3 ms·

This is a great project. Is it bad that I look at every library now, no matter how awesome, and think "Well this didn't exist on or before September 2021, so d

This is a great project. Is it bad that I look at every library now, no matter how awesome, and think "Well this didn't exist on or before September 2021, so do I really wanna learn this?" I mean surely an interface of the form blah blah blah, "Now, take the content and turn it into slides using a document creation API" is the best right? Should every library include a documentation endpoint for Scraper Plugin of the right context length? Is prompt injection an attack, or an opportunity? This is new ground for adoption.

mixedmath 3y ago
I don't quite understand. What happened in September 2021?
- rhd 3y ago
  Chat gpt learning cutoff
  - mcmcmc 3y ago
    Wow, that’s an incredibly lazy and short-sighted take from the GP then. The training set will definitely get updated at some point… or you could just ask Bing, or Bard
    - aspyct 3y ago
      Or just actually learn the library.
    - gjvc 3y ago
      Agree. This is telegraphing the prospect of people reduce their efforts in favour of using LLM systems, and in time the snake starts eating its tail
cjonas 3y ago
This is a real effect where newer (and maybe objectively better) tech is now at a disadvantage because LLM hasn't been trained on it (or there aren't enough examples to yield good results). Of course being an "early adopter" has always been a challenge, but I suspect the lack of LLM capabilities will slow adoption even more.