I've noticed a number of people using AI Dungeon to test GPT-3's abilities. While it's a great way to see how GPT-3 can power an interesting application. It's a poor test of GPT-3's abilities in general. The first generation of any custom prompt is actually GPT-2.

3:31 PM · Aug 2, 2020

20
28
11
172
This was put in place to prevent backdoor access to the OpenAI API. Additionally we have finetuned on a specific dataset and use parameters optimized for our use case making AI Dungeon not necessarily representative of GPT-3 in general.
2
8
1
49
Replying to @nickwalton00
You nave the infrastructure, why not offer a pure gpt3 experience as an option for subscribers? Might be quite popular!
1
0
0
4
OpenAI would have to agree to it
0
0
0
15
Replying to @nickwalton00
Good to know. It would be great to see what actually gets submitted to GPT-2 and GPT-3.
0
0
0
3
You’re unable to view this Tweet because this account owner limits who can view their Tweets.
I imagine it came up during OpenAI's security review
This tweet is unavailable
0
0
0
1
Replying to @nickwalton00
You wrote a whole blog post about how you’re using GPT-3 in Dragon. I bought the premium version specifically to get GPT-3. Where did you disclose this previously?
2
0
0
24
Here’s your blog post and settings screen for posterity — you claimed the model runs GPT-3. You clearly knew this is the reason people were paying. I suspected from the output quality you weren’t running GPT-3 and now I know why. I’m cancelling now.
2
2
0
22
Replying to @nickwalton00
Damn, then GPT-3 must be EXTRAORDINARY!
1
0
0
0
I managed to get Dragon to code a full (very basic and buggy) Java program at its will. According to Dragon, it should ask the user for some numbers and then output which ones were even/odd, it even compiled and run the program and showed me its output. When I run the program...
1
0
0
1