> What is the best way to talk about a startup idea in the very early stages?
> The best way to talk about a startup idea in the very early stages is to talk about its users.
Sounds about right!
Show HN: AskPaulGraham
61–70 of 75 posts
Re: Show HN: AskPaulGraham
#62Earlier quoted context omitted.
Are you using their semantic search API for pre-filtering response candidates?
I'm following along their question answering guide: https://beta.openai.com/docs/guides/answers
I haven‘t been active and don‘t know if they have changed this. But when I was using GPT-3, their semantic search endpoint didn‘t actually use GPT-3 for ranking. It used some other, „cheaper“ algo as a pre-filter (tf-idf, I believe). And only the those results that this „cheaper“ algo returned on top made it into the evaluation by the actual GPT-3 model you had chosen.
(There was a setting that controlled how much of your data was actually ranked by GPT-3 - top_k or something. But to chose anything even approaching a sizable percentage of a large corpus such as the one you are dealing with would be prohibitively expensive. You‘d be looking at several Dollars per search.)
Anyway. This is also why switching the model from Ada to Curie to Davinci on the semantic search endpoint has (or had?) very little effect. It still often missed highly relevant snippets from the source material.
When they came out with their „answers endpoint“, it really was just a semantic search piped into a completion prompt.
I actually found it better to keep my own implementation of this pipeline, because then you at least get to control how the results of the semantic search are prompted to the generation model.
Re: Show HN: AskPaulGraham
#63Just to point out the obvious in case some readers are not aware GPT-3 has nothing at all to do with state-of-the-art dialog systems/chat bots.
(It is a so-called pre-trained neural language model that can calculate the most likely next word or sentence, given some input sequence, very much "in the spirit" a statistical n-gram model.)
Re: Show HN: AskPaulGraham
#64> What do you think of sociology as a field? > I think it’s a very important field. LOL!
Makes me think of that scene in Archer talking about anthropology.
I'm pretty sure somebody's already named all the different spiders, Noah!
Re: Show HN: AskPaulGraham
#65Q: How much wood would a woodchuck chuck if a woodchuck could chuck wood?
A: A lot.
Re: Show HN: AskPaulGraham
#66Q: What has been your best investment?
A: My best investment has been a company that has been a good investment.
Very good.
Re: Show HN: AskPaulGraham
#67Q: What is eating the world?
A: Software is eating the world.
Q: Why is software eating the world?
A: Software is eating the world because the world is software.
Q: Why is the world software?
A: The world is software.
Re: Show HN: AskPaulGraham
#68Q: What is 1+1?
A: 1+1 is 1
Re: Show HN: AskPaulGraham
#69One glaring deficiency that is common among chat bots like this is the lack of any sort of learning/memory. I can get pretty convincing answers to some questions, but I can never get an appropriate answer to questions like "what was the question I just asked you?".
That would be pretty interesting. I wonder if there would be a way to smoothly integrate that with OpenAI's API.
https://youtu.be/VjK8j1C3IR4
Note the “For how much?” question, which refers back to the previous one about Zip2.
(This was done by keeping a log of topics and injecting into the prompt, along with the new question.)
Re: Show HN: AskPaulGraham
#70Q: What is the sound of one hand clapping? A: Sorry, I don't know the answer to that question. Q: What is the sound of two hands clapping? A: It is the sound of one hand clapping, and the sound of one hand clapping. Q: What is the sound of one hand clapping and how can it be monetized? A: One hand clapping is the sound of a startup that can't get funding. Q: What is the sound of two hands clapping and how can it be m…
> Q: What is the sound of one hand clapping and how can it be monetized?
>A: One hand clapping is the sound of a startup that can't get funding.
LOL, that’s some serious zen insight right there, brilliant!