Hey community, did anyone try creating a self-learning AI agent that gets better with the task over time through the human based feedback loop?
Would love to know how well did that go and if there are any caveats/limitations to be aware of.
Main objective:
- If you give an AI agent consistent and clear feedback, it should get better and better at doing the same repetitive task - without needing complex, detailed instructions upfront.
- In other words: the agent learns by doing + feedback, not by following a huge premade instruction manual.
Please share your insights, learnings and challenges you had.
P.S. Would also appreciate some hints to where dig deeper on this topic (maybe even some SOTA research papers, etc?)