by Oliver Waters
Panic about runaway artificial super-intelligence spiked recently, with doomsayers like Eliezer Yudkowsy prophesising that, if current progress continues, literally everyone on Earth will die.
In a way, he’s right. Humanity will probably die out as technology progresses, but not quite in the depressing way he imagines. ‘Humanity’ after all, refers to two distinct things. The first is our biological species – homo sapiens. The second is our collective cultural existence – the beliefs, attitudes, and creations that make us truly who we are. Our biology is merely a platform for our humanities to dance upon, and it is a platform that transhumanists wish to eventually supersede.
If all goes well, we will design and build better bodies and neural systems that more seamlessly integrate with our information technologies. On this trajectory, there will be no grand dichotomy between humans and ‘Artificial General Intelligences’. All persons, whether descended from apes or designed in a lab, will be constantly upgrading their cognitive systems to be ever wiser and sturdier.
The first big step on this journey will involve scanning your brain for all relevant neural information, which would then be loaded into a synthetic brain, attached to a comfortable artificial body. Science fiction has already imagined the countless benefits of existing in such a digital, computational brain. Practical immortality is at the top of the list, given you can ‘download’ your backed-up state of mind into a new body if you happen to be hit by a train or are notoriously assassinated.
The prospect of truly immersive virtual reality also has profound economic implications, given it makes the scarcity of most physical resources a thing of the past. It would be the ultimate step in de-materialisation – providing ourselves with better, richer experiences with far fewer resource requirements and impact on the natural world.
It’s understandable to cringe at this sales-pitch now, given the rudimentary current state of simulation technology. Who wants to put on a clunky headset to watch glitchy, pre-recorded footage of Mark Zuckerberg on his private beach, gesturing at you to join him in the MetaverseTM? No one. It’s lame, and there are no piña coladas. But in a simulation truly indistinguishable from the physical world mediated via an artificial brain, the drinks will taste just as good, if not better.
Our hardy synthetic brains and bodies could also survive and thrive in any physical environment, be it deep underwater or outer space, and would allow us to teleport anywhere at light speed by transmitting our neural data to wherever another suitable artificial body is stationed.
This is all huge…if true. There are obviously many unknowns about this speculative ‘transition’. What precise level of detail of brain information will need to be copied over to preserve one’s personality? And what kinds of other materials, if any, could support consciousness?
But perhaps the most troubling of all concerns is whether the artificial version of you will really be you. Indeed, all these wonderful benefits would count for nothing if you simply died and someone else entirely was created in your place.
A film that captures this quandary nicely is The Prestige (2006) by Christopher Nolan, in which magician Robert Angier asks the famous inventor Nikola Tesla to build a special device. This machine creates an exact copy of its user nearby, leaving their original body intact. Angier exploits this to carry out a spectacular magic trick, where he suddenly disappears amid electric sparks only to reappear instantly up on the theatre balcony. Unbeknownst to the audience, the version of him on stage has dropped down through a trapdoor into a flooded glass box, which then is automatically locked. As this version of him drowns in agony, the one on the balcony receives a standing ovation.
Angier decides to go through with this trick again and again, because he values the looks of awe on his audience members’ faces more than he fears the risk of really being the man drowned in the box. The question the film poses is whether you would make the same gambit. Do you think you would survive performing this illusion?
Obviously your original self and your artificial copy are not strictly identical, since they exist in different locations and are made of different physical stuff. But of course, by this standard, no two things in the universe are ever identical. This means we need to let go of the mathematical concept of identity (i.e., ‘1=1’) if we’re going to talk sensibly about identity in the real, physical world.
Take the example of two five-dollar bills. There are some senses in which they are the same, and others in which they are not. The bills are identical with respect to paying for things, but they each have different histories and precise atomic compositions. Similarly, when thinking about personal identity over time, we need to distinguish between the senses in which you and a copy of you are the same, and the senses in which they are different, and decide which senses matter to us and which do not.
