Transmigrating human consciousness into a machine body: three arguments for the implausibility of this idea
One of the recurring tropes in transhumanist literature is the idea of “downloading your consciousness” into a robotic body. Despite its prevalence, however, I will now argue that this is in fact one of the more spectacularly implausible ideas ever put forth in science fiction. The more we learn from neuroscience, neurological medicine, and cognitive psychology about the nature of human consciousness, the more we find it to be intimately and irrevocably intertwined with the functioning of a biological organism embedded in a concrete social context. My skepticism rests on three main points.
1. The technology remains utterly beyond our reach.
In order to transmigrate human consciousness into a machine, the first prerequisite is presumably the ability to meaningfully extract and decode the total pattern of neural firings in a human brain. This is not necessarily impossible (in principle), but we need to take seriously the sheer magnitude of what such a scanning technology would actually require. Today our most advanced technologies for recording neural activity in a human brain are capable of tracking the firing of several dozen individual neurons at most. Most contemporary brain-computer interface devices operate by loosely reading the overall activity of particular regions of the brain, rather than the activity of individual neurons.
Your brain contains about one hundred billion neurons. According to today’s best estimates, each of these neurons makes on average about 7000 synaptic connections with other neurons. This means that your brain has somewhere between 100 and 500 trillion synapses. Developing a technology for keeping track of each and every one of those firing (or quiescent) synapses, and recording their activity, will be, shall we say, challenging.
Then comes the further requirement of meaningfully decoding what those synapses are saying to each other. This we can do today, at the level of two or three individual synapses at a time. Understanding what all the trillions of synapses are saying to each other requires not only an aggregation of their total activity, but – far more difficult – a complete model of the functional architecture of the human brain. We remain, shall we say, far from having such a model.
I am not arguing that creating such a technology is necessarily impossible. I am merely observing that we do not even know enough today to say with any confidence whether it is impossible or not. We are still so clueless in this regard that we can’t even realistically assess the magnitude of our own cluelessness.
2. A fully scanned brain doth not a person make.
Let us suppose, for the sake of argument, that scientists succeed one day in developing a full model of the functional architecture of the human brain, as well as a sufficiently sophisticated device for accurately reading and decoding the brain’s total activity. You would still be a long way from being ready to download my personhood into a machine. Why? Because the snapshot-pattern recording of my brain’s activity in this moment does not constitute or capture the essence of who I am. I am much more than the sum total of sensations, thoughts, emotions, impressions, focus of attention, and unconscious mental processes going on in my brain at any given moment. The real meaning of my personhood only unfolds over much longer periods of time, as my brain interacts with my body, my brain/body interact with my immediate environment, and my brain/body/environment interact with other people over time. It would be like capturing a millisecond of Beethoven’s Fifth Symphony on a recording device, and triumphantly proclaiming that you had thereby apprehended the essence of the entire symphony. Persons, like symphonies, are intrinsically diachronic entities, whose identity and meaning only reveal themselves as they play out in time.
In order to generate an accurate scan of my personhood, therefore, you would have to find a way to record my brain’s activity over many years of its development, self-modification, and interaction with my body, my environment, and with other people. This, too, is not necessarily impossible in principle, but it certainly sets the bar a lot higher than just stepping into some fancy machine for a quickie snapshot.
3. The recipient machine would have to be a full-scale functional replica of a human body and brain, embedded in a human social context.
Let us suppose, for the sake of argument, that scientists find a way to conduct such a diachronic brain scan – perhaps through a brain implant embedded early in your childhood and carried with you to record your brain’s total activity over many years. Now we presumably have a recorded trans-temporal pattern that comprises the interaction of all your brain’s synapses over a very long stretch of your experience as a person.
Meanwhile, scientists have carefully prepared a machine into which they now intend to download the “total time-pattern” that is you. Will it work? In Chapter 15 I suggest that human consciousness is irrevocably tied to embodiment in a human physical organism, and to a form of personhood rooted in a human social context. Take these two factors away, and you may indeed generate some sort of consciousness in your downloaded entity, but it will in no way be recognizably human. A mechanical robot body will not do, unless it faithfully replicates most of the principal capabilities and limitations that go along with the functioning of a human biological organism. This means, among other things, hormones, pain, emotions, irrationality, focalized consciousness, disease, psychological dispositions, hunger, inattention, hope, lapses of memory, and so on. But even this would not be sufficient. The machine would have to exist in a human social context, replete with parents, friendships, enemies, peer groups, jealousies, unfairness, moral codes, ideologies, culture, demands for reciprocity, labor, love, aggression, loneliness, and so on. Anything less than this would result, once again, in something quite different from human personhood.
At this point, therefore, we might reasonably ask: What’s the point? Why take all the trouble, if the result is just another humanlike person, not much different from any existing human biological individual? A proponent of downloading consciousness might reply that the new mechanical body was more durable or robust than a living biological substrate. It would also presumably be duplicable. For example: if you were about to get run over by a bus (and if you saw it coming), you could swiftly send your consciousness over the Web into a replacement machine body waiting for you at home. Perhaps. This would indeed be an advantage, but it would be a far cry from the vision most sci-fi writers and transhumanists have in mind when they imagine the possibility of inhabiting a machine body. They want all manner of new capabilities, well beyond the species-typical performance profile of biological humans. I argue in Chapter 15, however, that they would buy such a radically transformed performance profile at the expense of their own humanity. They might indeed acquire a completely novel set of capabilities, but they themselves would no longer be present to enjoy the experience. Their selfhood would have been left behind, in the scanning machine. The new entity would be deeply, utterly different from anything human.