Well, assuming that we are Turing machines (which seems probable) then we could say that a mathematical formula is not our way of expressing reality, but actually our way of perceiving it (since all possible thoughts are then just mathematical relations). Since we can never see anything but the mathematical formula, trying to differentiate between reality and the formula becomes fairly meaningless.
Bah, I tend not to trust people who's idea of excitment (here, moral building or something) is indistiguisable from being really really angry. And who can't dance. And who scream thing's like 'Give it up for me." and "I Love this company." I mean really. seriously. But they're funny to watch.
Ballmer Dancing
and
His hypnotic voice.
I am not trying to be contentious, but, I would like to say that I personally find the idea that it would be easier to 'simulate' the growth and development of an entire organism rather hard to envision. But, I may vastly be underestimating the power of computers twenty years down the road, so I will do a back of the envelope calculation. Now, Human DNA codes for around 30,000 to100,000 proteins. It Takes 19,000,000 Gflop too compute roughly 10ns of a 36 Amino acid protein. (1) Lets say that a protein takes around 500ns to fold, (fastest 10ns, some take several milliseconds) Assuming linearity, then an average protein might take 7,900,000,000 Gflop. Now, there will undoubtedly be some short cuts you can take, and you will not have to calculate every inch of those 30,000-100,000 proteins, so lets assume you only need 10%, then a minimum of 2.37 * 10^13 Gflop or (2.36*10^13*2^30 flop) are needed. So now, by moore's law we would expect computing power in 30 years to be what, 10*9*2^(30*12/18)= ~10^15 flops; it takes roughly 7.11*10^13*2^30/10^15 seconds, or roughly 1 year on the low or 3 on the high end.
Crap.
It appears I vastly underestimated the advances in computing power. If moors law holds out and we develop efficient enough algorithms, then we might have the protein products of DNA computed in 30 years. Well that's pie in my eye. Keep in mind you still have to run interactions between these proteins in cells, which should be an order of magnitude more complex, and do it in trillions of cells over at least 9 months of development, but my minimalist argument has failed. And I am sad.
Just to retain some dignity here, I would like to point out that DNA cannot be treated as a blue print, which is like a scaled down representation of an object, but should be treated like a recipe, (or even more accurately, a chemical synthesis scheme.) which wile it still gives you a very complex finished product, the step beat eggs for ten minutes, has no simple correlation to a particular area of the finished product (if you skip this step you don't get half a cake you get no cake at all) as Dawkins so cogently pointed out. And, I personally still think I am right, but I suppose I will stop arguing at this point. *sigh*
I suppose I could fudge the numbers, I doubt any one will bother checking, but what would the point be? To misquote Huxley; "a beautiful theory, killed by an ugly, nasty little fact."
(1) http://www.psc.edu/science/Kollman98/kollman98.htm l; "T3E, four times faster, the next 800 nanoseconds took about another two months" The T3E operates at 307 Gflops.
T3E, four times faster, the next 800 nanoseconds took about another two months.
Quick Update, from Duane Chapman's Enviromental Economics; 2000, p132. Source / Cost per kWh Solar Photo-Cell / $.50 Solar Thermal / $.33* Wood / $.12 Wind / $.09** Nuclear / $.09 Coal / $.06 Hydro / $.06*** * author notes that cost estimates are as low as $.08 from some sources, and that at any rate, should decline around $.07 annually. ** I think this has gone down since 2000 *** By this I believe he means dams, not newer tidal systems. Finally, these rates do not include government incentives for alternative fuel.
Quick Update, from Duane Chapman's Enviromental Economics; 2000, p132.
Source:Cost per kWg
Solar Photo-Cell$.50
Solar Thermal$.33*
Wood$.12
Wind$.09**
Nuclear$.09
Coal$.06
Hydro$.06***
* author notes that cost estimates are as low as $.08 from some sources, and that at any rate, should decline around $.07 annually.
** I think this has gone down since 2000
*** By this I believe he means dams, not newer tidal systems.
Finally, these rates do not include government incentives for alternative fuel.
