http://www.kurzweilai.net/carnegie-mellon-ai-beats-top-poker-pros-a-first
The Supercomp running the code (learning based algos) is still on the order of 1000x times "weaker" than the human brain so there's some room for improvement ;-) Also, let's not forget that when we build a computer with a comparable number of interconnections (Current estimates say 10^15 [Quadrillion] -- Synapses connecting Neurons in humans) then we will be no match for that computer for 1 main reasons, processing speed! Our synapses fire at ~0.1 - 2Hz (compare to a typical i7 firing at 3 Billion Hz [GHz]), furthermore, the speed at which these signals travel is very slow in Humans (286 Mph) compared to nearly at the speed of light in computers (186,000 Mph -- or the actual speed of light in Optical computers).
If we think about the firing rate as a kind of time keeping device, we can see why seconds became our shortest (in most cases) units of time (1 Hz is 1 cycle per sec)...compare that to a computer running at 1 Billion cycles per sec, convert that to 1 cycle per Nano sec. If the computer equivalent of a second is a Nano second, then 1 Billion Nano seconds (i.e. 1 sec) is a long time to think for a computer...it is the equivalent of saying 1 Billion seconds is a long time to think for a human (10^9 sec = ~32 years!). Consider how much you can accomplish in 32 years! These computers that reach our level of interconnections, with their much faster processing speed can do as much each second! I found a link that show's we are still years away from this (I say only years again due to the exponential increases) so you have some time for that info to sink in and make peace with it haha (I'll remind my readers again, this is why I say we join them rather than try to beat them...I'll leave the details of that for a future post!).
PS Just saw this hilarious comment regarding the last link (simulating 1 sec of 1% of the human brain)
Posts
0 comments:
Post a Comment