Neuromorphic Brain Chips Microchips Hardware

DARPA has a five year program called Systems Neuromorphic Adaptive Plastic Scalable Electronics or SyNAPSE for short. I already previously mentioned this project that the US military is undertaking in another blog post. Neuromorphic is basically the creation of digital neurons on a microchip that mimic the functioning of real neurons. Meaning they are creating computer hardware that can replicate how a neuron works and interacts with other neurons, like the synaptic connections and firing patterns. Ultimately it is hoped that neuromorphic chips will be able to process information in much the same way that a real brain can.

The SyNAPSE project itself consists of 5 different phases. The first Phase 0 will take approximately 9 to 12 months. In this phase they will demonstrate the basic architecture of the neuron and synapse development on a chip. This will be done in order to model the chips interaction with a sensory environment. In this first phase they are going to need to demonstrate that the technology can be scaled up to higher densities and that it has a low power consumption. They are also going to develop hardware capability to model spike time dependent plastic synapses.

Apparently the goal of this project is to be able to replicate a rat or cat level intelligence on a neuromorphic chip. So when they mention scaling it up it made me wonder whether they are going to try to model a simpler brain first before they went on to a higher level of animal intelligence. The simplest brain to model that I can imagine would probably be an insect. Surprisingly, I do tend to think that insects are conscious as opposed to unconscious robots

Insect brains also have many things in common with animal brains as well, so it might be a good place to start. That's just speculation on my part, though. Anyways, I'm not exactly sure how beneficial having rat or cat like intelligence would be. Presumably they want to build an intelligence that would take limited orders. You obviously wouldn't want to have an artifical brain that was as intelligent or more intelligent than people right away.

In Phase 2 then are going to start running benchmark tests on the previously developed technology. Chips will have already been created that replicate the densities of synapses and neurons. In this phase they are also going to start testing the chip in a sensory environment. This could perhaps be done in a virtual reality training ground. A goal of this phase is that the artificial brain would be able to function in this sensory training ground and be able to seek out resources and survive. This sounds a little scary. They want to make a brain that actually wants to survive in a training environment. The first thing that came to mind when I heard about this were those von neumann probes in the movie batteries not included. So I can imagine in the future they might use this for military hardware that would be able to seek out resources (as in fuel) to maintain its own existance. Perhaps this would eliminate the need for continuous refueling on the part of its creators for an unmanned vehicle. In Phase 3 they will have created full-scale chips along with the hardware. They hope that the artificial brain will be able to have vision, touch, symbol manipulation and survivability in the sensed virtual environment. Other phases will consist of further testing of the chip.

The jury is still out though on whether this will actually accomplish anything. It seems like a chip that successfully replicates a brain would have to be concious. Can you really have a functionally equivalent chip that is not concious but still performs all of the functions that a real brain does? Conciousness is a complex emergent function of the many aspects of brain processes at multiple levels. It's not clear if conciousness can be replicated merely by modeling all of the computational capacity of the brain via artificial neurons and synaptic connections. I think it would really be amazing if they could get this neuromorphic chip to actually behave in a manner similar to a real brain. This goal of this program is like that of the Blue Brain project that I mentioned previously. If the brain-chip had a dogs level intelligence for instance, this would allow it to follow and carry out many simple commands. I'm not positive how useful an animal level intelligence would be to the military. Also, I wonder what brain areas they are going to try to model in order to successfully replicate an animal intelligence. Many brain areas may be associated with negative aspects of animal intelligence (such as aggression), that might not be such a good thing to model. Overall this is a fairly fascinating project that would have many implications if successfully carried out. One can even imagine a future neuromophic chip that replicates human level intelligence or greater. Maybe the technological singularity really is close at hand.