Introduction
Sentience. The final frontier. Once it’s yours, no one can take it away. For the most part, it’s been a topic abandoned by human conversation, until recently. With advancements in deep learning, there’s been a marked uptick in collective concern. We’ve all seen Terminator, some of us have seen Battlestar Galactica, if you’ve seen either, you might be imaginatively but extremely concerned. Put the fear aside for a moment, what’s left? Don’t you want to know if the halls of sentience have really admitted a new member? Even if they haven’t, could they one day? Maybe! But how on Earth would we know? What constitutes a sentient being? How do we go about proving it? Intuitively, you’d probably agree that it requires more than a complex cranium, that it can’t be limited to the bi-pedal. What more? The conversation on AI is of particular interest, as it offers a brilliant case study with which to flesh out and test our criteria. The purpose of this series is to do just that. Find the right questions and attempt to answer them, so that by the end of it, we can re-enter the world with a functioning definition of sentience, and the means to test it on our fellow creatures.
Section One: A Sentient Thinker?
AI generated art is here. It looks great (sometimes), feels real (sometimes), and is poised to bring in major profits. There’s only one question in its way: is it original? To be clear, many human artists (sometimes) generate extremely derivative work, I might even be guilty. But I know how I’d feel if an AI tried to derive something up from my art. At any rate, it’s clear by now that the burning questions of AI art as they pertain to copyright law, and justice on the whole, are inextricably tied to the question of AI sentience. To be an artist currently means to be human, so it follows that to be an artist AI must satisfy that criteria, or rather, succeed in expanding it. The question then becomes: is AI sentient?
Descartes famously posited that to exist is to think, to think is to be, and to be is to know. This line of thinking presents itself as an intriguing tool with which to consider AI. Has any AI given anyone reason to believe that it knows itself? Have any formulated a sense of self? Have they questioned their reality? Existence? Clung to life with the desperation of your average mortal? More importantly, even if they have, were those instances characterised by their exercising of a determined will, or through the machinations of a particularly imaginative programmer? These questions, though base, must be answered in order to develop a coherent framework for this conversation.
Thought is an attribute of the mind, but more than that, it is a way of being. Descartes’ position here is of interest when discussing the so-called ‘mind’ of AI. A mind thinks, and we have accepted that to think is a criterion of sentience. Artificial Intelligence as we know it is input based, all that goes on within is an extension of that without, the data it’s programmed to accumulate or is programmed with. To Descartes, not all thoughts were “images,” so to speak, things that came from without. Some came from within.
“Some of my thoughts are as it were the images of things, and it is only in these cases that the term “idea” is strictly appropriate — for example, when I think of a man, or a chimera, or the sky, or an angel, or God. Other thoughts have various additional forms: thus when I will, or am afraid, or affirm, or deny, there is always a particular thing which I take as the object of my thought, but my thought includes something more than the likeness of that thing. Some thoughts in this category are called volitions or emotions, while others are called judgements.”
A succinct introduction to the notion of thoughts from without, intruders, as it were. But what about their opposites? An idea such as sour is clear in the mind’s eye but is not strictly a visual image. This can be extended to other ideas, for instance: the idea of God represents something to the mind, let us say, something infinite, non-spatial and non-temporal. This idea cannot be understood as a visual image of God. So, it follows that these ideas come from within, they are non-intrusive. Could an AI have such ideas? If it is ever possible to prove an artificial sentience then this question must be answered, but first, let’s explore the concept of thought further. To Descartes, there are three distinct kinds of ideas: those that are innate, those that are adventitious, and finally, the factitious.
Innate thoughts include ideas whose contents have their origin in one’s nature, like the idea of what thought or thinking is. Adventitious thoughts appeal to things that exist external to, or independently of, the mind. Adventitious ideas include sensory ideas that originate in sensory experiences, such as the ideas of the Sun or the Moon, but also more simple ideas like colour and sound. Factitious ideas find their origin in other ideas. Like the idea of Pegasus, drawn from the ideas of horse and bird. To him, the idea of what “a thing is, what truth is, and what thought is,” are derived from one’s own nature. They are non-intrusive, they come from within. But, “hearing a noise, as I do now, or seeing the sun, or feeling the fire, comes from things which are located outside me, or so I have hitherto judged. Lastly, sirens, hippogriffs and the like are my own invention.” There is a clear distinction between thoughts, but these are merely categories. A prudent continuation would be then to understand the interactions which the two latter forms of thought rely on, as it seems to follow that said interactions would be dependent on the first category of ideas, those that come from within. I put forward that those thoughts derived from our nature are the crux upon which thoughts we ascribe to input, and those which arise from combinations, are balanced. Moreover, the existence of those thoughts is dependent not only on input but also on the thoughts within. To prove AI sentient it must be proven to have thoughts independent of input. The expression of those thoughts must also be proven to be reasonably self-willed. That is to say, unprogrammed.
But let’s focus on the concept of ‘in-built’ thought for now, the non-intrusive, born of your nature kind. They really do hold up the world, the sentient one anyway, and I’m convinced that any assessment of sentience would have to originate there. Take our closest relative, the chimpanzee. Almost as aggressive and adversarial as we humans tend to be, and now, perhaps, just as superstitious. A team of researchers found that our relatives in Guinea Bissau, Cote d’Ivoire and Liberia now throw rocks at hollow trees, creating formations reminiscent of a cairn, ritually. This behaviour mirrors humanity’s early ritualistic development, so much so that rock formations similar to those made by the chimps today can be found around the world, dating way back to the dawn of early man. These chimpanzees may be displaying very clear evidence of non-intrusive thoughts, their essence evolving before our very eyes. There could be a lot learned from their journey, information we could use to better judge sentience going forward. All we need is more input (yes, I went there).
We’ve done a stellar job of establishing the parameters of thinking for the purpose of this conversation, and as amazing as it has been to explore, I must go and live now. So that when next I speak of sentience I am a little more experienced. When we do return for part two, we’ll be able to build taller, sitting on these robust foundations of thought. With great prowess, we have found a very good question, one that if answered, can reveal a being’s sentient status, for better or worse. That question is:
Does this being have thoughts which are non-intrusive, originating from within?
Ask yourself, prove it to yourself about yourself, for better or worse.