After the moment of your copy’s creation, you both clearly have different perspectives on the world. Does this one difference make you meaningfully different people? Imagine instead that ordinary ‘you’ goes to sleep, and while you are unconscious, someone transports you to Japan. When you wake up in Tokyo, are you a different person? No. You’re the same person, just with a different perspective.
Many object to the idea that you and your copy only differ with respect to perspective, because of course, your copy is made of entirely different stuff! An obvious problem with this response is that your ordinary self does not consist of the same stuff over time, given the components of most human cells get replaced every seven or so years. Yet we tend to think we remain the same person over that period of time.
Defining something in terms of its precise physical components makes sense in certain contexts. For instance, the question of whether a painting was the one that Claude Monet painted is very important for art dealers. A forgery can be ‘perfect’, in the sense that it replicates the exact appearance, smell and feel of an artwork, but it is still worthless in the eyes of collectors. They care about the history of the actual atoms that compose the forgery. Did that splash of red paint come off the very brush of the master?
Some of us, however, don’t care much about the origins of art, and would enjoy a perfect forgery just as much as the original. For us, it is about the experience that the art affords us, not the higher social status of owning something rare.
This, I suggest, is how we should think about our minds.
A mind is best conceived as information stored and processed in a physical nervous system. This means that just like a novel or computer program, we shouldn’t try to define a ‘self’ in terms of its physical components. Rather, we should define it in informational terms. Bodies and brains are certainly necessary for our physical instantiation as selves, but the specific ones we employ today are likely not the only possible ones.
In his monumental 1984 book Reasons and Persons, Parfit argued for what he called a ‘reductionist’ viewpoint on personal identity, which entailed that all that matters for your survival is psychological connectedness and continuity, not any physical connection with your prior body. That is, so long as you share the memories, dispositions and other psychological properties with your past self, you should be considered to be the same old you.
Parfit’s approach addresses the problem of one’s personality changing gradually over the years. As a 60-year-old, you might not be strongly psychologically connected to yourself at age 10 – indeed you might remember hardly anything of that version of you. But so long as you have a strong psychological connection to who you were at age 50, and that 50-year-old version of you has a similarly strong connection to who they were at age 40, this chain of personal identity continues all the way back to your youngest conscious self. Most importantly, Parfit thought this way of thinking about a person’s identity over time made far more sense than the singular ‘Cartesian ego’ whose identity must be ‘all or nothing’.
Parfit’s dismissal of ‘physical connectedness’ is still controversial among philosophers today. Philosopher Jeff McMahan, a long-time colleague and friend of Parfit’s, still suspects that some kind of physical connectedness is necessary for personal survival. In a recent podcast, McMahan expressed his intuition that a mere copy of himself continuing to exist elsewhere is no consolation for the death of his original body and brain. He would be comfortable however with his atoms being disassembled and then reassembled perfectly back as they were. In other words, he thinks that our personal identity depends upon the very same atoms coming along for the ride.
This strikes me as an untenable view given atoms of the same type are truly identical, so it shouldn’t make any difference whatsoever whether some or all of the atoms in our bodies are replaced with equivalent ones. Zooming out from the atomic scale, it seems likely that even different chemical properties could give rise to the same higher-level psychological properties, due to a phenomenon called ‘multiple-realizability’. There seems to be no reason in principle why biological neurons are the only possible substrate for thoughts and feelings.
The philosopher Susan Schneider also doesn’t buy the kind of view Parfit is selling. In her 2019 book Artificial You, Schneider argues against the ‘patternist’ conception of a person, as promoted by Nick Bostrom and Ray Kurzweil, which has many of the same implications as Parfit’s view. Schneider rules out this position by imposing a requirement of ‘spatiotemporal continuity’ upon a person for them to be considered identical to their past self. It offends Schneider’s intuition that a person’s mind could jump instantaneously across time and space between bodies like an internet signal. In her view, to remain the same mind, you must enjoy a smooth, physically continuous trajectory through space-time.
