AI is dead

There’s a lot of discursive kerfuffle around AI progress these days, especially with respect to so many implausible predictions that machine consciousness will emerge from advanced autocorrect and predictive text. Of course, the mere mention of the word consciousness is enough to generate a host of knotty disagreements, even as the same word lulls others into a comatose drooling contest. I still think it’s worth keeping this discursive kerfuffle in the back of our minds, even as we consider a related issue.

Here, I simply want to look at the distinction between live things and dead things; between life and non-life. This is less an argument than an exercise in definition. I’m aiming for a sound understanding of certain basic facts, even if some conclusions are likely to be implied. I’m well aware that perception is always already a form of judgment, but that only means that we should ensure that we are perceiving rightly.

I submit that if we have a solid definition of the distinction between life and non-life, this’ll help us to arrive at a fairly stable philosophical foundation to refer to as we consider AI and its meaning for people. If we can demonstrate, with adequate reasonableness, that in the end there is an absolutely unbridgeable gap between human beings and AI, we’ll have less trouble switching the machines off if or when they go rogue. We’ll be less conflicted when deciding in favour of the human against any desire to replace the human by technological means.

Even with so many different opinions flying around us, we’ll be able to divide the world into two groups with respect to AI: vitalists and mechanists. Vitalists, in this context, are those in favour of life; who may or may not find a use for AI but who would, at a moment’s notice, shut it all down if they thought life was threatened by it. They are life-appreciators and life-celebrators. Mechanists, in contrast, are those who confuse electricity and code with life and who, probably because they are numb to their own existential depths, should not be trusted in the least with deciding what life is or should be about. As far as I can tell, this latter group is driven by a number of overwhelming cognitive blind spots and are unlikely to be convinced of the distinction between life and non-life that I’ll be articulating here.

Actually, before we get to it, there’s another key reason I have in mind for clarifying the distinction between the live human being and the dead machine, based on a basic distinction between life and non-life. Whether AI can or cannot become conscious will likely be less important than whether people think (at scale) that it is conscious. I suspect people will be less likely to make this rather embarrassing mistake if they can better recognise what most of us know intuitively: there really is a stark difference between people and the dead computers or devices they’re using to converse with digital zombies.

Let’s not start our defining from scratch, though. Thankfully, philosophers have made various attempts to articulate the distinction between life and death, and we may as well stand on their shoulders instead of on their toes. Aristotle, for one, understood living things as having a soul that organises matter in service of self-nourishment. I like this. A live thing, in brief, resists dissolution and perpetuates itself from within. The leading biological definition of life similarly sees life as a self-producing and self-maintaining system. In other words, life is essentially autopoetic. Death, in contrast, means the irreversible breakdown of any autopoetic network; it’s when a system no longer regenerates its own parts or affirms its own boundaries.

This approach to definition is rather nifty, in my view, because it knows what to do with somewhat ambiguous cases. Viruses, for example, are not autopoetic and are therefore dead. Computers are similarly dead. For simplicity’s sake, I’m conflating deadness and non-aliveness, although I can also see a useful distinction between made here.

In fact, a number of horror movie monsters, such as the Xenomorph in Alien (1979) or ‘Calvin’ in the movie Life (2017), or any zombie, are clearly devoid of life by this definition. Such monstrous entities aim towards technical perfection in keeping with the momentum of a mere engineered mechanism. Such monstrosities strike us as horrific precisely because there is so much anti-life in their persistence. They act less like conscious beings than like machines. In them there is no fragility born of real dependence. There is no ecology or world for the dead thing. The horror exists, for those who are horrified and for those who aren’t, because of the reduction of life to a mere function at the expense of any consideration of an enworlded, ontological centre.

That said, a mildly more complicated but also more robust way of seeing the distinction between life and non-life is found in the work of Michael Polanyi, and it’s his definition we’ll be working with from here on out.

