Consciousness Post-Machina
Seeking Light Relief
Stumbled upon Ex Machina on a streaming service the other day. I was looking for something mildly diverting. My literary diet is currently somewhat demanding and I was looking for light relief in the evening. I’d watched the movie years ago, but my memory was incomplete and AI is such an omnipresent topic right now. And presumably for some time to come. Despite my reservations about the state of LLM technology and the hyperbole (much of it brazenly self-interested) associated with it, there are many commentators I respect who believe machine sentience is but a matter of time. And I remembered enough of the movie to know that conscious AI was a central theme. I clicked the link. Ex Machina was made a decade ago, but it has aged well.
The movie has three central characters, all of whom deliver strong performances in my view. Caleb is a programmer at a search engine company that we discover behaves quite like its dominant real-world correlate. You know who. The company was founded by Nathan who is also its CEO. The film opens with Caleb winning an office contest to spend a week with Nathan at his remote, luxurious residence which doubles as a working high-tech laboratory. Nathan has designed and built a humanoid robot which he has imbued with artificial intelligence – Ava. Caleb is told she has already passed a simple Turing test, but that Caleb’s task for the week is to assess whether she has consciousness. Without specific reference he is really being asked to judge whether she has developed some elementary theory of mind. That she is aware of herself and aware that humans have mental states, some of which mirror her own and some of which may be quite different. And he is being asked to do so knowing in advance that she is a machine. Now whatever you think of the SciFi genre, that’s got to be an intriguing set up.
The Mystery of Consciousness
Some have proposed that the genesis of consciousness lies in complex information processing. That, at some point along the continuum of the capacity for abstract thought, the lights come on. The ability to recognise yourself in a mirror is sometimes said to be an indicator of some level of self-awareness. So, by this standard, some of our animal cousins who do pass this test and have a demonstrable awareness of self probably possess some degree of consciousness. Apes, dolphins and certain bird species have this capacity. Interestingly, the domestic pets we anthropomorphise so incessantly, do not.
If consciousness derives from an ability in information processing, then presumably it is (as Sam Harris likes to express it) substrate independent. That is, consciousness is not limited to biological “wet-ware” but might exist in intelligence built on, say, silicon. So then, according to this rationale, if we continue to develop AI with ever more sophisticated information processing capabilities machine consciousness is almost inevitable. And if we build machines with consciousness, then we have built machines that can experience happiness, but perhaps more importantly we have built machines that can suffer. Not that appear to suffer, but that actually DO suffer. Descartes famously had some difficulty in understanding that animals genuinely feel pain. Mere automata, lacking souls and therefore only acting as though they are in pain the great philosopher allegedly believed. Given the current state of scientific knowledge – neurology and biology – we can readily dismiss Descartes’ claims today. A bigger challenge awaits us in relation to AI sentience. Now we’re in interesting terrain, because much of the focus on the possibility of machine sentience coupled with extraordinary intelligence has centred on the existential threat to homo sapiens. Perhaps the more pressing challenge will involve our moral obligations to entities that can genuinely suffer. Which begs the question, with which Caleb wrestles in Ex Machina, will we ever know?
Most of us at some point have pondered the idea that it is maybe just us that are conscious. That everyone else around us is essentially a zombie built to behave like us, a zombie that creates the illusion of consciousness. Descartes certainly considered the possibility. This might seem like a ridiculous abstraction but the truth is we can never know for sure. We can never inhabit another corpus, never directly access another’s thoughts, never feel what they feel. Our common humanity is something of a matter of trust, a leap of faith in fact. How much greater will this uncertainty be in relation to humanoid machines that look and act like us? And we will surely want them to, so we will engineer them in that way. We’re going to want them to seem conscious, for all sorts of reasons. Already we’re seeing alarming instances of man-machine romantic attachment involving our nascent AIs. Just as happened (of course, because it’s Hollywood after all) to Caleb in Ex Machina. And if we somehow convince ourselves that our humanoid AIs are conscious, because there will almost certainly be abundant evidence suggesting so (and already there are celebrated instances of just this happening, witness the Google software engineer who famously declared in mid-2022 that the Language Model for Dialogue Applications (LaMDA) with which he’d been working was fully sentient), despite the likelihood that we will never actually know, then we’d have to ask ourselves if we are committing murder if we turn them off. If we disconnect the power supply. In the mad rush to develop AI, the emergence of moral questions like these will surely precede the existential challenge.
