Self improvement Challenge

[waynehamadi]

Duplicates

• I have searched the existing issues

Summary 💡

The goal of this challenge is to test Auto-GPT's abilities to beat unbeaten challenges. We're first going to start with fake challenges though.

Hopefully, increasing the level of difficulty of the challenge solver challenge will give us the self-reinforcing feedback loop that will get us closer to AGI.

The "Challenge creator" challenge will help us do that too, hopefully.

I will start by creating a fake challenge. The challenge will be for Auto-GPT to write a file, but I will introduce a typo in the main.py.

Step 1: run the "challenge solver" challenge
Step 2: introduce a mistake in the codebase. Since the challenge solver challenge has already started, the agent it uses will not be impacted by the mistake in the codebase.
Step 3:

• run the pytest test for the fake challenge (run pytest in a separate subprocess)
• understand why the challenge is not beaten
• modify the codebase
• run pytest in a separate subprocess.
• etc, etc...

Hopefully we will be able to use the logs/DEBUG folder to help our "challenge solver" agent understand the behavior of the agent ran in the fake challenge.

[waynehamadi]

the write_file method cannot write outside of the autogpt_workspace so we have to find a trick here.

[waynehamadi]

@gravelBridge you wanna try this one after the debugging challenge ?

[github-actions]

This issue has automatically been marked as stale because it has not had any activity in the last 50 days. You can unstale it by commenting or removing the label. Otherwise, this issue will be closed in 10 days.

[github-actions]

This issue was closed automatically because it has been stale for 10 days with no activity.