The problem with Schneider’s move is that whether our journey through life is physically continuous depends on the scale at which you examine it. Yes, from our intuitive human range of experience, it looks as though we move through the world in one fluid motion. But if you were to zoom right down to the quantum level of quarks in our neurons and muscle fibres, you would see entities discontinuously jumping around – sometimes acting like particles and sometimes acting like waves. This makes Schneider’s requirement of spatiotemporal continuity seem totally arbitrary – a product of our parochial scale of observation.
If these counterarguments don’t stack up, why are people so reluctant to give up on a strictly physical conception of personal identity?
Evolution has not equipped our instincts to find multiple instances of ourselves plausible. By analogy, consider the modern practice of being injected with a vaccine with a sharp syringe. For most people, no amount of rational deliberation on the scientific validity of vaccination can overcome the visceral emotional response of being jabbed in the arm. This feeling of personal violation is a product of millions of years of evolving in environments where such events were uniformly bad for us.
Similarly, the highly intuitive perspective that there can only ever be one physical continuation of me will be very hard, if impossible, to shake. But we should try our best to consider the issue from a more objective viewpoint. Many of us may end up deciding to go ahead with the brain-copy procedure despite our strong intuitions to the contrary, just as most of us still grit our teeth and bear a painful injection to defend ourselves from nasty infections.
One ‘intuition pump’ that can help assuage such doubts is the ‘gradual replacement’ hypothetical transition scenario. While you remain fully conscious throughout, I start replacing each of your neurons one by one with artificial ones. We converse happily for the duration, and you never feel a thing. Are you still the same person after this long operation? If you think you’re not, the challenge is to describe the exact moment you somehow switched from being the real you to being a mere forgery. This seems impossible to do. And it’s unclear why replacing all your neurons at once should make any difference to the outcome.
Those of us who do choose to take this leap will certainly be generating some very strange moral and legal conundrums.
A scientific or causal analysis of the situation is very comfortable with gradations of ‘you-ness’. There is no essential, singular, authoritative version of you – the Cartesian ego. This means there is no reason there can’t exist many, many copies of you with slight variations, which we simply label you1, you2, you3…etc. In the context of morality and law however, we can’t deal in infinitesimal shades of grey. We think in terms of individuals, each with certain rights and responsibilities.
For instance, let’s say we can scan your brain non-invasively and create a copy of you in an artificial body. You might naively think that you can both go off and live your lives as distinct people. But then you try withdrawing money from your bank account, only to discover that half the funds have been drained. That damn copy! He knows your passwords, address, date of birth, and your mother’s maiden name. You immediately file a lawsuit to get the money back, claiming you are the rightful owner of the bank account.
As the original, biological version, you may argue that only you are the real you. After all, you inhabit the body and brain that signed the contract with the bank and that acquired the money in the first place. Your copy’s lawyer counters, in true Parfitian style, that these details are not essential to personal identity. What matters is the information of which your mind consists. That information was perfectly copied in the procedure, so your copy shares your exact personal identity.
The judge sides with your copy, and you are required to relinquish half your wealth to them. You curse yourself for ever making that stupid decision to copy yourself. You thought you could get more stuff done by having someone else on your team with the same values and goals. Instead, you just created an annoyingly autonomous version of yourself with increasingly divergent interests. But there’s no time for all this self-berating, because you’re due in the next court session over who is entitled to be married to your wife and to have custody over your kids.
Things get very messy when there are multiple versions of you running around, because our most fundamental concepts in morality and law have hitherto been premised upon everyone having a singular physical identity. Should you be held responsible for crimes your copy commits? We’ll explore this in a future essay, but the short answer is yes. Merely having no memory of committing the crime will be no excuse when you and the perpetrator share the same core identity.
Artificial brains promise extraordinary benefits but will also challenge our deepest intuitions about who and what we are. Getting our heads around the implications now will help us steer wisely through the next great technological age: the fusing of humans and machines.