
I heard on the wireless yesterday that the current head of Microsoft lies awake worrying that people are forming relationships with AI tools as if they were human. He used the expression ‘seemingly conscious AI’ and I think that is a good place to start.
When , in normal circumstances do questions of consciousness arise?
Let us suppose you are out for a walk. Rounding a bend, you see ahead of you a body lying across your path. The first question that springs to mind is probably, ‘is that body alive or dead?’ (these are, after all, the two categories in which a body can be placed). The next might be, ‘if it is alive, is unconscious, or not?’ – for it might be a footpad, feigning injury to ensnare the unwary, or even a very enthusiatic entomologist, getting up close and personal with an insect that has caught his eye.
Once, again, these are the only two categories to which a living body may be assigned: it must be either conscious or not.
(there are inbetween states, as in the scene with Lady Macbeth:
Doctor: You see, her eyes are open
Gentlewoman:. Ay, but their sense is shut.
or, in John Donne’s words,
As virtuous men pass mildly away,
And whisper to their souls to go,
Whilst some of their sad friends do say
The breath goes now, and some say, No:
but these are not problematic)
On approaching more closely, we realise that a combination of aging eyesight and poor light has deceived us: it is not a body at all, merely a suggestive bundle of discarded clothing or a log or even a dummy. And at that point, the question of consciousness ceases to trouble us: while we supposed that what we were looking at was human (and indeed this applies equally to animals – had we taken it for a stricken deer, say) it was valid; but now, in the light of evidence to the contrary (it’s not a body at all) it disappears.
Let us approach by a different route. I have an online chess partner – let us call her K – with whom I have been playing for years. I have no idea of her age, appearance or occupation, and indeed my only ground for believing her female is her name. I know her to be an English speaker because we occasionally converse online in relation to our games. Now, though I am ccnfident that such is not the case, it is fair to say that all my interractions with K could very easily be managed by an AI tool. So how would I ascertain if that was the case?
The obvious course is to attempt to meet her in person. Yet even before I set out, I know the range of possibilities is binary: either there is an actual person K or there is not. If there is not, with whom have I been playing chess and conversing all these years? It might be that I discover that a group of pranky felawes (for some unfathomable reason) take turns to play the role of K, in which case there are actual people, but K herself is a figment and does not exist in flesh and blood; or I might find that K is a computer program operating from some university basement and that I am (somewhat unethically, as my daughter has pointed out) being used as the corpus vile in someone’s PhD experiment.
But if we take that last to be the case, would you continue to wonder whether or not K was conscious? On what grounds? Well, you might say, hasn’t K interacted with you for years in an entirely human fashion? But that is the narrow range of evidence that led me to suppose K to be human, with all the customary human attributes (of which consciousness is one). It is, if you like, an instance of synecdoche: I have been supplied with a part from which I inferred a whole – a part expressly designed to be taken for a human being in a narrowly-defined set of circumstances.
But now that I have established by my investigation that the part is all there is – that there is no person K, nor even sundry parsons pretending to be K – why would I persist with any speculation that K might nevertheless be conscious? The additional evidence that has come to light – including the fact that K has been expressly designed to be mistaken for a human being (and so, by implication, is not one) – all points in the other direction. Is this not as if, having learned to my asonishment that what I had taken for a library in a grand house was in fact a trompe l’oieil painting, I nevertheless persisted in wanting to go in and examine what books were on the shelves?
