11 February 2007

Some thoughts on uploading

The topic of uploading human minds into computers has confronted me in a number of contexts recently. Here are a few related ideas I have been thinking about. Most people familiar with transhumanism will already have their own views on these matters, but this posting may be of interest to those who are new to the subject.

What makes me "me"?

What is the self? When we speak of preserving the existence of a human individual, what exactly is it that needs to be preserved?

To begin with, I do not believe that there is any supernatural component to it – a "soul" or "spirit". Phenomena such as free will and consciousness, which are the very basis of our subjective experience of being ourselves, seem so different from the observed behavior of "ordinary" matter and energy which the laws of physics can explain, that many people assume they must be of an entirely different nature and origin – some sort of divine spark bestowed upon us by a deity. But this is just the latest repetition of an error we have made again and again through history – that if we cannot yet find a natural explanation for a given phenomenon, there cannot be one, and the phenomenon must therefore be supernatural. At various times sunshine, lightning, rain, and many other phenomena were attributed to the actions of deities by humans whose science was not yet advanced enough to figure out what really caused them. When it did become advanced enough, it turned out that these were purely natural phenomena and the supernatural explanations were discarded. I have no doubt that the same will happen with the mysterious qualities of our minds.

Note also that other mammals possess all or most (depending on the species) of the mental traits that we do, to a lesser degree of sophistication – and that the more an animal’s brain resembles the human brain in complexity and anatomical structure, the more similar its intelligence and other mental abilities are to our own, to the point that the great apes (whose brains are the most similar of all to ours) are essentially accepted as "people", even if primitive ones, by most humans who have worked with them for long periods of time. This bolsters the view that our minds, like our brains, are ordinary natural phenomena produced by biological evolution.

(I also reject the idea that, because free will cannot be explained by the presently-known laws of physics, it must be an illusion. This would be equivalent to early humans deciding that lightning didn't really exist because they couldn’t think of a natural explanation for it. In any case, even if free will were found to be an illusion in some sense, it is so subjectively real that the "fact" of its being an illusion would be of no practical relevance.)

So if selfhood is a natural phenomenon rooted in the physical brain, what is it? What needs to be preserved in order for us to say "this person still exists"?

Is it the continuity of the physical matrix? Fortunately for the subject under discussion, the answer seems to be no. You may think you have had the same brain all your life, but the organic processes of your body are constantly replacing its substance molecule by molecule, so that of all the molecules which made up your brain twenty years ago, there is probably hardly a single one that is still there today. The physical matter changes, but the arrangement of it – the pattern – remains. It’s that pattern that is important.

When you look at all the mental traits which make an individual that individual, as distinct from other individuals – his feelings, opinions, specialized skills, and so on – it’s actually remarkable how little the continuity of these things seems essential to the continuity of identity. Your opinions can change, your aspirations can change, you can learn a new skill or language, without any weakening at all of your sense that you are still the same person.

It seems to me that the key to continuity of selfhood is memory. I’ve changed in all sorts of ways over the last twenty years, but I still remember what I thought, did, and was, through all that time, and it’s that continuity of memory that makes me feel I am the same person now that I was then.

(This, by the way, is why I wouldn’t be impressed by the concept of reincarnation even if I believed in it. Believers in reincarnation generally hold that a reincarnated person has no, or almost no, memory of his previous life. If I die, and later another entity which is supposed to be a continuation of me is born, but that entity has no memory of being me, then it isn’t me.)

Fortunately, we don’t need to choose. All mental phenomena – memory, consciousness, will, beliefs, emotions, everything – must, since they can exist at all, be natural phenomena operating within the brain. As natural phenomena, they can eventually be completely understood; and having been completely understood, they can eventually be preserved independently of the brain.

A digression on brain complexity

The human brain is a massively-parallel organic supercomputer which processes information at the synapses (connections between neurons) and constantly reprograms itself by changing the arrangement of the synapses, forming new connections between neurons and dropping old ones (learning a new skill, for example, involves the formation of many new synaptic connections in the relevant part of the brain – this process has been photographed). For a neuron to go through one complete cycle of firing and returning to its ground state takes about 1/200 of a second, so a neuron can fire no more than 200 times per second. The brain’s true complexity lies in the enormous number of synaptic connections – there are about 100 trillion of them. Thus we can say that the maximum computational capacity of the brain is about 20,000 trillion operations per second. The staggeringly complex patterns of neuron-firing which are continuously running on this "computer" are the "programs" which we subjectively experience as memory, consciousness, will, and everything else that makes up the self.

