This will only work with slang from before ChatGPT’s knowledge cutoff, though (2021). Any slang newer than that (or if it just doesn’t know) it’ll likely just make up an answer.
As always, take anything a GPT algorithm generates with a grain of salt (though it got it right in OP’s post).
make an updateable slang DB, tie it to knowyourmeme and other sources, have it extract to a vector db for use when prompting the model.
now it stays up-to-date and you correct bad translations. it would be capable of translation as well as using the encoding sets in any way you can think of.
This will only work with slang from before ChatGPT’s knowledge cutoff, though (2021). Any slang newer than that (or if it just doesn’t know) it’ll likely just make up an answer.
As always, take anything a GPT algorithm generates with a grain of salt (though it got it right in OP’s post).
make an updateable slang DB, tie it to knowyourmeme and other sources, have it extract to a vector db for use when prompting the model.
now it stays up-to-date and you correct bad translations. it would be capable of translation as well as using the encoding sets in any way you can think of.
Is this true using gpt4 with browsing? I feel like it would at least make an attempt to use newer knowledge in that case.
Ahh true, that would definitely help.