3 ms·

big if true? :) Exciting to see competition across LLMs for increasing context window size. I can't find updated pricing anywhere. Previous prices are here: h

big if true? :) Exciting to see competition across LLMs for increasing context window size. I can't find updated pricing anywhere. Previous prices are here: https://cdn2.assets-servd.host/anthropic-website/production/images/apr-pricing-tokens.pdf https://cdn2.assets-servd.host/anthropic-website/production/... but don't seem to be embedded directly on the Anthropic website. I tried messing with the URL (apr -> may/jun) but 404'ed.

kordlessagain 3y ago
> Exciting to see competition across LLMs for increasing context window size. Maybe. I think the debate is going to continue about prompt optimization vs. context window size. A while ago, I had a rather interesting conversation with GPT-3.5 about forgetting things. Knowing what to forget, or delete from the prompt, may be just as important as what to put in it. Putting the kitchen sink into the prompt probably isn't going to help much, past a certain point and it may be putting certain things in there based on time and context is a better strategy.
- SeanAnderson 3y ago
  Yeah, there's definitely diminishing returns. I just wanted to talk to ChatGPT about a game I'm developing. I have pages upon pages of product design notes and I'm not able to just copy/paste the whole thing in and start talking to it at 8k context length. There's not really duplicate information as far as I can tell since each section covers new topics. I'm sure there's a way to express the same ideas more succinctly, but I kind of want ChatGPT to do that for me rather than me figuring out how to do that just to interface the ideas into it.
  - kordlessagain 3y ago
    Hey! Try this: https://github.com/featurebasedb/DocGPT https://github.com/featurebasedb/DocGPT Holler if you want help with it. I have some more code I'm adding to it this week.

3 ms·

big if true? :) Exciting to see competition across LLMs for increasing context window size. I can't find updated pricing anywhere. Previous prices are here: h

by SeanAnderson 3y ago

kordlessagain 3y ago
> Exciting to see competition across LLMs for increasing context window size. Maybe. I think the debate is going to continue about prompt optimization vs. context window size. A while ago, I had a rather interesting conversation with GPT-3.5 about forgetting things. Knowing what to forget, or delete from the prompt, may be just as important as what to put in it. Putting the kitchen sink into the prompt probably isn't going to help much, past a certain point and it may be putting certain things in there based on time and context is a better strategy.
- SeanAnderson 3y ago
  Yeah, there's definitely diminishing returns. I just wanted to talk to ChatGPT about a game I'm developing. I have pages upon pages of product design notes and I'm not able to just copy/paste the whole thing in and start talking to it at 8k context length. There's not really duplicate information as far as I can tell since each section covers new topics. I'm sure there's a way to express the same ideas more succinctly, but I kind of want ChatGPT to do that for me rather than me figuring out how to do that just to interface the ideas into it.
  - kordlessagain 3y ago
    Hey! Try this: https://github.com/featurebasedb/DocGPT https://github.com/featurebasedb/DocGPT Holler if you want help with it. I have some more code I'm adding to it this week.