Artificial intelligence is an incredibly important topic these days, and weak AI is already in use all around us. Neural networks and machine learning are being used in a wide variety of industries to autonomously perform tasks that would have been impossible for a computer to handle just a few years ago. But, even as sophisticated as those artificial intelligence systems are, they’re not truly “intelligent” in the way that animals, and particularly humans, are. However, a new chip that acts as an artificial synapse may change that in the future.
This novel chip is made up of an array of artificial synapses (Photo Credit: Armantas Melianas, Scott Keene, Stanford)
A strong AI would need to be able to think like you and I do. What that actually means, however, is difficult to pin down. Sure, we know it would need consciousness and sentience, but what do those terms really mean? From the perspective of hard science, no one knows. No one can even say for sure if consciousness is even real in the way we think it is, or if that’s just an illusion that our brains have evolved to produce because it’s beneficial to our survival.
Philosophers have spent millennia trying to come up with an understanding of consciousness, and they too have struggled. The best example of that, as it relates to artificial intelligence, is the philosophical zombie thought experiment. In this thought experiment, you take it for granted that you possess true consciousness. Now, imagine a person who doesn’t possess consciousness, but who could perfectly emulate the appearance of having consciousness. By all outward appearances and scientific tests, they are exactly like you. How could you ever prove that they aren’t conscious?
Once conclusion to come to after going through this thought experiment is that it simply doesn’t matter. If a being can emulate consciousness so convincingly that nobody can tell that they aren’t actually conscious, then they are conscious for all practical purposes. That may be an uncomfortable concept to consider, but it’s likely the conclusion we will eventually come to about artificial intelligence. But to get there, we first need to create a synthetic mind that can be that convincing.
To accomplish that, many artificial intelligence experts believe we’ll need to create a synthetic brain that works in the same way as a biological brain. That is because that biology itself may be key to our consciousness—or at least the way we perceive it. Your mind doesn’t exist in a vacuum, it’s an organ that is part of your body. The physical construction of the brain, and sensory input and motor output through your body, may be vital to consciousness. It’s entirely possible that that simply can’t be replicated with computer hardware as we know it.
This chip, which was developed by researchers from Stanford University, may eventually make that possible. Unlike a conventional processor that is made with many transistors, this chip has artificial synapses that more closely resemble batteries. That’s important, because it more accurately mimics how your brain’s synapses operate. This kind of brain-inspired design is called neuromorphic. In tests of the chip, it performed better than conventional processors.
They tested the chip with just nine artificial synapses. That’s hardly enough to evaluate its capabilities as they relate to artificial intelligence, but they could compare it to traditional computers. In those tests, the artificial synapse chip could carry out the same tasks while consuming just a tenth of the energy as current state-of-the-art computer systems. A large part of that is due to the chip’s unique construction, which combines logic and data storage into a single step—something that normal computers have to do in stages.
The physical chip may have had an array of only nine artificial synapses, but the researchers were able to simulate how that could be scaled up. In a computer simulation, they analyzed an array of 1,024 x 1,024 synapses. That’s a total of 1,048,576 individual synapses. Based on that simulation, they determined that a physical setup like could run off of small batteries, like what is inside your cell phone.
There was some concern that the physical construction of the chip, which is made from polymers instead of silicone, wouldn’t hold up as well as conventional chips. But the chip exceeded those expectations, too. Under ideal operating conditions, the researchers’ tests indicate that their unique chip is just as resilient as the ones we already rely on today. However, more testing will be required to see how well the chip works under adverse conditions.
Eventually, chips built on this technology may perform more efficiently than what we use today. That alone would be major news. But the more interesting aspect of this development is how it could be used in artificial intelligence. If strong artificial intelligence does actually require something akin to a biological brain, there is a good chance that it will be chips built on this technology that facilitate that.