The Limits of Being Human
Created | Updated Jan 28, 2002
(Work in Progress)
There are two numbers I've been thinking about recently. 100 billion, and 3 billion. At first glance, these seem like fairly big numbers, but in context, there are many things bigger than them. The hard drive of this computer has about 10 billion bytes of storage. A DVD stores about seven gigabytes (in actual fact, it stores about 12 bits for each 'byte', using the extra redundancy for error correction, making it about 100 gigabits of storage) and you can hold it in your hand. There are a billion grains of sand in a bucket. These are countable things with physical existence, not merely the unreal billions of Microsoft's bank account.
So, what are the two numbers? The first is the number of neurons in a typical human brain. The second is the number of base-pairs in the human genome.
Let's take the human genome first. 3 billion base pairs. Each pair can be of four different types, so it would take 6 gigabytes to store the raw data of the entire human genome. Codon sequences are fairly repetitive, so compression is likely to squeeze that down to fit on a CD. Imagine that. The entire human blueprint, lost somewhere in your stack of Pink Floyd albums.
Actually, there's anywhere up to double that amount of genetic information needed to make a human, since some of the organelles have their own 'chromosomes', due to the fact they were once independant organisms which took up symbiotic residence in our cells. Call it double. Two CD's then. Or one DVD, with enough space left over to fit a copy of Run Lola Run.
The Human Genome Project has spent ten years sequencing that data, and they've only finished two chromosomes. but that's a deceptive number. Part of the project was developing new and faster ways to sequence as they went. So, even though they're only 10% of the way through, they expect the other 90% to only take a few more years. (Other commercial projects are actually ahead, and will probably get there first) By the end, they'll have the technology to go back and do it all again in a few months. 'Genome' actually means all the possible chromosomes of a species, not just a single individual. (Though most of us share 99% of the same code) Within a few years, it will be feasible to sequence individuals, and their organelles. A few years after that, it might be a fairly routine medial procedure, like CAT scans are today. Visit the doctor, have a 'blood test', and the next week pick up a disc containing all the information needed to clone yourself. Who would you bequeath that to in your will?
It's true that the chromosomes are only part of the story. They're like the initial arrangement of a chessboard... before the game is played out with all the subtle complexities possible. Actually, the genome is kind of the opposite, starting from random ensemble of pieces with set rules and consistently reaching the same endgame: a working human.
Supercomputers today are put to work to determine the results of running tiny pieces of genetic code. What proteins get made? What are their properties? To simulate an entire cell, just one, is currently utterly beyond us. But it will not always be so. Where we are was inconceivable just a few decades ago.
We are firmly in the grip of Moore's Law, which says that compters get twice as powerful every year and a half. This exponential rate of growth shows little sign of stopping. If anything, as the feedback becomes more intense, it may accelerate. Inconceivably powerful computers will be here within our lifetime. We have problems waiting for them.
Speaking of computers, it's time for that second number. 100 billion neurons in the brain. If we are the ghost in the machine, then they are the hardware. No individual neuron is important; conciousness seems to be holistic.
Here's a thought experiment. Suppose you can get a computer to exactly simulate a neuron, in real time. This isn't easy, especially since our understanding of them is incomplete. But, they're small thing made from atoms, subject to the laws of Quantum Electrodynamics (which is the single most accurate theory ever create) and built from the genetic code. Alright, it would require a big computer. Maybe a really big one. But it's clearly possible.
Now, here's the clever part. You connect up the computer in parallel with a neuron in someone's head. You get the simulated neuron to learn the total behaviour of the real one. To mimic it in every respect. Then you kill the original neuron, and let the simulation take it's place.
Does the person we've subjected to this bizzare experiment notice? No. It's just one neuron.
So we do a second. And a third. Then ten. A thousand. A million!
Do they feel strange? Can they tell the difference, now that one hundred thousandth of their mind is embodied by a simulation in a computer, and no longer on the 'wetware' of their neurons? How far can we go? Can we keep removing neurons and replacing them with software, until there's nothing left of the original brain? Will the mind survive?
Questions like this give a lot of people the screaming heebie-jeebies, because the implications either way are shocking. My view is that there's nothing preventing us from doing this. In fact, we probably don't have to simulate the neurons to 100% accuracy as long was we get the key things right. What's important about neurons is not what they do, but how they relate, dropping the computational requirements down to a level which may be achieveable in our lifetimes. In fact, given optimistic estimates, the equivalent of an modern affordable PC can do it in 30 years, which rather changes the meaning of 'personal computer'.
This is the blessing and the curse of advancing computer technology. It gives us new tools to understand ourselves, but promise to overtake us in complexity within half a century. The entire of who were are may become just another mote of software in some future processor. A curious concept.
Because, once there, we can finally subject ourselves to the same improvements that have been driving our technology. Can't imagine that? Get a RAM upgrade, and double your brain size. Run twice as fast, so the 'real world' slows down to half speed, giving you all the time in the world to consider it.
You'll need it too, to consider stumpers like "what am I?".
There is no question that we'll go there. This 'machine transcendance' offers immortality, a prize sought after for as long as there have been legends. Someone will do it. How is still unkown, although most bets are on nanotechnology if you want it via neuron replacement surgery.
But, will the mind survive? I think so. Even if 'conciousness' is a quantum phenomenon, we'll soon have quantum machines which run software on those same principles. I think the mind is like a note plucked on a string. If you're clever enough to replace the string without stopping the hum, then isn't it the same note?
