While medicine will advance in the next half century, we are not on a crash-course for achieving immortality by curing all disease. Bodies simply wear down with use. We are on a crash-course, however, with technologies that let us store unthinkable amounts of data and run gargantuan simulations. Therefore, well before we understand how brains work, we will find ourselves able to digitally copy the brain's structure and able to download the conscious mind into a computer.
If the computational hypothesis of brain function is correct, it suggests that an exact replica of your brain will hold your memories, will act and think and feel the way you do, and will experience your consciousness — irrespective of whether it's built out of biological cells, Tinkertoys, or zeros and ones. The important part about brains, the theory goes, is not the structure, it is about the algorithms that ride on top of the structure. So if the scaffolding that supports the algorithms is replicated — even in a different medium — then the resultant mind should be identical. If this proves correct, it is almost certain we will soon have technologies that allow us to copy and download our brains and live forever in silica. We will not have to die anymore. We will instead live in virtual worlds like the Matrix. I assume there will be markets for purchasing different kinds of afterlives, and sharing them with different people — this is future of social networking. And once you are downloaded, you may even be able to watch the death of your outside, real-world body, in the manner that we would view an interesting movie.
Of course, this hypothesized future embeds many assumptions, the speciousness of any one of which could spill the house of cards. The main problem is that we don't know exactly which variables are critical to capture in our hypothetical brain scan. Presumably the important data will include the detailed connectivity of the hundreds of billions of neurons. But knowing the point-to-point circuit diagram of the brain may not be sufficient to specify its function. The exact three-dimensional arrangement of the neurons and glia is likely to matter as well (for example, because of three-dimensional diffusion of extracellular signals). We may further need to probe and record the strength of each of the trillions of synaptic connections. In a still more challenging scenario, the states of individual proteins (phosphorylation states, exact spatial distribution, articulation with neighboring proteins, and so on) will need to be scanned and stored. It should also be noted that a simulation of the central nervous system by itself may not be sufficient for a good simulation of experience: other aspects of the body may require inclusion, such as the endocrine system, which sends and receives signals from the brain. These considerations potentially lead to billions of trillions of variables that need to be stored and emulated.
The other major technical hurdle is that the simulated brain must be able to modify itself. We need not only the pieces and parts, we also the physics of their ongoing interactions — for example, the activity of transcription factors that travel to the nucleus and cause gene expression, the dynamic changes in location and strength of the synapses, and so on. Unless your simulated experiences change the structure of your simulated brain, you will be unable to form new memories and will have no sense of the passage of time. Under those circumstances, is there any point in immortality?
The good news is that computing power is blossoming sufficiently quickly that we are likely to make it within a half century. And note that a simulation does not need to be run in real time in order for the simulated brain to believe it is operating in real time. There's no doubt that whole brain emulation is an exceptionally challenging problem. As of this moment, we have no neuroscience technologies geared toward ultra-high-resolution scanning of the sort required — and even if we did, it would take several of the world's most powerful computers to represent a few cubic millimeters of brain tissue in real time. It's a large problem. But assuming we haven't missed anything important in our theoretical frameworks, then we have the problem cornered and I expect to see the downloading of consciousness come to fruition in my lifetime.
(I originally published this essay in John Brockman's This Will Change Everything, a collection of answers to the Edge.org annual question)
See my comments on the Amazon series Upload for more on the possibility of transferring consciousness to a computer.
