To solve problems using computer programs, we have developed a wide set of tools / control flow statements such as For Loop, If-Else, Switch-Break Statements and so on. How natural are these control flow statements? Since, we do not know, how exactly would AGI work, and the understanding of Neural Networks, so far, does not tell us about origin of "intelligence", Could a modern AI evolve itself into a system of such control-flow elements? (An appropriate architecture for computation is , anyways, necessary for any inorganic intelligence system and it is of no doubt that control-flow statements as such just makes the computation much more organized) If so, could an AI be killed in an infinite loop created by itself?

P.S. The question isn't baseless, it really questions the kind of computational infrastructure a modern AGI would require, and would it be able to alter itself without any intervention.

  • 1
    $\begingroup$ Maybe take a look at "Could a paradox kill an AI?" at ai.stackexchange.com/questions/1768/could-a-paradox-kill-an-ai $\endgroup$ – GJZ Dec 21 '16 at 23:05
  • $\begingroup$ @GJZ, The questions are very different. $\endgroup$ – Parth Raghav Dec 23 '16 at 21:53
  • $\begingroup$ But the answers are quite alike $\endgroup$ – GJZ Dec 23 '16 at 22:05
  • $\begingroup$ Would a human create an infinite loop for itself? $\endgroup$ – MilkyWay90 Jun 5 at 16:43

This is a hard to answer question because a truly correct answer would involve static analysis of a given Intelligence to determine whether it has the computational capability to generate a looping state (e.g. some state which reproduces itself in the next instance), and in fact whether these looping states can even exist in the given architecture.

Generally speaking, assuming we can build AI, there is no reason it would be impossible build an AI that has the potential to get stuck in a computational loop and become non-functional. You can imagine a recurrent neural network where the continuous state inside the ANN dominates its behavior causing it to reproduce states with small variety (as input will always be relevant in a connected ANN) but are converged around a particular state space. Whether that is akin to death is a philosophical question for your own peace of mind.

However, if we designed an AI agent that commonly did this, that would reflect poor design; that wouldn't necessarily devalue the contribution of the designed agent but it would bring the agent's other activities under scrutiny. If this is a bi-product of some simulated intelligence, is the process going on really that intelligent?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.