1) In my original post I was not claiming that complex systems had to be evolved, which is non-sense (cars, and supercomputers) but rather that complex systems are either a) evolved or b) created. So this idea of AI just 'waking up' proposed in the FA was non-sense. Since it would not be created, it would have to be evolved. And there is good reason to believe from evolutionary theory that this is highly unlikely, being as the computer virus it would have to, by necessity, evolve from, will likely chose the near universal evolutionary answer of simplicity over the extremely uncommon answer of complexity. (Try to imagine a 8 GB computer virus infecting your computer. . . )
2) Realizing that I am not in the field of AI, and in all likely hood neither are you, so both of us don't know what we are talking about; Your faith in computers aiding us in simulating the mind is entirely misplaced. I will acknowledge that some one may make a detailed map of the brain through any of a dozen different methods, and this may in fact be the way that we arrive at the first AI, and I claimed nothing to the contrary in my post. However, we are talking about speculating on the time it will take us, and not the method that it will be arrived at. I am somewhat skeptical that mapping the brain will be as trivial as you assert. (Though I am not a neurologist and will bow down to superior wisdom if contradicted by a neurologist.) Realizing that not only the place the connections are made between neurons must be mapped, but also the receptors involved (and we still do not know all the neuroreceptors) and we must figure out weather the each receptor on each neuron is acting in an inhibitory or stimulating manor, and figure out by what internal states of the neuron its activities can be modulated, and finally to top it all off we have to understand the function of glial cells, which have recently be indicated to play a role in neurological activity, and make up some 90% the brain . . . I would say I am not all that 'blind' in doubting the 20 or 30 year figure. The brain is more complex than it thinks it is.
In conclusion: I'm not saying that this or that will be the way we finally create an AI, or that this or that process is going to be so hard that it will take 100 years to figure out. I'm not saying it is impossible that it will be here in 20 years, just saying I'm not convinced. And that is the point. No one has tried (with much success) to map the brain into a computer and simulate it. The amount of processing power it would take, and the techniques it would require to be efficient are unknown. Unless you have some evidence, such as some one having simulated an ant brain, (which may have been done, but a quick search of JSTOR didn't show anything.) then you don't know how long it will take. Because it hasn't been done. And that is what I am saying.
Ps: I consider the LEST feasible way to figure out how the brain works to be looking at DNA, because it ought to be immensely easier just to study the finished product.
I have an alternative theory. Governments are generally unwilling to drop tens of billions of dollars into technologies that may or may not end up being useful.
1) We have at least thousands of years of coal left at any reasonable consumption rate (much of it in the US!) so we are not running out of coal.
2) As for that global warming thing, the US government seems to be saying bring it on! So even though I am swayed by this argument, GWB and others don't seem to see a problem with the 1-4 degree (F) increase in temperatures.
3) Worst of all, you can't really claim fusion power will necessarily be cheaper than fission power, and what keeps us from using fission power is largely the fact that it cost twice as much as coal to make electricity. (Though wind power now runs cheaper than coal in some places, even without government incentives, making it a nice alternative.) We have little reason to believe that fusion power would be economical. The plants will be very expensive to build, and will be expensive to tear down (commission / decommission) and since nothing last forever we can only expect to generate power from a plant for a certain period of years (which is where the fudging on 'projected' cost are.) If it is not economical, then we are unlikely to use it, and billions of dollars were wasted that could have gone to school budgets or (more likely) to buying new bombers and helicopters!
I am not saying it would or would not be economical. I have no idea, but it is a complex issue at any rate, and if the govenrmnt thought it was open and shut i suspect we would have fusion power plants already.
Honestly. Although some of the advances in artificial intelligence of been impressive (deep blue, what have you) for the most part they are still disappointing, and defiantly work nothing like the way we think. We have very little understanding of the PROGRAMING of the brain. Until we understand the laws that govern the internal logic of the brain, or of any cognitive system, speculating on when or if this will occur is like talking about those dancing angels. (Though I will assert point 2 will never happen, because evolution would be the only way to create such an undoubtedly complex system. And for too many reasons to count, that will not happen.) Right now, we just don't know.
