I talked with a graduate computer science who said one challenge of making artificial human-like is making random decisions, and that computers can't be random, that they always need a "seed." But, if a computer's outcomes are determined by the chaotic movements of electrons, it doesn't seem like it should be difficult to program inherent uncertainty into a computer. So, what exactly is stopping people from harnessing this basic component of reality to allow artificial intelligence to make randomized decisions? I mean, all you'd need is different neural pathways that rely on the superposition of electrons, and that's it.

  • $\begingroup$ Simply because we are at a level of Artificial Narrow Intelligence or ANI $\endgroup$ – quintumnia Jul 2 '17 at 10:15
  • 1
    $\begingroup$ True Randomness doesn't exist in classical physics (or world). It only exists in Quantum World. Systems doing random things is not very useful in itself. Though today we have many that can stimulate a certain degree of randomness that is useful in various fields. Imagine a system that performs its actions randomly and decided randomly to kill you or any other human or something even worse. Careful what you wish for. $\endgroup$ – Ugnes Jul 3 '17 at 12:21

Computers (the processor in combination with the memory) are designed to be deterministic. Otherwise no software would ever work, because the computer would be executing it randomly.

The computer's outcomes are not determined by the chaotic movements of electrons, but by the much strong deterministic currents and voltages between the transistors. A digital machine is, simply speaking, not affected by noise. The noise is always present, but 0.13 volts or whatever gets rounded to 0 V (logic 0) and 3.02 volts or whatever gets rounded to 3.3 V (logic 1).

It is however possible to take advantage of the chaotic movement of electrons as well. This is done in special hardware called "hardware random number generators", (some use different forms of true randomness such as nuclear decays).

| improve this answer | |
  • $\begingroup$ I mean, I can't really find where I stated computers aren't currently designed to be deterministic, so I'm not sure why you brought that up. I'm speaking in reference to making an AI random so that it emulates chaotic decision making that humans sometimes have. I'll look into the hardware random number generators though. $\endgroup$ – RayOfHope Jul 1 '17 at 23:43
  • $\begingroup$ You can't simply program it to become analog. It actually requires specialized hardware. $\endgroup$ – Oskar Skog Jul 1 '17 at 23:45
  • $\begingroup$ Right, but I'm wondering why that has been so difficult given it's such a fundamental property. $\endgroup$ – RayOfHope Jul 1 '17 at 23:55
  • $\begingroup$ It's not difficult, it's just not needed in most cases. There's enough randomness from the outside world (microsecond timing of IP packets and mouse movements for instance) to seed a cryptographically secure pseudo random number generator. $\endgroup$ – Oskar Skog Jul 2 '17 at 0:00
  • $\begingroup$ And installing a high quality hardware RNG is not difficult, but it costs money and very few people would ever need it. $\endgroup$ – Oskar Skog Jul 2 '17 at 0:01

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.