DARPA, memristors, nanotechnology, and an approach to AI

Foresight’s core concern is understanding the changes in the human condition that will result from transformative technologies—particularly high throughput atomically precise manufacturing and artificial general intelligence (AGI). A Foresight Briefing document from 1987 “Dimensions of Progress” makes the point that progress in fabrication ability (for example, through atomically precise manufacturing) can accelerate progress in design ability (through AGI) and vice vesa. A striking example of this synergy already provided by current incremental nanotechnology begins with memristors, a novel circuit element based upon a property that is insignificant at the microscale, but becomes substantial at the nanoscale. Mathematically, the behavior of memristors is similar to the behavior of neurons, suggesting they could be useful to build systems that work like brains (see this and this Nanodot post from this past April). Brian Wang brings to our attention a new DARPA memristor-based approach to AI that consists of a chip that mimics how neurons process information, which was featured in the December IEEE Spectrum “MoNETA: A Mind Made from Memristors” and discussed a couple weeks ago in a Next Big Future post. Brian emailed us his further thoughts:

IEEE Spectrum – DARPA has funded a new memristor-based approach to AI consists of a chip that mimics how neurons process information.

Nextbigfuture discussed how memristors look very promising for creating human scale synapse networks.

The goal of the MOdular Neural Exploring Traveling Agent (MoNETA) project is to develop an animat that can intelligently interact and learn to navigate a virtual world making decisions aimed at increasing rewards while avoiding danger. The animat, which is a virtual agent living in a virtual environment, is designed to be modular: a whole brain system, initially including fairly simple modules, will be progressively refined.

Trying to create artificial animal intelligence on the way to human.

Memristors will be used as analog synapses—CPUs and GPUs will be used for neurons (there also could be custom chips)

DARPA SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics) goals:

  • 1 million neurons per square centimeter
  • 10 billion synapses (memristors) per square centimeter
  • 100 milliwatts per square centimeter
  • total power 1 kilowatt

The total system would then be about 10,000 chips with a combined 100 trillion synapses and 10 billion neurons. The human brain has about 100 billion neurons and 100 trillion synapses. The human brain is 50 times more energy efficient than the DARPA Synapse goals.

DARPA is trying to get to this in 5-8 years.

I have looked at the memristor – synapse before: http://nextbigfuture.com/2010/11/mind-uploading-brain-emulation-and.html

Consciousness transfer—What I think might work. Need to perfect mind machine interfaces and have brain prosthetics and integrated co-processors. You would need systems about an order of magnitude or two beyond the DARPA targets.

Then if you have your wet mind and the computer mind extension working together for a decade and the computer appliance had the same or higher level of complexity the result could be your consciousness might exist and span the substrates (both in the brain and in the artificial systems). So if you lost the biological substrate then you could have consciousness that continues in the machine portion.

Leave a comment