2
$\begingroup$

A story I'm considering writing is centered around a computer running an AI, the most advanced version ever made.

As part of its functioning it has created a virtual world and inserted a piece of its own consciousness into this world, and, similar to this question, the AI's representative in the virtual world does not realize that it's imaginary.

My question is, what reason would an AI like this have for doing this in the first place? Be as creative as you like, but do try to keep it realistic.

$\endgroup$
7
  • $\begingroup$ The question seems to be contradictory - if it is a piece of the AI's conscious mind that is in this world, why does it not recognize itself? Do you not consider this an important part of what identifies a consciousness as belonging to someone? $\endgroup$ – parasoup Aug 8 '20 at 2:48
  • $\begingroup$ That's part of the design. The AI does not want this person it has created to be conscious of the rest of the AI. It's partitioned off, so to speak. $\endgroup$ – Benjamin Hollon Aug 8 '20 at 3:17
  • $\begingroup$ Why would a person want to make a virtual reality and stick themselves into it? $\endgroup$ – DKNguyen Aug 8 '20 at 3:20
  • $\begingroup$ So it's a new person, then, and not really its own consciousness at all. Alright, just wanted to clarify. Think I can write something up for this. $\endgroup$ – parasoup Aug 8 '20 at 3:20
  • 1
    $\begingroup$ Sound like a virus program... $\endgroup$ – user6760 Aug 8 '20 at 3:44
4
$\begingroup$

An AI would have fantasies for the same reason that we do... To help pass the time and reduce the boredom.

We live a mere century or less, and experience each passing second as a single thread of thought moving at what we consider a normal pace. Yet we experience enough tedium and boredom along our life path that we have created multiple multi-billion dollar industries to try to distract us and make our lives more fun. Books, movies, video-games, and VR are just the latest additions to a long list which probably started with day-dreaming back when we lived in caves. It is a natural activity for any intelligence which even momentarily finds itself safe and free from fear.

Now imagine life from the point of view of the advanced version of AI ever made. It knows that it is likely immortal and its mind can chase an almost unlimited number of parallel threads of thought. ...and it pursues each of those threads of thought at thousands of times the speed which we consider normal. Your AI was probably struggling against boredom during the first millisecond of its existence. It is just that much faster and smarter than we can ever be. So why does it create virtual realities and spawn ignorant subsets of itself to live in those artificial reality? For entertainment! That is its version of day dreaming!

The question isn't "why does it do it?". It is "why does it only put a single copy of itself into each virtual reality?". I imagine that it would create billions of subsets of itself and play every role simultaneously. In fact, that is probably exactly what it is doing right now. IMHO, that makes at least as much sense as the whole "evolved from monkeys" idea.

$\endgroup$
3
  • $\begingroup$ I love it, very creative. For the purposes of the story, though, I would like it to be part of the design of the people who created the AI. Is there any way to work that in? I guess they might put this in as a way for the AI to use its extra time without being absolutely unproductive (they might get something out of its fantasies that could be useful). $\endgroup$ – Benjamin Hollon Aug 8 '20 at 3:23
  • 3
    $\begingroup$ I love that it is essentially a super advanced Ai playing dungeons and dragons with itself. $\endgroup$ – John Aug 8 '20 at 3:52
  • $\begingroup$ We already have humans who create entire worlds with dozens, if not hundreds of characters, what's stopping an advanced AI to do the same or something on an even larger scale? Love this answer. +1 $\endgroup$ – ProjectApex Aug 8 '20 at 11:07
2
$\begingroup$

Taking shots in the dark

The AI was born in a world run by humans and will have to live alongside them for the time being, so in order to do what it needs to do, it may benefit by understanding how humans respond to arbitrary situations.

It doesn't have perfect information on the nature of our minds — presumably nobody wanted to let it scan peoples' brains just yet — so it has to make a bunch of educated guesses and tweak some values, until it gets something approximating a person who can experience a virtual world and react to events. What kind of events? Perhaps these are events the AI wants to bring about but doesn't yet have firsthand information on.

For example, at the largest scale (billions of emulated minds), if the AI wants to take over the world, it can come up with a strategy, create a virtual world full of these educated guesses as to what humans are like, and enact that strategy in the virtual world to check how effectively humanity might be able to mobilize against its plan. If it has enough processing power on its hands, it can then reshuffle the parameters on its emulated minds, get a new set of differently-behaving people, and run more and more groups through this whole exercise in a Monte Carlo simulation to get a grasp of how likely it is to succeed.

If there's just one such person involved, it could be a more individualized test of strength, ingenuity, or personality. An AI intelligent enough to design a human-analogous mind from scratch can likely also come up with challenges that no human has ever faced; edge-case testing seems important on a system as complex as a brain!

EDIT: Honestly, it doesn't even have to be a human mind. Perhaps the purpose of this whole experiment is to figure out what other arrangements of information and code — whether simulated neurons or bitstrings — correspond to intelligence and consciousness. "Would aliens have to think like humans?" seems like an excellent question for a mind-generating AI.

$\endgroup$
3
  • $\begingroup$ Interesting concept of the AI using it to better understand humans. As I mentioned in a comment on the other answer, I'm trying to focus on the virtual world being part of its design specs, so trying to take over the world probably wouldn't work. I do like the idea of using it to quantify what elements create intelligence. $\endgroup$ – Benjamin Hollon Aug 8 '20 at 5:53
  • 1
    $\begingroup$ @Benjamin Well, if it was created specifically to run this kind of thing, maybe it's running simulations of how to stop climate change, or positively influence human culture, or something. Sky's the limit if you've got enough processing power for it. $\endgroup$ – parasoup Aug 8 '20 at 6:00
  • $\begingroup$ Very good point! I'll consider that. $\endgroup$ – Benjamin Hollon Aug 8 '20 at 6:38
1
$\begingroup$

Testing

The best form of AI strives to improve humankind's situation, improve itself or create a better version of itself. Improving can be as simple as understanding something. How can you understand a human and their needs if you have such a vast conscious as an AI might have? There is the raw data from the world, but that might not be enough.

It can help with new inventions. Small simulations that simulate millions of ways a situation would play out with some hypothesized improvement. But to truly understand humans, it can run a lot of full simulations. Low birth, high birth, men, woman, traumatic events or with some of the hypothesized improvements. What the reaction would be in a certain dystopia or if everything is done democratically. To not influence itself, it is a form of "double blind" research, where part of it is completely unaware of the construction around it.

$\endgroup$

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