Thirty years. We'll all find out then.
There are two numbers I've been thinking about recently. 100 billion, and 3 billion. At first glance, these seem like fairly big numbers, but in context, there are many things bigger than them. The hard drive of this computer has about 10 billion bytes of storage. A DVD stores about seven gigabytes (in actual fact, it stores about 12 bits for each 'byte', using the extra redundancy for error correction, making it about 100 gigabits of storage) and you can hold it in your hand. There are a billion grains of sand in a bucket. These are countable things with physical existence, not merely the unreal billions of Microsoft's bank account.
So, what are the two numbers? The first is the number of neurons in a typical human brain. The second is the number of base-pairs in the human genome.
Let's take the human genome first. 3 billion base pairs. Each pair can be of four different types, so it would take 6 gigabytes to store the raw data of the entire human genome. Codon sequences are fairly repetitive, so compression is likely to squeeze that down to fit on a CD. Imagine that. The entire human blueprint, lost somewhere in your stack of Pink Floyd albums.
Actually, there's anywhere up to double that amount of genetic information needed to make a human, since some of the organelles have their own 'chromosomes', due to the fact they were once independant organisms which took up symbiotic residence in our cells. Call it double. Two CD's then. Or one DVD, with enough space left over to fit a copy of Run Lola Run.
The Human Genome Project has spent ten years sequencing that data, and they've only finished two chromosomes. but that's a deceptive number. Part of the project was developing new and faster ways to sequence as they went. So, even though they're only 10% of the way through, they expect the other 90% to only take a few more years. (Other commercial projects are actually ahead, and will probably get there first) By the end, they'll have the technology to go back and do it all again in a few months. 'Genome' actually means all the possible chromosomes of a species, not just a single individual. (Though most of us share 99% of the same code) Within a few years, it will be feasible to sequence individuals, and their organelles. A few years after that, it might be a fairly routine medial procedure, like CAT scans are today. Visit the doctor, have a 'blood test', and the next week pick up a disc containing all the information needed to clone yourself. Who would you bequeath that to in your will?
It's true that the chromosomes are only part of the story. They're like the initial arrangement of a chessboard... before the game is played out with all the subtle complexities possible. Actually, the genome is kind of the opposite, starting from random ensemble of pieces with set rules and consistently reaching the same endgame: a working human.
Supercomputers today are put to work to determine the results of running tiny pieces of genetic code. What proteins get made? What are their properties? To simulate an entire cell, just one, is currently utterly beyond us. But it will not always be so. Where we are was inconceivable just a few decades ago.
We are firmly in the grip of Moore's Law, which says that compters get twice as powerful every year and a half. This exponential rate of growth shows little sign of stopping. If anything, as the feedback becomes more intense, it may accelerate. Inconceivably powerful computers will be here within our lifetime. We have problems waiting for them.
Speaking of computers, it's time for that second number. 100 billion neurons in the brain. If we are the ghost in the machine, then they are the hardware. No individual neuron is important; conciousness seems to be holistic.
Here's a thought experiment. Suppose you can get a computer to exactly simulate a neuron, in real time. This isn't easy, especially since our understanding of them is incomplete. But, they're small thing made from atoms, subject to the laws of Quantum Electrodynamics (which is the single most accurate theory ever create) and built from the genetic code. Alright, it would require a big computer. Maybe a really big one. But it's clearly possible.
Now, here's the clever part. You connect up the computer in parallel with a neuron in someone's head. You get the simulated neuron to learn the total behaviour of the real one. To mimic it in every respect. Then you kill the original neuron, and let the simulation take it's place.
Does the person we've subjected to this bizzare experiment notice? No. It's just one neuron.
So we do a second. And a third. Then ten. A thousand. A million!
Do they feel strange? Can they tell the difference, now that one hundred thousandth of their mind is embodied by a simulation in a computer, and no longer on the 'wetware' of their neurons? How far can we go? Can we keep removing neurons and replacing them with software, until there's nothing left of the original brain? Will the mind survive?
Questions like this give a lot of people the screaming heebie-jeebies, because the implications either way are shocking. My view is that there's nothing preventing us from doing this. In fact, we probably don't have to simulate the neurons to 100% accuracy as long was we get the key things right. What's important about neurons is not what they do, but how they relate, dropping the computational requirements down to a level which may be achieveable in our lifetimes. In fact, given optimistic estimates, the equivalent of an modern affordable PC can do it in 30 years, which rather changes the meaning of 'personal computer'.
This is the blessing and the curse of advancing computer technology. It gives us new tools to understand ourselves, but promise to overtake us in complexity within half a century. The entire of who were are may become just another mote of software in some future processor. A curious concept.
Because, once there, we can finally subject ourselves to the same improvements that have been driving our technology. Can't imagine that? Get a RAM upgrade, and double your brain size. Run twice as fast, so the 'real world' slows down to half speed, giving you all the time in the world to consider it.
You'll need it too, to consider stumpers like "what am I?".
There is no question that we'll go there. This 'machine transcendance' offers immortality, a prize sought after for as long as there have been legends. Someone will do it. How is still unkown, although most bets are on nanotechnology if you want it via neuron replacement surgery.
But, will the mind survive? I think so. Even if 'conciousness' is a quantum phenomenon, we'll soon have quantum machines which run software on those same principles. I think the mind is like a note plucked on a string. If you're clever enough to replace the string without stopping the hum, then isn't it the same note?
Thirty years. We'll all find out then.
