News for hackers. Stuff that matters.
Re:Ask HN: Why do so many assume we’re on the cusp of super-... by TylerLives on Thursday May 25, 2023 @02:27PM (#36074248)
>Progress in LLMs has come much more rapidly than expected

I hear this very often, but I'm not sure, what were the expectations 10 years ago? What are the expectations for the next 10 years?

Re:Ask HN: Why do so many assume we’re on the cusp of super-... by bjourne on Thursday May 25, 2023 @03:08PM (#36074744)
Almost to the date seven years ago: http://karpathy.github.io/2015/05/21/rnn-effectiveness/

"We downloaded the raw Latex source file (a 16MB file) and trained a multilayer LSTM. Amazingly, the resulting sampled Latex almost compiles. We had to step in and fix a few issues manually but then you get plausible looking math, it’s quite astonishing:"

My emphasis.

Re:Ask HN: Why do so many assume we’re on the cusp of super-... by june_twenty on Thursday May 25, 2023 @04:22PM (#36075681)
10 years ago we knew you needed a huge amount of data. These LLM are proof of that
Re:Ask HN: Why do so many assume we’re on the cusp of super-... by digging on Thursday May 25, 2023 @04:44PM (#36075954)
Well, it's what I've heard from many "doomerists" in the last year. It's not even a question of 10 years; ChatGPT at launch was pretty surprising to a lot of people. Could be many experts weren't surprised by ChatGPT, though, since we got Sydney and GPT-4 and others pretty soon afterward.