Perhaps computers in the next twenty-thirty years will have the processing power of human brains, as measured by the amount of external data they are able to process. But this (of course) does not even make them close to the human brain - it is the software that runs on a machine, not the hardware. In fact, the hardware may be totally irrelevant except in cases where interaction with the outer world is needed - a ai program running on a really old computer is still ai, isn't it. This page has some interesting arguments about the subject of speed and intelligence; the author argues that intelligence may be dependent on the interior baud rate of the machine, and the baud rate with which it reacts to the outside world.
Yes, computers will design other computers, as it has already been proven, but it will be much more than seventy years before they are able to genetically generate something equal to the human brain. Consider this, if a genetic evolution were to be simulated for the human brain: approximately 1 billion humans participated in the evolution of humanity as we know it (this is a very conservative estimate of 400 generations with an average of 2.5 million people participating). And humans have generally 15 billion neurons. By another conservative estimate, they think along the lines of 1/100 of a second; they don't use their entire brains during this time, but simulation software couldn't figure this out (and neither can we). Many neurons (or so I am told) connect in no fewer than 10000 places; thus they must be represented by at least 14 bits. If this is all multiplied out, we must sample 15 million trillion neurons every 1/100 of a second (each connecting at least 100 others, since not all connect to very many others), or 0.15 trillion trillion member accesses per second to conquer the problem in the aforementioned 70 years, if we are able to split up the hierarchy of evolution to parallelze all 1 billion lives into life spans of 70 years, and considering that any action of a human may influence its life, death, and ability to create offspring. To hold all this, we might need 210 billion trillion bits, or 26.25 billion trillion bytes. And that is merely the very last piece of evolution, a second compared to the billion years that got us here.
That shows some very fuzzy reckoning of how hard such a simulation problem would be to do with todays technology. Even with quantum states, the numbers are on the order of a mole (!)
Somewhat more irrelevant: I feel that even tough the problem is daunting, we need to try it; I think that humans are de-evolving as a result of modern medicine; genetic mutations are piling upon out human genotype; so what if machines replace us? We can make them better than us.
Slashdot Mirror
Yes, computers will design other computers, as it has already been proven, but it will be much more than seventy years before they are able to genetically generate something equal to the human brain. Consider this, if a genetic evolution were to be simulated for the human brain: approximately 1 billion humans participated in the evolution of humanity as we know it (this is a very conservative estimate of 400 generations with an average of 2.5 million people participating). And humans have generally 15 billion neurons. By another conservative estimate, they think along the lines of 1/100 of a second; they don't use their entire brains during this time, but simulation software couldn't figure this out (and neither can we). Many neurons (or so I am told) connect in no fewer than 10000 places; thus they must be represented by at least 14 bits. If this is all multiplied out, we must sample 15 million trillion neurons every 1/100 of a second (each connecting at least 100 others, since not all connect to very many others), or 0.15 trillion trillion member accesses per second to conquer the problem in the aforementioned 70 years, if we are able to split up the hierarchy of evolution to parallelze all 1 billion lives into life spans of 70 years, and considering that any action of a human may influence its life, death, and ability to create offspring. To hold all this, we might need 210 billion trillion bits, or 26.25 billion trillion bytes. And that is merely the very last piece of evolution, a second compared to the billion years that got us here.
That shows some very fuzzy reckoning of how hard such a simulation problem would be to do with todays technology. Even with quantum states, the numbers are on the order of a mole (!)
Somewhat more irrelevant: I feel that even tough the problem is daunting, we need to try it; I think that humans are de-evolving as a result of modern medicine; genetic mutations are piling upon out human genotype; so what if machines replace us? We can make them better than us.