To upload a person, there will be two technological requirements: (1) the ability to scan an individual brain at high enough resolution to capture every detail of all this complexity – not just the arrangement of a hundred trillion synaptic connections, but details of the structure of individual neurons which influence exactly how each one fires; and (2) a computer with the capacity to create a simulation of the scanned brain to that same degree of detail, so that the same "programs" could be run on it just as they run on the brain. Both of these things are obviously beyond the capabilities of the equipment we have now – but not by such a large margin as you might suppose. That maximum brain capacity of 20,000 trillion operations per second, for example, is only about 70 times greater than the capacity (280 trillion operations per second) of the most powerful man-made computer in existence today. Ray Kurzweil estimates that to fully simulate an individual brain at the level of detail needed to actually upload a person, a computer would need a computational capacity about 100 times that of the brain itself. But since the capabilities of computer technology grow at an exponential rate (and a very fast one), the necessary capabilities in both scanning and computation should be achieved much more quickly than the layman today would probably expect. Kurzweil estimates that human uploading will become feasible in the early 2030s.

(This is, of course, a drastically simplified summary of the issues covered. For a full discussion, read Kurzweil’s book.)

Uploaded life

Uploading would bring us as close to true immortality as it is probably possible to get. Even if the forecasts of thinkers like Aubrey de Grey are borne out and aging and disease are eradicated in the developed world by the 2020s, people will still be vulnerable to death from accidents, terrorism, and so forth. An uploaded person could be protected from death by all the same techniques by which any extremely-valuable computer program and stockpile of data can be protected – including the keeping of regularly-updated backup files from which the person could be restored in the event of a catastrophic system failure.

An uploaded person would emphatically not experience his existence as a mere disembodied consciousness. He could interact with people and the physical world through a sensory and motor interface with a physical body specially grown for the purpose, or (more likely) function in virtual-reality environments sensorily indistinguishable from physical reality. The technology required for either of these options would be much less sophisticated than that required for the uploading process itself. Indeed, I expect that by the time uploading becomes feasible, a great deal of ordinary human interaction will routinely take place in virtual reality anyway.

In certain cases, uploading may even enable us to literally restore the dead to life. Modern cryonics is already using vitrification to preserve the brains of dead people, in the hope that future technology will be able to repair the damage done to the brain by death and by the vitrification process itself, and restore the brain to normal functioning, thus bringing the dead person back to life (presumably in a specially-grown body). But even if this cannot be done, if the process preserves the entire pattern of synaptic connections in the brain, it should in principle be possible to scan the vitrified brain and upload it.

The uploaded mind will be freed from the limitations imposed by the organic brain. Its computational capacity will be free to increase as computers continue to become more sophisticated. Eventually humans will be able to increase their intelligence to whatever level a particular person needs to solve a given problem, design and run a given virtual-reality environment, or whatever else he wants to do – even to levels trillions of times greater than our present organic intelligence.

I expect that by the middle of this century, uploaded people working, playing, and interacting in virtual-reality environments of their own choosing will be the norm, while having a limited physical body and dealing with the messy and intractable "real world" will be an option which fewer and fewer people bother with. I also expect that I and most of the individuals reading this will still be around to experience this era.

As other technological innovations have done in the past, uploading will confront us with new ethical and philosophical questions. An uploaded person whose mind had grown sophisticated enough could create sub-programs with whatever characteristics he chose, as a novelist creates characters in his mind (or as a god creates new life, perhaps?). If these sub-programs possessed self-awareness and volition in their own right, their creation and use would give rise to a thicket of ethical issues. Another problem (or opportunity) is the existence of multiple versions of the same person. If my brain is scanned and its complete synaptic pattern is uploaded into several computers instead of just one, then there will be several people (identical at first, though steadily diverging thereafter due to differences in experience), each one of them equally "me" by any standard one cares to apply.

As challenging as such issues will be, I personally am not frightened or discouraged by them. Every technological advance in history has created some unexpected problems, but on balance the benefits have almost always been well worth the drawbacks. And whatever conundrums uploading presents us with, we – unlike any previous generation – will have unlimited time and intelligence with which to address them.

Labels: ,


Post a Comment

<< Home