#ai #llm
Created at 2023-02-02
# [Anonymous feedback](https://www.admonymous.co/louis030195)
# [[Epistemic status]]
#shower-thought
Last modified date: 2023-02-02
Commit: 0
# Related
- [[Computing/Siri Apple Shortcuts ideas]]
- [[Computing/ChatGPT]]
- [[Computing/Intelligence/Large language models are not personal enough - how to fix it]]
- [[Business/Entrepreneurship/Personalized prompt engine as a service]]
- [[Computing/Ideas/GPT index and ParlAI]]
# TODO
> [!TODO] TODO
# My strategy idea to open source ChatGPT
The hardest part is data.
How do you get data?
Build something people want, that they get value from, collect data
[[Computing/Obsidian ava|Ava]] has an optimized UX for using LLM within a software that have also been optimized for thinking
Opt in setting to contribute to sharing your LLM usage to a public dataset automatically updated (hf hub) - ANONYMISED SOMEHOW
https://github.com/LAION-AI/Open-Assistant
## Detailed strategy
- Ava gets 1000 prompts daily
- Open Assistant expect 50k rows
- let's say 50% of people accept sharing their anonymised prompts
- that's 500 prompts daily
- let's say ai community x10 the scale of ava popularity
- 5000 prompts daily
- in 10 days we reach goal