Agenda
- 05:00 - How AI Dungeon is used with GPT3 engine
- 13:00 - Frequency penalty vs. Presence penalty Settings
- 18:21 - Doing detect and regenerate settings
- 25:30 - Customer influence responses
- 30:00 - Gpt3 to use in summarization
- 33:00 - Trevor exits the call
- 33: 50 - Nick’s experiences between GPT2 and GPT3
- 36:00 - Sampling Process
- 40:00 - Using machine learning for other things
- 41:00 - What is AI Dungeon?
- 46:00 - Quests
- 52:00 - Give a little overview on how AI Dungeon works now
Notes
- Fine tuning can make a great difference
- Safe mode - that’s on by default - reduces the chances of having an explicit token in the Space
- Detection on the user side - Child Friendly version