So, I have been developing an experimental open source Cognition Architecture, QMind.

The idea is:

What if instead of intelligence being simply next-token prediction, it was a competition of interfering probability states within a symbolic graph memory system?

A few things I'm trying out:

- rather than very large embeddings, a graph-based form of memory,

- explicit reasons you can examine,

- contradiction coexistence/reconciliation,

- probabilistic “interference” between ideas, and

- symbolic + causal reasoning,

- ongoing change of world models.

Not quantum computing rather, quantum-inspired reasoning mechanisms put into cognitive architecture.

Basically:

What would spring from the melding of symbolic AI, graph systems, probabilistic inference, and bizarre cognitive experiments in a single, clear reasoning engine?

Not sure whether this will come in handy in the long run, but it has been a lot of fun to experiment with and a great way to express yourself.

Project is MIT open source if anyone wants to roast/contribute/build weird ideas on top of it.