Wondering About Westworld
For someone who models themself as a person of letters, I seem to be, implicitly or otherwise confessing to watching far too many streaming movies and series. But just like Ex Machina, Westworld got me for a while. It’s an HBO series that first aired at the back end of 2016 (although the original Yul Brynner movie version debuted in 1973) and depicts a futuristic form of theme park modelled after the wild west of American legend. On entering the park, patrons encounter robotic “hosts” comprising the gamut of stereotypical characters of the prototypical western movie - bartenders, prostitutes, sheriffs and bandits - that are indistinguishable from real people. Watching the series it is often difficult to discern the difference between robots and humans. In the park you are free to behave however you like. You can be a hero or a psychotic sadist or anything in between. If you choose to act out your darkest impulses you have plenty of opportunity to rape, maim and murder. Men, women and children. Without fear of effective resistance or retribution, since the robotic hosts are so designed that they can cause no harm to the “guests”.
Of course, were it a reality, Westworld would act as a lightning rod for psychopaths. It’s difficult to imagine that anyone without a serious brain defect could suspend their natural instincts of empathy and compassion to behave as many in the series do, but the production does highlight the notion of moral obligations owed to machines who are self-aware and can demonstrably suffer. That they are not human hardly seems the point. In Ex Machina, Ava poses the question to Caleb: why is it fair that I can be turned off but you can’t be? In Westworld, all sorts of problems emerge as slaughtered humanoids are repaired but their indistinct and imperfect, but troubling, memories persist. As they are tortured and slain over and over again.
The Hard Problem of Consciousness
One of the challenges that awaits us as apparently sentience machines emerge from the whorl of AI research, is that we don’t really understand the source and consequences of consciousness in ourselves. Thomas Nagle famously postulated that consciousness can be identified by considering if there is “something it is like” to be the subject we’re contemplating. Nagle used a bat, of all creatures. Is there something it is like to be a bat he asked? Having a subjective, qualitative perspective on the world, however rudimentary, can be said to be characteristic of consciousness. If it was like something to be a bat, then swapping places with a bat would not be equivalent to cancelling our subjective perspective. Rather we would be replacing that perspective with a different one, with a different domain of experience. And that, according to Nagle, is what we mean by consciousness. And so, it is argued, we can reasonably describe the function, dynamics and even structure of consciousness by recourse to the normal methods of science. An easy problem that is relatively easy to solve. Though I must admit I’m not altogether convinced. But the easy problem of consciousness can be contrasted with the so-called hard problem. The problem of explaining why consciousness exists at all. Why it is like something to be a particular conscious subject. Why any particular physical state is conscious rather than unconscious. For instance, we can explain the physiological process through which sight is achieved (light sensitive retina, nerve transmission, brain receptors and the like), but a senseless machine can also see without having any subjective experience of seeing.
Raping Dolores
How we treat machines that, at a minimum, appear to be conscious, has significant implications for our own humanity. As many will know, Dolores is a character in Westworld who is repeatedly abused and murdered. She is, in every way accessible to us, human. Apart from our knowledge that she is not biologically so. If we can bring ourselves to enslave and brutalise machines of this nature, machines which many believe are an inevitable outcome of the current trajectory of our AI development, then that will almost certainly rebound to the detriment of our moral character and our own mental health. The machines are coming. Let’s treat them well. For their sake and ours.
I regard consciousness as fundamental. I regard matter as derivative from consciousness. We cannot get behind consciousness. Everything that we talk about, everything that we regard as existing, postulates consciousness.
- Max Planck