Sunday, June 16, 2013

GF 2045: Second Sunday Session

Randall Koene

Title: Whole Brain Emulation: Reverse Engineering A Mind.

Trying to take the big picture and look top down at how to understand the mind in order to duplicate.  Motivation for doing this is to avoid external catastrophes (meteorites, supervolcanoes, etc) - make ourselves more adaptable so we can cope with changes.  (Talk about projection - technology is the leading cause of change!)  Goal is to get to substrate independent minds.

Desire is to abstract the neurons to the extent feasible, without losing the large scale behavior of the brain.  Presentation seems like building a road map way too far ahead - all very abstract discussion of the tasks that would be required, but not grounded in actually working on a concrete project.  The neuroscientists were a lot more impressive because they are actually building stuff and dealing with all the concrete detail required to really make something work.

Mentions Candida Elegans again - 300 or so neurons.  Two projects working on simulating it, but confirms that we can't currently simulate the behavior of this organism.  That again suggests that individual neurons must be pretty damn complex.

Questions - gets asked about quantum consciousness and has a much better answer: it's much too soon to say, but it's not the most obvious place to start in trying to understand how the brain works.

Ken Hayworth

Neuroscientist that wants to be able to upload state of a mind by mid-century.  Central thesis is that human identity is the connectome - the state of connections between all the brain neurons.

ACT-R is the leading cognitive model of the brain today.  Believes consciousness is basically having a self-model in the brain, and this can be duplicated in a machine.

Discussion of different techniques for sectioning brain tissue to produce exact model of the layout of the neurons in the tissue.  Now discussing how to build a connectome for the entire brain using these kinds of destructive techniques.  Basically using electron microscopy to image the tissue as it's gradually ablated away with an ion beam.  This is too slow for a whole brain.  Uses a hot diamond knife to losslessly cut brain into pieces small enough to be imaged.

In the last year, neuroscientists have gotten to the point where they can produce an electronmicrograph of an entire mouse brain!

Has a very believable road map for this kind of destructive imaging of the brain.  Very unclear how this proces could be reversible.

Now calling for the ability to preserve our brains against the day when they can be imaged!  Has started Brain Preservation Foundation.

Witali Dunin-Barkowski (with an assist from a young graduate student)

Discussion of data formats used by which different parts of the brain communicate.  Something to do with waves of neuron firing - not enough detail to follow.

No comments: