Here are the high level goals / sub projects I had in mind when I started Project Phasor. How else can we facilitate the NeuroAI and Neuromorphic ecosystems together?

I want to elaborate on each of these soon. But before I do, I'd like to hear from everyone about what they think of each of these topics. Where would be the best "homes" for each of them? Who would be interested in tackling each of these? How can we inclusively get folks on board?

1 >>> Establish an early-access joint research fund of AI companies and investors big and small that attracts the best neuromorphic and NeuroAI research.
2 >>> Build a neuromorphic virtual machine that is easy to hack on with a focus on architecture search.
3 >>> Work together on an MLIR-based neuromorphic compiler dialect/framework.
4 >>> Consider all types of neuromorphic HW, including analog, in holistic heterogeneous solutions.
5 >>> Attack neuromorphic training at data-center scale, where efficiency wins will be huge.

We are organizing an open proposal for AMD to start a neuromorphic program of its own. The bones of it are good, but it could use some polish and supporting references,

To co-sign or express interest, use this form.

Here is a draft of the full proposal: link.

Comments, edits, and forks welcome! Let's aim to finalize the proposal by December 2.

Initial summary copied here for reference:
"""
Alarming progress indicates AMD must invest now – but how?

Neuromorphic computing (NC) has advanced significantly and presents a strategic opportunity. NC achieves superior efficiency by co-locating compute and memory, optimizing for sparse connectivity/activity, using stateful neurons, and harnessing the analog domain. While other players focus on productizing neuromorphic inference at the edge, AMD can become the leader in neuromorphic training with a cost-effective software-first approach. With the potential to disrupt AI, neuromorphics cannot be ignored and offer a rare chance for AMD to leapfrog the competition.
"""

The Neuromorphic and NeuroAI ecosystems have structural gaps and inefficiencies.

What collaborative efforts do YOU think really need to exist? Think big. Think meta. Think speed-of-light efficiency. We are just getting started!

To stay in the loop or help steer this effort, register your interest at: https://interest.project-phasor.org.