Neural Science Augments AI

Despite of proliferation of wearable sensors, current crop of fitness applications and personal assistants is not even close to meeting user expectations. Reason for this failure is lack of understanding of biology of human body and human brain, which AI technology strictly separates into two distinct realms of body and mind. AI assistants, such as Siri, are disembodied, relying on naive assumption that brain is a sort of computer CPU, “controlling” largely passive body, using algorithms and logic computations, much like pilot controls an airplane from the cockpit.

 

The good news is that there were plenty of AI successes over the last 40 years. The bad news is that with one exception, they all occurred in Hollywood. The exception is Demis Hassabis Deep Mind, acquired recently by Google. Their software can learn to play most ATARI 2600 games without instruction, just by watching the screen, playing with UI controls, attempting to maximize the score. The program uses deep neural networks, but more importantly, it uses real neural science in form of hippocampal replay which according to their white paper is “criticaly important”. At Embody Corp, we take this approach further.

Rather than starting with symbolic AI or deep learning neural networks, Embody Corp started with neural science and deep understanding of unity of body and brain, amply demonstrated in recent advances in neural science in work of Gerald Edelman, Gyorgy Buzsaki and Walter Freeman. Brain is complex indeed, orders or magnitude more so than AI technologists are willing to admit. Each neuron, far from being a simple On/Off switch is a complicated resonator, more like a Stradivari violin than a transistor switch.

 

Neocortex exhibits coherent oscillations of neuron populations at several temporal and spatial scales, forming not just static network structures of connected synapses, but also dynamic, short duration AM spatial fields emerging during fast, but continuous, non-linear phase transitions. Rather than starting with “information processing” and “computations”, we started with listening to and capturing the orchestra of hundreds of billions of resonating neurons, using not language of discrete digital modeling but language of non-linear dynamics and chaotic attractors.

 

Our brain is inseparable from our bodies and a plethora of next generation sensors entering the marketplace are perfect means for listening to our body, harvesting myriad of events in motor, visual, audio, tactile and other modalities and constructing meaningful understanding of unity of action of our body and our brain in context of goals of our present activities.

 

Using proprietory technology, Embody Corp assistant can actually capture the real, embodied meaning of my actions, understand what I am trying to do and help me with real time advice to accomplish my goals, whether they be to get more fit through exercise or refresh my memory with real time advice. He understands what I say, what I see and what I do, in context of my present task.

(c) 2015 Embody Corp