PROBLEMS QUEUE

[maxxflyer]

A list of all ideas/problems to be solved. Same as the output folder, it should contain all the inputs waiting to be solved. Anyone should be able to contribute to it by pushing requests.

[Torantulino]

How does this differ from Issues?

There's also a discord coming due to popular request where we can put a more casual list!😊

[maxxflyer]

is autogpt already reading the issues? I am talking about a db of problems waiting to be solved because not all the code is available.

[maxxflyer]

the machine should constantly check the list to see if the newly created output can be used to solve a previously unsolved problem.

[Torantulino]

Ah, sorry, I misunderstood.

So you're talking about a task-list for an Auto-GPT instance to work through?
It already keeps a short-term bulleted list, but I high-level overview task list is a great idea. :)

[maxxflyer]

actually... you made me think about the github issues list, as an alternative. it should in fact read all the issues and consider them as a possible overview task list too. or...it should read the discussions tab and discover new possible inputs on its own. Basically it should pull possible inputs from the discord server too. or from twitter comments. No input should be lost. This comment I am writing should become the next auto-task as well. :)

[rkeshwani]

I saw that a to-do list was started. So use github api to access the project, and review items in the backlog and it starts working on them? using the list as directives?

[edmondsylar]

@ALL
hello, hope this is not a bad start. This is a great project, hands down!! it actually got me thinking of the feasibility of a simple recursive-self-improvement model for it, the autonomy feature makes it almost completely possible to do it now, and in response to this, I think we could experiment with a kind of "autonomous service" that can keep tabs on the current state of the project (have the entire project source code as context), and also monitor the comments and issues raised , we can pass these chatGPT-4 and use the project's current state as context to engineer a solution or a suggestion for the team which they can apply or have the system patch itself. The implementation will definitely be more complex than my explanation but i hope you get the picture.

What do you think?