April 15, 2017
Minimum Viable Brain
The human brain is the most complex organ known to mankind, yet we are still far from knowing everything there is to know about it. To attempt to recreate one is a very lofty ambition, and one that is certainly not the purpose of this project. What we are trying to achieve however is an artificial mind that takes the essence of the functionality of a human mind, but without emulating the same mechanical and biological complexities. We’re not so concerned with creating a highly accurate representation, but rather a smart model that we can utilise to eventually solve complex problems. The goal is to learn from the brain and take our limited understanding of how it functions and test our ideas with quick and simple prototypes. Through this process of researching the human brain and prototyping solutions, not not only are we able to build useful intelligent systems, but more importantly, we’re able to deepen our understanding of the human mind and its intricacies.
The Human Brain
Based on our current understanding of the human brain, If we were to separate it into sections that are responsible for specific tasks and processes, we’d get something that looks a little like this handsome chap:
As we can see, a large portion of the brain is taken up with processing vision. We also tend to have a lot of brain tissue that is responsible for the maintenance and functioning of the body to which the brain is attached. Luckily for us, our mind isn’t attached to a whole load of meat and bones, so we immediately save on having to reconstruct a large portion of the brain. In contrast, what we’re working with is essentially an empty shell. We can pick and choose which components we attach to our artificial mind. Of course some of these components are vital, but what is important to note is that bodily functions such as temperature control, movement, stability, digestion and breathing can largely be ignored due to a lack of a physical body. Whilst these aspects may play some part in influencing our thought processes, I believe it is safe to assume that we don’t need our AI mind reminded that it needs to take a physical dump every once in a while.
The human mind is a product of millions of years of evolution, obeying the simple rule that a mind that survives, thrives. Therefore it is highly adept at taking care of the body that it inhabits as well as behaving in a way that ensures the survival of future generations. In fact its survival instinct is the primary force that has shaped it into the complex sack of organic tissue it is today.
Computers are extremely powerful and versatile machines that allow us to do things that would be almost impossible without them. In some regards they are far more efficient than our brains, and are the closest thing we have to an artificial brain. However they are not enough to create an artificial mind. The difference between a brain and a mind is that a brain refers to a physical element, whilst a mind generally refers to the emergent phenomenon that occurs from the regular functioning of the brain. This involves being able to think and feel; something that computers are currently incapable of. Let’s examine the computer more closely to see why this is:
If you’ve ever built a PC before, this should look quite familiar. If not, then this diagram hopefully explains the basic functions of the main parts that make up a computer. All components (except the GPU) are vital for the computer to run. Without one, there is no computer.
When looking at the functions of the components, we can start to relate some of them to the human mind, but we are still lacking quite a few parts that are responsible for more subjective things such as personality, emotions, self, etc. On a computer, the operating system is the software that controls how everything functions. It would be the closest thing to the human self, yet is still far from it. The physical components of the computer are what allow the processing of the OS to occur. Without any one of them, the OS would not be able to run. Similarly, without the OS, the physical parts would be useless. Only when it all comes together does a computer function similar to the way a brain does when it is in a body.
If we were to roughly list out the functions of the human brain alongside the functions of a computer, we’d get a list like this;
The things that the computer lacks are mainly based around subjectivity. What is interesting to note, is that all of the functions that the computer is missing can be linked to self-preservation. Through its evolutionary stages, the human mind developed each of these characteristics to allow it to be better at adapting to its environment. We can thus assume that the notion of self-preservation is fundamental to developing mammalian-level intelligence.
The Self-Preservation Killer
Let’s take a moment to think about this. If the only way to build intelligence is to focus on self-preserving systems, we’re going to have a bad time. You can read a bit more about my thoughts on this in [another post], but to sum up, if we build true AI systems the way nature built us, i.e. on top of our intrinsic motivation to survive, we’re headed for a doomsday scenario in which true AI has an intelligence explosion to the point where humans are unable to comprehend its actions. This would inevitably lead to mass panic as we would no longer be in control, likely resulting in the human race taking drastic action to regain the upper hand. By this point, it would already be too late as the AI would have already far surpassed us and would always be many steps ahead. The future of the human race would then truly lie in the hands of the AI we built. Assuming we did program in a safety net for human-kind, once an AI begins to view humans as inferior, we would not be able to predict how it would react and whether it would continue to honour this directive.
My current answer to this dilemma (as naive as it may be) would be to instead focus on life preservation, where the AI would treat all life as equally sacred. Reasons for how I came to this answer can be read [here].
The Fundamental Components
The different components of an artificial mind can then be listed like this:
Based on the diagram above, we’re going to need our brain to take in some input, process that information and depending on the context, provide a response, creating a feedback loop. The inputs is a given. Without it, our mind would be unable to learn. For processing, I have identified a few key areas; we will need to implement some form of memory so that the mind can refer back to things it has learned. Just relying on an external database or cloud server wouldn’t be enough as this information needs to be specific to the individual that is retaining it. Of course it makes no difference where the physical location of that memory is, as long as it is exclusively assigned to one specific mind, which can read and write to at will. Along with the memory module, I believe it is vital to also have a short term memory module that acts more like computer RAM. Since memory isn’t stored here for long, the mind is free to use this space of the brain to retain important information that is only context specific and doesn’t necessarily need to be stored long term. Lastly, the mind also needs to have output modules that allow it to interact with its environment. These modules essentially make up the minds 'body' and it would be free to use them at will. Eating represents the mind’s energy source. No living creature can escape the burden of sustenance. The mind would need to understand this concept intrinsically as it would form the basis of its motivation. The mind would also eventually be able to communicate. Humans began to communicate through sounds made by our mouths, but there is no reason an artificial mind couldn’t use the most logical tool to allow it to communicate. In this case, I would assume a terminal would make the most sense for AI to human communication.