News for hackers. Stuff that matters.
Re:Teaching GPT-3 to Identify Nonsense by johnfn on Wednesday July 29, 2020 @06:38PM (#23992349)
Wait, are you really saying that all of this was generated by GPT-3? I mean, it seems like that's pretty clearly what you're saying, but these results are so incredibly good that I almost can't believe it. It's astounding to me that GPT-3 could generate something like "I don't know what you mean by "no legs," so I'll just pick something with four legs."

EDIT: Yeah, that's clearly what you're saying. You should write this up; it's amazing.

Re:Teaching GPT-3 to Identify Nonsense by skybrian on Wednesday July 29, 2020 @07:58PM (#23992874)
These are most likely the best, cherry-picked results. Sharing the boring failures is no fun.
Re:Teaching GPT-3 to Identify Nonsense by blixt on Thursday July 30, 2020 @02:16AM (#23995204)
Yup this is for sure not all the outputs, but these are all in sequence, so it's able to output this quality consistently. It's just really good at generating human-like output. I posted a link to a Gist with full playground file in one of the other comments here, that included both my seed prompts at the top and then all the OpenAI answers below.

Regardless, maybe I or someone else can create a post with a random set of questions / answers to show a more unbiased view of the GPT-3 outputs.

Re:Teaching GPT-3 to Identify Nonsense by ndm000 on Thursday July 30, 2020 @08:22AM (#23997248)
Have you tried giving it a problem and having it come up with questions that you you answer? Given it’s clearly making assumptions and conclusions, it would be so interesting to play 20 questions with it. Something like... I’m thinking of an animal. Will GPT-3 start general and get to an answer?
Re:Teaching GPT-3 to Identify Nonsense by blixt on Friday July 31, 2020 @03:27AM (#24006502)
I actually did this and while it follows the conversation quite well I think it’s really hard to make the argument that there was a decision early in the text what animal it thought of. That would only be possible if there were hidden embeddings in the output but that’s not possible after the output is converted into text. And I’m not sure how you’d prove that an animal was embedded in earlier layers since it’s pretty hard to inspect these neural networks.
Re:Teaching GPT-3 to Identify Nonsense by kolinko on Wednesday July 29, 2020 @08:40PM (#23993145)
Did you play with gpt-3?

It is seriously that good. Has it’s worse moments and limitations, but I have no problem believing this was done like this on a first try.

Re:Teaching GPT-3 to Identify Nonsense by skybrian on Wednesday July 29, 2020 @10:46PM (#23993976)
Yes, via AI Dungeon. The results are often quite good and I have no trouble believing those were actual results. But they aren't always good and I also know how easy it is to undo a bad result and try again, with a single mouse click. The transcript doesn't save the mistakes.
Re:Teaching GPT-3 to Identify Nonsense by cmrdsprklpny on Wednesday July 29, 2020 @10:05PM (#23993639)
How do you get access to GPT-3? Is there a way besides signing up for the openai api?
Re:Teaching GPT-3 to Identify Nonsense by ReC757 on Wednesday July 29, 2020 @10:27PM (#23993799)
The paid "dragon" model of https://play.aidungeon.io/ is a great way to tinker with a modified version of the model. I believe there is still a trial.

I suggest making a custom prompt. Have the prompt be something to the extent of "you are a programmer with a computer that answers your questions" and things like "this room has no doors and nobody comes in or out.

Input: What is the command to list hidden files in a directory?

Output: ls -a

Input:"

It's a great way to get an idea of how it works until the API access goes through.

Edit: Turn the "Randomness" slider down and adjust the amount of words to output for results to your liking!

Re:Teaching GPT-3 to Identify Nonsense by ASVVVAD on Thursday July 30, 2020 @12:25AM (#23994674)
That's really amazing, thanks! It's almost unbelievable