Absolutely fascinating. I'm going to play the devil's advocate here and say that this is the kind of research we need if humanity plans to survive the singularity of AI. Why? Because there should be no distinction between artificial intelligence and humanity. Otherwise, artificial intelligence will simply displace and replace humanity.
Now to be a pedant, I have to stress that strictly speaking this is NOT Artificial Intelligence we are talking about. It is Computational Intelligence as defined by the Institute of Electrical and Electronics Engineers (IEEE).
What's the difference?
AI refers to the use of symbolic logic (i.e. math algorithms) to try to break down and process intelligence with digital (number based) machines. Computational Intelligence, in spite of it's ill-chosen name, refers to computational methods that are much closer to biological computers (brains) and encompass analog computation such as Op-Amp computers and such - we sued to do a lot of analog computation before the digital revolution. Aircraft control systems, elevator control circuits, and just any type of electrical controller before WWII would be under computational intelligence.
The human brain does NOT work like digital computers. We don't use symbolic logic to process information. Our brain uses Neural Networks. Google is just simulating neural networks using digital software, much in the same way that SIRI's voice recognition subroutines are basically virtual neural networks. You could always take them out of software and hard-wire them into a chip that is not digital in nature.
These neural networks sound exactly as what they are. The mimic the neurons in biological brains. Each neuron is connected to many other neurons and communicate with a digital-analog hybrid method, specifically synapses firing, where each neuron's pulse is either zero, or some variable voltage (infinite number of possibilities). The response from the other neurons will depend no only on whether the input is zero or not zero, but also the intensity. An input of zero, an input of 0.2, and an input of 1 will all prompt different responses from the adjacent neurons. There will be an activation voltage which is like a threshold before your receiving neuron reacts. So say your activation threshold is 0.25, then neither the zero input nor the 0.2 input will illicit a response from the receiving neuron.
Depending in how you hook these neurons you will get different functions. Circular loops of neurons can store memories (Hopfield Networks). Tree-like networks (Feed Forward Networks) can apply fuzzy logic and recognize patterns.
Why do I know this? Because when I was in graduate school I had an interest in Cellular Automata and Neural Networks to create machines that would simulate fluid flow fields, with the neurons literally "transporting" virtual fluid molecules, like air or water.
So what? Well, the trick is this folks... A neural network is potentially MUCH faster than AI. What? YES. WE are slower than computers because we are made of squishy gelatinous salty matter. which requires electrochemical signals to communicate information. In other words, our brains are way too slow, because the speed of signal information is somewhere in the range of 300 mph, compare to the speed of information transfer in silicon circuits.
BUT as you know the human brain seems so much more capable than computers at many tasks. In fact on a brain complexity-to-problem solving ratio, Neural Networks are far more efficient than digital circuits, not in small part due to the fact that breaking problems into logical-mathematical constructs and the digitizing that information is very time consuming, and often redundant (multiplying zero times zero still takes computational power).
The reason digital computers out-pace our human brain is because they are made with silicon wafers. This is a materials problem. To put it bluntly digital computer's magic is all based on the material being used to transfer information, and the actual signal processing method (binary) is actually VERY inefficient at processing information -have you ever heard the expression "a computer can be the fastest idiot in the world?
So... I recognized that a neural network /cellular automata computer would simulate fluid flow MUCH faster than any digital computer could possibly do, PROVIDED I build the neural network in silicon wafers.
And that is how I came to know about Neural Networks.
So what does that matter to us?? Well. Very simple. You are all afraid of having AI take over the world, right? Pfft! Even smart people like Stephen Hawking, Elon Musk, and Bill Gates are all scared human intelligence is about to fall pray of AI.
O Ye of Little Faith, I say!! Pathetic.
If we can build a silicon-based neural network human brain, that brain will outpace ANY computer you put in front of it!
We don't need to be afraid of artificial intelligence. We need to OWN it. Arthur C. Clarke once said that "Any sufficiently advanced technology is indistinguishable from magic." Well a visionary by the name of Susan Armstrong also replied "A sufficiently advanced civilisation should be indistinguishable from nature."
Any artificial intelligence that humanity creates, if sufficiently advanced, should be no different than our genetic own descendants, for the simple reason that by caring for our old and sick people and living in sedentary lifestyles, we have basically STOPPED the natural forge of evolution as far as human biology is concerned (unless adapting your stomach to carcinogens and Taco Bell is considered evolution). Arguably some individuals have regressed a few steps back in the evolutionary ladder (Darwin Award anyone?).
Provided we don't kill ourselves (very likely as of late), then the only way humanity will advance in terms of intelligence is by way of engineering. Face the reality folks. The next step in the evolution of humans will be engineered. I'm not letting any computer take over my future or my children's future.
For starters, I want a copy of my brain in Silicon. Let's take it from there.
I remain AYS,
J. (that weird dude) Wilhelm
