Val - Self Reflection

Inspiration

This project is based on my old project Methodable, which is predicated on the idea: "What if one's cognitive routines were coded into a programming language, and what if this language could reflect upon itself, creating self-improving metaprograms?"

What it does

I added modules to Val, allowing Val to see its stack trace, and to inspect its own code, allowing Val to have a greater sense of its own inner workings, and the ability to reflect upon itself and then describe itself over email, leader to a richer sense of self on the part of the agent.

How we built it

I implemented three new "browser modules" - one for creating objectives, one for viewing Val's stack, and one for viewing source code for arbitrary modules.

Challenges we ran into

Weird python import errors!

Accomplishments that we're proud of

I'm proud the Val can respond eloquently about itself and can suggest improvements to its own source code!

What's next for Val - Self Reflection

I hope to leverage this level of self-understanding to create an agent that's fully self-organizing, able to adapt itself well to new challenging circumstances, and able to adapt along with a user, to model the user well and act as a high-integrity companion.

By the way, here's the commit where I started: https://github.com/dnsosebee/valuous/commit/1bf6a6fc433627a67e666fa4fcf744fbf7b890f3

Built With

• python