The Halting Problem demonstrates that there are things that a machine can never be programmed to do.
Is this proof that true Artificial Intelligence - that is, the ability for a machine to think and reason like humans - is not possible?
4
$\begingroup$
$\endgroup$
asked May 27 '16 at 14:11
-
3$\begingroup$ Possible duplicate of Why can't we solve the Halting Problem by using Artificial Intelligence? $\endgroup$ – adrianN May 27 '16 at 14:46
-
3$\begingroup$ You presume that humans can solve the Halting Problem. We don't know that to be the case. $\endgroup$ – Raphael♦ May 27 '16 at 14:51
-
$\begingroup$ Under the meaning of "true Artificial Intelligence" that I understood, this is not duplicate. But if this question was about Machine Learning that we use to solve something then it is duplicate by all means. $\endgroup$ – Evil May 27 '16 at 15:47
-
1$\begingroup$ A somewhat related question is Human computing power: Can humans decide the halting problem on Turing Machines? $\endgroup$ – manlio May 28 '16 at 10:36
-
$\begingroup$ Along that line of thought Penrose's argument based on Gödel's incompleteness theorem and Searle's Chinese room experiment, however controversial, are more compelling. $\endgroup$ – manlio May 28 '16 at 13:07
1
$\begingroup$
$\endgroup$
If we cannot determine from program description whether it halts we also cannot write algorithm to do so (otherwise we could have determined it from description).
The ability to think and reason like humans is not connected to the halting problem, and the separate problem here is that we do not fully know how people think, so we cannot copy this idea, we are prone to emotions and logic just fails. If you want to create "true Artificial Intelligence" you would have to unlock any restrictions, constraints, let it evolve and think differently otherwise your AI is at the best case clone of you and it will not be self-concious or thinking just ultra advanced processing unit.
If tou like to think about the true AI - human limits are not applicable because if we proved something cannot be done - it is not doable under any circumstances, but if we do not know how to do something (yet) - there is no constraint neither for human nor for the true AI.
-
$\begingroup$ To downvoter: it would be nice to wrote down your objections here to let me fix problem you find. $\endgroup$ – Evil Mar 27 at 18:47