Polanyi, we should keep in mind, thinks of reality as hierarchically stratified. Actually, all of us do so, albeit in slightly different ways. At lower levels, in Polanyi-speak, there are physical-chemical laws that exist in themselves as self-contained. Such laws are completely indifferent but are discernible and explicable. Here we can think of things like gravity, how atoms stick together, how electricity moves, and how chemicals mix. These lower-level things exist ‘in themselves,’ which is to say they do their own thing without being phased about what they’re doing or whatever else is happening around them. Such things rely on higher self-organising principles and laws directed towards higher ends such as avoiding threats, replicating themselves, seeking sustenance and metabolising. Some personifying of impersonal things leaked into this paragraph. My apologies.

Anyway, this is to say that higher level things limit what lower level things do. They can restrict or harness them. Polanyi calls these restrictions or rules boundary conditions. Importantly, they can’t be derived from or reduced to what’s happening at the lower levels. Take a bicycle, for example. All the bits and pieces (metal, rubber, etc) that compose this remarkable invention are ‘in themselves’. The intentionally-created design of the bicycle is the higher level thing that gets all the parts to act a certain way. Obviously, its design can’t be explained by its materials alone. Yes, we can think about these things in other ways, say, via Aristotle’s four causes. But this simple distinction between higher and lower order things will suffice for our purposes.

Already in the above we have a few inklings regarding how Polanyi distinguishes between things that are alive and things that aren’t. Polanyi uses the idea of being-in-itself to define what is not alive. Entities governed only by physical-chemical laws or by externally imposed principles are beings-in-themselves. They don’t have any live core or soul or inner principle to integrate various parts into a self-referential, autopoetic whole. They have no real responsiveness to the world around them. They have no conatus essendi, to use Leibniz’s idea, and definitely no passio essendi, to think of the work of the magnificent William Desmond.

Beings-in-themselves have no intrinsic desire to strive, to achieve or to overcome failure relative to some intrinsic desire to persist. They are not sensitive to existence, whether their own or that of others, and they have no clue about their dependence upon other things. Things happen to them and they don’t do anything to either oppose or actively comply with what happens. They do not orient or maintain themselves for themselves. They couldn’t care less or more because care of any kind isn’t a feature of their existence. That’s a key idea: care isn’t a feature of their existence.

In contrast to the being-in-itself is the being-for-itself, which is alive. The being-for-itself is situated hierarchically above the being-in-itself because it enacts those operational principles that function as boundary conditions. You might say, as Fr. Robert Spitzer does in his book The Soul’s Upward Yearning, that “all physical processes within a being-for-itself are orientated towards the preservation and replication of the higher-order system”. And this is to say that “in living systems, physical processes are oriented towards objectives that lie beyond them — towards preserving the higher-order system with its higher level of self-organization.” Again, a mere being-in-itself has no interest in anything, including the higher-order system. It is indifferent to its own nature.

As already intimated, the being-for-itself indicates a core or centre of individuality, a sense-of-self (used in a very weak sense) that aims to achieve (even if it fails to achieve) its own continued form in genuine responsiveness (and not mere reactivity) to its environment. That’s a messy sentence, but in brief: the being-for-itself is ensouled and wants to keep itself together. Note, however, that this for-itselfness is not really even about conscious will. We’re dealing with something tacit, something intrinsic — something ontological. Even the simplest organism demonstrates the ability to respond to the world in a way that helps it to persist wherever it happens to be.

I like this distinction because it nullifies ambiguity rather nicely, and that’s exactly what we need to interpret the difference between even the most sophisticated machine and a person. For Polanyi, life is not merely an intense or complicated version of non-life but something qualitatively distinct from non-life. A thing either possesses irreducible operational principles, working through a centre of individuality in service of self-maintenance and environmental resonance, or it doesn’t. There’s no middle ground.

And it’s with this in mind that we can take another look at AI. For starters, every single thing about any computer or device that runs AI has been externally imposed upon it. It operates according to certain rules but it has no individuality, no for-itselfness, no self-generated desire to keep striving to be. A computer does not ‘care’ if it breaks because caring is not a feature of its nature. It has no tacit sensitivity to its own existence. It is entirely in-itself. No life is at stake.

The same goes for any AI that runs on a machine. AI is not a comprehensive biotic organism but a clear example of a contrivance. Although it is a complex thing, a sophisticated arrangement of physical-chemical processes by some higher-order entity called a human being, it relies on higher-level rules imposed from the outside. It serves entirely external aims rather than its own stubborn persistence within an eco-system.

