Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Dynamic expert spawning #5181

Closed
Pwuts opened this issue Sep 9, 2023 · 3 comments
Closed

Dynamic expert spawning #5181

Pwuts opened this issue Sep 9, 2023 · 3 comments

Comments

@Pwuts
Copy link
Member

Pwuts commented Sep 9, 2023

We know that a good match between the AI's given personality and its task can render better results.

Proposal

A function that takes a task: str as input and returns an AIConfig (name, personality, best practices) tailored to the task

Challenges

  • 90% of this deliverable comes down to prompting. This will take time to tune and iterate.
    • 💡 There may be opportunities w.r.t. auto-tuning the prompt using an LLM-powered feedback loop.
  • In tuning and optimizing the prompt, the risk of overfitting should be taken into account and mitigated (e.g. by using a medium-large varied set of test cases).
@Boostrix
Copy link
Contributor

Boostrix commented Oct 5, 2023

Wouldn't this be overlapping with/specialization of @merwanehamadi idea about dynamic prompting #3937

PS: Sorry if these issues are not meant to be discussed here, just let me know plz

@Pwuts
Copy link
Member Author

Pwuts commented Nov 14, 2023

I think you're right @Boostrix, thanks for cross-referencing! :)

@Pwuts
Copy link
Member Author

Pwuts commented Nov 15, 2023

Resolved in 36e2dae. This will need a follow-up to improve performance on specific verticals, but this issue can be closed.

@Pwuts Pwuts closed this as completed Nov 15, 2023
@github-project-automation github-project-automation bot moved this from 🔖 Valuable & Actionable to ✅ Done in AutoGPT development kanban Nov 15, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Archived in project
Development

No branches or pull requests

2 participants