Shadowfiend mid vs. Shadowfiend mid (which doesn't make sense because Dota 2 doesn't allow opposing teams to choose the same hero) would indeed be an extremely micro-heavy ordeal. This was very fascinating.
+1
+1
-1
Vote comment up/down
ubershin (Old Spike)
there are some caveats to this AI. they had a few rules- no runes, no raindrops, and some other stuff I forgot in detail. Impressive but still limited in its early stages.
+1
0
-1
Vote comment up/down
blue_devil (Short Spike)
Beating humans in chess and similar games is just a matter of calculating more possibilities then the human and picking the best move. It's basically based on a huge decision tree with a huge math basis. It's almost completely math and computers have been beating humans at math for a very long time. You may say that math is dead simple and stupid at the same time.
Computers will always have the upper hand in straight up linear calculations like math problems but make the environment more like reality and they fail miserably. This is the realm of parallel calculation and even closer to reality would be unclear rules . This is the reason we don't have maid robots like we thought we would in the future. For this they need artificial neuronal networks aka animal brains.
The last calculus revolution brought about the transistor and microprocessor, the next would bring about something similar to a neuron that you can link and re-link in networks and poof animal to human to god level A.I.
Given a complete map of a animal or human brain with all the charge potentials at every synapse and all the complicated stuff in the background we could probably build something that could ether simulate a animal or human brain or even an A.I that rivals humans but the thing would be fucking massive i mean several stadiums of computing power and probably a nuclear power plant to run it. So it would not be feasible any time soon. For that we need to go to a new field of science aka nanotechnology.
Nature on the other hand has been evolving and optimizing neurons for at least several hundred million years if not over a billion where as we just scratched the surface on nanotechnology a few decades ago so i wouldn't be to worried about human level A.I for the next several hundred years.
Comments
(Long Spike)
Shadowfiend mid vs. Shadowfiend mid (which doesn't make sense because Dota 2 doesn't allow opposing teams to choose the same hero) would indeed be an extremely micro-heavy ordeal. This was very fascinating.
(Old Spike)
there are some caveats to this AI. they had a few rules- no runes, no raindrops, and some other stuff I forgot in detail. Impressive but still limited in its early stages.
(Short Spike)
Beating humans in chess and similar games is just a matter of calculating more possibilities then the human and picking the best move. It's basically based on a huge decision tree with a huge math basis. It's almost completely math and computers have been beating humans at math for a very long time. You may say that math is dead simple and stupid at the same time.
Computers will always have the upper hand in straight up linear calculations like math problems but make the environment more like reality and they fail miserably. This is the realm of parallel calculation and even closer to reality would be unclear rules . This is the reason we don't have maid robots like we thought we would in the future. For this they need artificial neuronal networks aka animal brains.
The last calculus revolution brought about the transistor and microprocessor, the next would bring about something similar to a neuron that you can link and re-link in networks and poof animal to human to god level A.I.
Given a complete map of a animal or human brain with all the charge potentials at every synapse and all the complicated stuff in the background we could probably build something that could ether simulate a animal or human brain or even an A.I that rivals humans but the thing would be fucking massive i mean several stadiums of computing power and probably a nuclear power plant to run it. So it would not be feasible any time soon. For that we need to go to a new field of science aka nanotechnology.
Nature on the other hand has been evolving and optimizing neurons for at least several hundred million years if not over a billion where as we just scratched the surface on nanotechnology a few decades ago so i wouldn't be to worried about human level A.I for the next several hundred years.