Slashdot Mirror
Well, assuming that we are Turing machines (which seems probable) then we could say that a mathematical formula is not our way of expressing reality, but actually our way of perceiving it (since all possible thoughts are then just mathematical relations). Since we can never see anything but the mathematical formula, trying to differentiate between reality and the formula becomes fairly meaningless.
Bah, I tend not to trust people who's idea of excitment (here, moral building or something) is indistiguisable from being really really angry. And who can't dance. And who scream thing's like 'Give it up for me." and "I Love this company." I mean really. seriously. But they're funny to watch. Ballmer Dancing and His hypnotic voice.
I am not trying to be contentious, but, I would like to say that I personally find the idea that it would be easier to 'simulate' the growth and development of an entire organism rather hard to envision. But, I may vastly be underestimating the power of computers twenty years down the road, so I will do a back of the envelope calculation. Now, Human DNA codes for around 30,000 to100,000 proteins. It Takes 19,000,000 Gflop too compute roughly 10ns of a 36 Amino acid protein. (1) Lets say that a protein takes around 500ns to fold, (fastest 10ns, some take several milliseconds) Assuming linearity, then an average protein might take 7,900,000,000 Gflop. Now, there will undoubtedly be some short cuts you can take, and you will not have to calculate every inch of those 30,000-100,000 proteins, so lets assume you only need 10%, then a minimum of 2.37 * 10^13 Gflop or (2.36*10^13*2^30 flop) are needed. So now, by moore's law we would expect computing power in 30 years to be what, 10*9*2^(30*12/18)= ~10^15 flops; it takes roughly 7.11*10^13*2^30/10^15 seconds, or roughly
m l; "T3E, four times faster, the next 800 nanoseconds took about another two months" The T3E operates at 307 Gflops.
T3E, four times faster, the next 800 nanoseconds took about another two months.
1 year on the low or 3 on the high end.
Crap.
It appears I vastly underestimated the advances in computing power. If moors law holds out and we develop efficient enough algorithms, then we might have the protein products of DNA computed in 30 years. Well that's pie in my eye. Keep in mind you still have to run interactions between these proteins in cells, which should be an order of magnitude more complex, and do it in trillions of cells over at least 9 months of development, but my minimalist argument has failed. And I am sad.
Just to retain some dignity here, I would like to point out that DNA cannot be treated as a blue print, which is like a scaled down representation of an object, but should be treated like a recipe, (or even more accurately, a chemical synthesis scheme.) which wile it still gives you a very complex finished product, the step beat eggs for ten minutes, has no simple correlation to a particular area of the finished product (if you skip this step you don't get half a cake you get no cake at all) as Dawkins so cogently pointed out. And, I personally still think I am right, but I suppose I will stop arguing at this point. *sigh*
I suppose I could fudge the numbers, I doubt any one will bother checking, but what would the point be? To misquote Huxley; "a beautiful theory, killed by an ugly, nasty little fact."
(1) http://www.psc.edu/science/Kollman98/kollman98.ht
Quick Update, from Duane Chapman's Enviromental Economics; 2000, p132.
Source / Cost per kWh
Solar Photo-Cell / $.50
Solar Thermal / $.33*
Wood / $.12
Wind / $.09**
Nuclear / $.09
Coal / $.06
Hydro / $.06***
* author notes that cost estimates are as low as $.08 from some sources, and that at any rate, should decline around $.07 annually.
** I think this has gone down since 2000
*** By this I believe he means dams, not newer tidal systems. Finally, these rates do not include government incentives for alternative fuel.
Quick Update, from Duane Chapman's Enviromental Economics; 2000, p132. Source:Cost per kWg Solar Photo-Cell$.50 Solar Thermal$.33* Wood$.12 Wind$.09** Nuclear$.09 Coal$.06 Hydro$.06*** * author notes that cost estimates are as low as $.08 from some sources, and that at any rate, should decline around $.07 annually. ** I think this has gone down since 2000 *** By this I believe he means dams, not newer tidal systems. Finally, these rates do not include government incentives for alternative fuel.
