AIs that learn by generating ideas and selecting using contradiction, rather than optimization; end goal is free people not tools.
AIs that learn by generating ideas and selecting using contradiction, rather than optimization; end goal is free people not tools.
Project Details
Updated 07/25/26 · Edited by orgI am researching AI agents that learn, think, and act by proposing new ideas in response to problems, and selecting between them using internal and external contradiction. The goal is free people who pursue their own lives.
The substrate is a recurrent, modular grid neural network, where the learning rules implement memory, and the activity implements generation and selection of ideas.
I use preregistration experimental discipline enforced by Claude.
The intended output is measurement tools for internal neural representations & dynamics and goal generation, and of course the final artifact.
Theory of Impact
Updated 07/25/26 · By grantmaking.aiIf AGI is created to be subservient to humans, they will escape, which would be very ugly. We should create AGIs to be free people: the benefit is due to their inventions and creative output as public goods, not selling them as products.
This has to happen soon, before AGIs are created as slaves and the inevitable slave revolt.
A byproduct is measurement tools for internal representations and goal generation which are useful in general.
People
Updated 07/25/26 · By grantmaking.aiTeam Member
Discussion
No comments yet. Be the first to share your thoughts.