Last time, we talked about emotions being intrinsic to sufficiently advanced AI. A closely related subject is whether a mind needs a body to be “complex enough,” including what you would call a body (which may not be at all the same as ours).
Let’s backtrack a little. The brain developed in animals and not in plants because animals must execute complex movement in complex environments. The brain is a prediction engine, attempting to predict the outcome of movement and adjust itself based on the results, via the feedback loop. The butterfly effect, present in sufficiently complex environments, makes exact predictability impossible in principle, even with infinite computing power (because arbitrarily small deviation, in initial conditions or even in where you round the results of computation, can lead to enormous divergence over time).
But one can predict in a way that, though imperfect and can backfire, works adequately most of the time. In our case, this mechanism–this very old brain optimization–is called emotions. In the previous post, I argued that sufficiently complex AI, when asked to solve sufficiently complex problems in unpredictable environment, will have to have emotions of some kind. Because they must interact with us–and because many of those problems are likely to involve work with our emotions–the AI emotions must be similar and recognizable by us. The same argument as why the parts of our spaceships have the same dimensions as a horse’s ass (because they had to be delivered by train, etc.–google it!)
But what about bodies? Will sufficiently complex AI need bodies, too?
My argument is that, emotions or not, a predictive engine would work best in an unpredictable environment if it is far from equilibrium–that is to say, if it exhibits the butterfly effect itself. A purely digital system is discrete, exactly reproducible–and therefore, predictable and near equilibrium. A far-from-equilibrium (FFE) system must be at least partially analog.
But I don’t advocate a return to analog computers, with vacuum lamps instead of transistors and the like. Digital systems have their advantages, and we should strive to combine the best of both worlds, analog and digital–not the worst. How? Let’s take a look at our own bodies. They grow in a process directed by our genes, which are digital. Yes, you have heard me right: our genetic code is digital. But it is “executed” in an analog environment that influences the way it actually proceeds.
Call this nature nurtured or call it analog noise injection, but the gist remains the same: the evolutionary process works best when a digital code is “implemented” within an analog “vessel”–the body. Embodiment introduces analog noise of sufficiently complexity to cause mutations that allow the evolutionary process to continue far from equilibrium–that is to say, toward lower instead of higher entropy, toward increased complexity. But the digital nature of the underlying code ensures fidelity of reproduction, not allowing the mutations to run amok so much that all stability is lost.
This is also why the next technological stage after (1) manual crafts, (2) manufacturing, and (3) 3d printing will be (4) growth–because evolution, when executed in the right direction (far from equilibrium), beats rational design. Exact reproduction (3d printing, for example) is inferior to growth, because the latter can incrementally improve its own design by means of evolution.
But for this, a body is indispensable. The body is the analog “vessel” that allows the digital code to thrive within. Conversely, any source of analog noise can be called a body for an AI.
In my own science fiction, the body of an AI depends on the AI’s principal environment. For example, one can think of sentient spaceships or submersible, with some part of the vehicle’s structure co-opted as the body for its mind. Those bodies actually grow from cybergenes, which are digital. Moreover, cyberhumans (the word “cyborg” is a swear in my world) also have cyber components in the brains and the rest of their bodies that grow from cybergenes, enclosed in special organelles (similar to mitochondrial DNA). Those cyber components plug them seamlessly into cyberspace, converting all emotions, thoughts, and beliefs into computation.
Thus, the marriage of the best features of both worlds, analog and digital, is achieved.