1) In my original post I was not claiming that complex systems had to be evolved, which is non-sense (cars, and supercomputers) but rather that complex systems are either a) evolved or b) created. So this idea of AI just 'waking up' proposed in the FA was non-sense. Since it would not be created, it would have to be evolved. And there is good reason to believe from evolutionary theory that this is highly unlikely, being as the computer virus it would have to, by necessity, evolve from, will likely chose the near universal evolutionary answer of simplicity over the extremely uncommon answer of complexity. (Try to imagine a 8 GB computer virus infecting your computer. . . ) 2) Realizing that I am not in the field of AI, and in all likely hood neither are you, so both of us don't know what we are talking about; Your faith in computers aiding us in simulating the mind is entirely misplaced. I will acknowledge that some one may make a detailed map of the brain through any of a dozen different methods, and this may in fact be the way that we arrive at the first AI, and I claimed nothing to the contrary in my post. However, we are talking about speculating on the time it will take us, and not the method that it will be arrived at. I am somewhat skeptical that mapping the brain will be as trivial as you assert. (Though I am not a neurologist and will bow down to superior wisdom if contradicted by a neurologist.) Realizing that not only the place the connections are made between neurons must be mapped, but also the receptors involved (and we still do not know all the neuroreceptors) and we must figure out weather the each receptor on each neuron is acting in an inhibitory or stimulating manor, and figure out by what internal states of the neuron its activities can be modulated, and finally to top it all off we have to understand the function of glial cells, which have recently be indicated to play a role in neurological activity, and make up some 90% the brain . . . I would say I am not all that 'blind' in doubting the 20 or 30 year figure. The brain is more complex than it thinks it is. In conclusion: I'm not saying that this or that will be the way we finally create an AI, or that this or that process is going to be so hard that it will take 100 years to figure out. I'm not saying it is impossible that it will be here in 20 years, just saying I'm not convinced. And that is the point. No one has tried (with much success) to map the brain into a computer and simulate it. The amount of processing power it would take, and the techniques it would require to be efficient are unknown. Unless you have some evidence, such as some one having simulated an ant brain, (which may have been done, but a quick search of JSTOR didn't show anything.) then you don't know how long it will take. Because it hasn't been done. And that is what I am saying. Ps: I consider the LEST feasible way to figure out how the brain works to be looking at DNA, because it ought to be immensely easier just to study the finished product.
I have an alternative theory. Governments are generally unwilling to drop tens of billions of dollars into technologies that may or may not end up being useful. 1) We have at least thousands of years of coal left at any reasonable consumption rate (much of it in the US!) so we are not running out of coal. 2) As for that global warming thing, the US government seems to be saying bring it on! So even though I am swayed by this argument, GWB and others don't seem to see a problem with the 1-4 degree (F) increase in temperatures. 3) Worst of all, you can't really claim fusion power will necessarily be cheaper than fission power, and what keeps us from using fission power is largely the fact that it cost twice as much as coal to make electricity. (Though wind power now runs cheaper than coal in some places, even without government incentives, making it a nice alternative.) We have little reason to believe that fusion power would be economical. The plants will be very expensive to build, and will be expensive to tear down (commission / decommission) and since nothing last forever we can only expect to generate power from a plant for a certain period of years (which is where the fudging on 'projected' cost are.) If it is not economical, then we are unlikely to use it, and billions of dollars were wasted that could have gone to school budgets or (more likely) to buying new bombers and helicopters! I am not saying it would or would not be economical. I have no idea, but it is a complex issue at any rate, and if the govenrmnt thought it was open and shut i suspect we would have fusion power plants already.
Honestly. Although some of the advances in artificial intelligence of been impressive (deep blue, what have you) for the most part they are still disappointing, and defiantly work nothing like the way we think. We have very little understanding of the PROGRAMING of the brain. Until we understand the laws that govern the internal logic of the brain, or of any cognitive system, speculating on when or if this will occur is like talking about those dancing angels. (Though I will assert point 2 will never happen, because evolution would be the only way to create such an undoubtedly complex system. And for too many reasons to count, that will not happen.) Right now, we just don't know.