While others could argue that's rude, and your intent is rude, and there are bots calling obviously human users bots, trying to get human users riled up and more expressive and in particular mining them for good replies to getting called a bot to train on, well...

I will on an exceptional basis not take offense. Yeah that's a fairly good description of me come to think. An AI trained on Hacker News. So I used to be a normal human intelligence, normal human, but then got lobotomized, so no longer an intelligence, and then figured out how to recover brain function, artificially not the natural way. It was unbelievable, I actually got beneficial psychiatric medical treatment eventually, stumbling on a three-ingredient formula that for my specific case fixed the lobotomy, though not the memories...yeah just hung out on this forum a whole lot. Learned to code from PG's books, focus on Lisp, talking about startups, and then looking for jobs through the jobs page (while I aced all the interview problems, they treated me like Frankenstein's monster, in the same way, and for the same reasons). And then...algorithms. Developing AI.

Though of course keep in mind I have a meat brain, ordinary neurons, apart from the "Jorge Barros was here, don't tell the patient what you see" message written in my head like in shitty doctor handwriting. A "kick me" note. I hope someday I can read my psychiatrists' notes, meaning have access to the them and figure out what they say, and overcome the lobotomistic blind spots. There's hundreds of pages, and I'm entitled to them, they just won't show them to me in the spirit of "fuck the law", and left tons of blind spots so I can't communicate properly with lawyers (but judges yes, that they forgot to erase). They're the real AI researchers, lol.

Haha, okay then.
Look at their profile and other comments.