To restate an earlier point, the boundary conditions of AI are established from the outside, by programmers, engineers and prompters. You might say that ontologically and teleologically speaking, AI is essentially parasitic. It is much more like a Xenomorph than a man. At best, AI is ‘in touch’ (a horrible personification) only with articulated human models of reality, and even saying this is a stretch. Being by nature out of touch with reality, and therefore lacking any sense of an ecosystem, AI also has no intrinsic centre.

To the above, I’d like to just briefly gesture towards the work of the phenomenologist Michel Henry, whose entire philosophy is built on a potent conception of life. It’s here that this gets more personal. For Henry, life is auto-affection; it is detectable in our ability to feel our own aliveness. Cut everything out of your thoughts for a moment, and turn inwardly to this surprising and amazing feeling of feeling alive and you’ll know the concrete referent that Henry is indicating. This is not merely biological life, seen and described from the outside, but phenomenological life — the real, immediate, intricate personal and radically immanent way that life reveals itself to itself. If I eradicate all interference and focus on the feeling, I become aware of the irreducible state of experience that is life-itself, which Henry holds to be the principle at the centre of everything.

In his book on Barbarism, Henry posits that modernity sets up various ways to turn life against itself. It makes it harder to stay in touch with our natural auto-affection. In more Polanyian terms, life stupidly elects to measure itself by what is non-alive: science, quantification, technology, mechanism, and so on. Henry says this is the true barbarism, which in obscuring auto-affection also sets up the conditions under which culture and morality will erode in favour of merely technical aims. Barbarism, in other words, is less about primitive savagery, at least for us today, than it is about sophisticated self-destruction. It’s the nihilism that destroys meaning even as it allows us to drive faster cars — possibly life’s attempt to make itself felt in the midst of so much non-life. This is not to claim in any way that science and technology are evil, however; rather, it’s to stress that they are relative goods that have been given far too absolute an importance. The hierarchy of being is inverted when what is higher is forced to serve what is lower. Life’s dynamic energy becomes monstrous when what is nearest and dearest to us is enslaved by bare functionality. If we could only learn how to feel again, this would be far less of a problem than it currently is.

Anyone who says AI is better than some human gift because it is more efficient is simply admitting that they have sided with non-life — I would say with death — against life. I don’t think AI is therefore to be discarded; I simply think its ontological status needs to be properly recognised. It doesn’t matter how amazing it becomes at doing things, its perfections are still rooted in a fundamental deadness.

Anyone who lets lower-order mechanisms like data optimisation and machine learning, all ‘in-itself’ things, dictate to higher-order life is acting suicidally. Anyone who claims that we are merely at the mercy of algorithms is doing the same: why remove from life what life can do, namely take responsibility for itself and strive for higher purposes. Of course, again, most of us intuitively know this. We feel our own aliveness in a remarkable way, and we even feel when we are being fed merely dead stuff. We are conscious of being alive because life is right there, at the very centre of our being.

AI, in contrast, doesn’t feel a thing, even as it steals the work of people, shreds the work of people, and hands back to us the mangled soul of what was once a human thought or feeling or experience. It offers so many scraps of disintegrated meaning, stitched together into the semblance of something that resembles (but is not) a lively coherence.

And yet, in the midst of this, we remain by default attracted to something deep that calls out to what is deep within us. We are attracted to a precious miracle: life. We see manifestations of life everywhere, always and everywhere echoing that part of ourselves that is ultimately irreducible in its intimate immediacy. This doesn’t mean that we live inside a zero-sum game, of course. Still, if things were to come down to a battle between man and machine, I will not be siding with the latter. I will be siding with life.

And that, I suppose, is that. Only, I want to close with reference again to the discursive kerfuffle going on about machine consciousness. If Polanyi is right, and I think there’s ample evidence to suggest he is, then what is known as the hard problem of consciousness has to contend, first and foremost, with the very hard problem of life and of living systems. As Spitzer writes, “If a living system cannot be reduced to physical processes, then how much more irreducible will be consciousness in animals and self-consciousness in people, which have even higher-order self-organization directed to even higher-order ends.”