Dr. Randal A. Koene

Abstract:

A principal requirement for patient-specific neuroprosthetics, whole brain emulation, and substrate-independent minds is the ability to reliably and correctly convert acquired brain data to a functional model that meets a set of success criteria. The field of reconstruction of neural circuitry from connectome data and additional data is in its infancy, because computational neuroscience has traditionally focused on predictions for hypothesis testing. Of the few attempts made so far, most attempt to dive directly into the problem of reconstructing an unknown brain from its microscopy data, without specifying success criteria or metrics for meeting those. We propose that a step-wise approach, beginning with fully-known ground-truth systems and a set of well specified success criteria are a path toward developing test-driven methodologies and rigorous protocols for whole brain emulation. As a first step, we propose a standardized whole brain emulation challenge available to all parties who wish to substantiate the promise of their proposed methods.

About Randal:

Dr. Randal Koene is a neuroscientist working on the functional reconstruction of brain circuits via neural prostheses and neural interfaces. Koene introduced the concept of whole brain emulation and is a lead architect of the scientific roadmap for its technological development. Aligned with this mission, Koene has built founding R&D teams for neurotechnology startups. He is the founder of the research non-profit Carboncopies Foundation, was the lead architect of brain data analysis company Kernel, software architect of high-throughput electron microscopy company Voxa, which provided the electron microscope technology that acquired the 10 petabyte cubic millimeter mouse brain connectome data recently published in Nature, and is now the cofounder of whole brain emulation company Nextup Neuro. Dr. Koene’s publications, presentations and interviews are available at http://randalkoene.com.

Category: 

Emerging and Frontier Computing (System Identification and Translation)