AI vs. Having a Body

I’ve written before about what I called “non-corporeal intelligence”, but I didn’t reflect deeply on what having a behaving body means to a developing intelligence. Of course, AI does not exist in a body and only acquires intelligence indirectly from human productions. This has profound implications for the limitations of AI and cautions that should be taken into account when dealing with AI.

In other writings, I’ve talked about behavior that is ritualized — what I termed liturgy — and its importance in passing knowledge among humans, particularly in pre-literate societies. I mentioned that, even though we now have many ways of passing knowledge among ourselves, we still have brains tuned specifically to enable liturgical knowledge acquisition. It is easy to ignore this aspect of learning and its social implications, but understanding it is critical to comprehending current religions, political groups, cults, and effects of media.

While articles about “mirror neurons” are revealing and somewhat disputed owing to some claims made regarding them, the central, undisputed aspect of them is that primate brains include representations of behavior that are activated regardless of whether the behavior is self-produced or merely observed. In other words, our knowledge consists not only of things and their relationships, but also of behaviors. Furthermore, just as things that are often encountered become paradigmatic, behaviors that are often observed become ritual, and those that are highly regulated become what I call liturgy.

Going further with this objects vs. behavior comparison, just as well-known objects become the core of the world as we know it, frequently-observed and ritualized behaviors are also part of the known world. We constantly expect certain behaviors from others and create them ourselves as a matter of daily existence, but we rarely appreciate the deeper implications of ritualized behavior. It is these implications that AI can only infer from human productions because it does not have a body, nor a brain that is tuned to observed behaviors.

Someone once explained to me their feeling in Catholic mass of religious liturgy. Their feeling had nothing to do with the content nor referents of the liturgy, but merely the feeling of comfort and ease that they felt while it was being performed. My interpretation of this is that the comfort and ease was related to familiarity of the liturgy — the same as what we feel when we are at home, or when we encounter any familiar situation. The cognitive match between perception and expectation is high and rewarding. In addition, the fight or flight reaction we feel in dangerous or unexpected situations is not present.

It is natural to ask why feeling comfortable and at ease helps create knowledge, and the answer is that it doesn’t. What it does do is create social relationships, which, we’ve argued elsewhere, help to create knowledge. In the long-ago past, these social relationships aligned with categories of things in the world — totems, for example. These days, such social relationships align with sets of core beliefs. Whether or not these beliefs correspond to objective reality is immaterial for those affected, and there-in lies the problem.

Very little knowledge that one acquires today is a result of direct experience. We learn from others and from recorded knowledge. We also learn from groups that we associate with, either face-to-face or virtually, and liturgical behavior plays a large role in our relationships to such groups. Therefore, liturgical behavior affects the “knowledge” that we acquire, fact-based or not. AI is not able to behave, nor perceive behavior except through recorded results. Nor, can it distinguish groups based on liturgy. Thus, AI is unable to weigh the impact such behavior has on humans. This aspect demands scrutiny when we assess AI and what it can do, in light of what it cannot do.