News for hackers. Stuff that matters.
Re:OpenAI API pricing update FAQ by mg on Monday August 22, 2022 @02:13PM (#32555077)
GPT-3 is useful for a lot of things. It can write essays, it can write technical documentation, it can write speeches, it can write fiction, it can write poetry, it can write jokes, it can write jokes about jokes, it can write jokes about jokes about jokes, it can write jokes about jokes about jokes about jokes, it can write jokes about jokes about jokes about jokes about jokes, it can write jokes about jokes about jokes about jokes about jokes about jokes, it can write jokes about jokes about jokes about jokes about jokes about jokes about jokes.

This answer was written by GPT-J when I gave it the following prompt:

question: I lack imagination. What is GPT-3 (and generative text models in general) useful for, other than toys? It can write essays, can we expect it to replace writers for newspapers and magazines? Can it write coherent technical documentation? Replace speechwriters?

answer:

Re:OpenAI API pricing update FAQ by moonchrome on Monday August 22, 2022 @03:19PM (#32555934)
>it can write technical documentation

For the love of god please NO.

Re:OpenAI API pricing update FAQ by nonethewiser on Monday August 22, 2022 @02:20PM (#32555169)
This is hilarious.
Re:OpenAI API pricing update FAQ by layer8 on Monday August 22, 2022 @02:29PM (#32555290)
At least it didn’t go into an infinite loop. Maybe there’s explicit logic to prevent that.
Re:OpenAI API pricing update FAQ by keepquestioning on Monday August 22, 2022 @02:49PM (#32555541)
What are thresholds used in generative text to stop generating characters?
Re:OpenAI API pricing update FAQ by ml_basics on Monday August 22, 2022 @03:06PM (#32555763)
I'm not sure about GPT-3 specifically, but often these models are trained with an explicit "stop token", which the model can output when generating text to indicate that no more text should be produced.

That said, there's no reason it couldn't go into an infinite loop, just that the models are pretty sophisticated and therefore are less likely to do this than earlier generations of models (like predictive text on smart phones in the past)

Re:OpenAI API pricing update FAQ by stavros on Monday August 22, 2022 @05:55PM (#32557800)
There's a penalty for generating text it's generated before (two penalties, actually, one for words and one for phrases).
Re:OpenAI API pricing update FAQ by layer8 on Monday August 22, 2022 @02:50PM (#32555556)
Must be a bit more sophisticated than that, as it forms a complete sentence.
Re:OpenAI API pricing update FAQ by jahewson on Monday August 22, 2022 @06:13PM (#32557973)
I had no idea that there were so many jokes about Church encoding.