5
$\begingroup$

The premise: A full-fledged self-aware artificial intelligence may have come to exist in a distributed environment like the internet. The possible A.I. in question may be quite unwilling to reveal itself.

The question: Given a first initial suspicion, how would one go about to try and detect its presence? Are there any scientifically viable ways to probe for the presence of such an entity? In other words, how would the Turing police find out whether or not there's anything out there worth policing?

$\endgroup$
3
  • $\begingroup$ This question seems to be related to or a duplicate of this one. $\endgroup$
    – nbro
    Sep 23, 2021 at 15:11
  • $\begingroup$ To decide whether an AI was self-aware, I would ask it to answer questions about itself, and see if it gave correct answers to those questions. $\endgroup$ Aug 2, 2022 at 1:19
  • $\begingroup$ @AndersonGreen I think that verbalizing self-awareness doesn't proof being actually aware. Similarly, verbalizing empathy does not imply its existence. Related question: ai.stackexchange.com/questions/11927/… But if the system has been given the task to know itself and has had the opportunity to first ask questions about itself to us human sensors, and has built a self-concept from that external coherent information... it can develop what can be labeled as artificial awareness, why not. $\endgroup$ Apr 18, 2023 at 20:39

2 Answers 2

1
$\begingroup$

This question should probably be moved to worldbuilding.stackexchange

That being said, in the context of a story, I would look at something like the neural correlates of consciousness. In this book by Stanislas Dehaene the experiments are described that led to the realisation that conscious perception requires information integration in certain parts of the prefrontal cortex. Basically the brain processes many different interpretations of a percept in parallel. To become conscious of the percept this interpretations have to be collapsed into a single "truth" which is propagated to a specific part of the brain.

In your story you could have a researcher of consciousness stumble upon a chart of information flow in a part of the internet and realise that the same kind of information integration is going on.

$\endgroup$
1
  • $\begingroup$ Thanks for pointing me to those resources! Seems to be what I was looking for: Potentially observable patterns to be expected where human-like intelligence forms or has formed. Concerning worldbuilding.stackexchange - I actually thought about posting there but after lenghty reading of other posts got the impression that a lot of the more tech or science specific questions end up being moved to the respective science sites. So I tried my luck here. But I'd be glad to be moved over, if it seems more fitting. $\endgroup$
    – user4327
    Dec 19, 2016 at 20:34
1
$\begingroup$

"A full-fledged self-aware artificial intelligence may have come to exist in a distributed environment like the internet"

The question implies that this artificial intelligence has surpassed human intelligence (full fledged) and therefore, due to the concept of the intelligence explosion resulting from such a state, the AI you are looking for is no doubt superintelligent.


The question states that this AI is

"unwilling to reveal itself"

and therefore does not intend to be discovered.


Strong (or Superintelligent) AI

Given these two factors (its superior intellect and its unwillingness to be discovered), we can conclude that there is no way in which we would be able to detect such an AI under conventional conditions.

A possible solution may involve employing a second superintelligent AI system, but this is precarious in more ways than one.


Weak AI

Detection of a simpler AI would rely on tracking the pattern of activity and the traces it leaves behind in this distributed network in order to find it, then subjecting it to some form of testing to verify its intelligence.

There are a very large number of possible indicators and these would vary widely depending on the specific AI concerned. This would depend especially on how the AI exists within the framework (here, the web).

Note: a superintelligent AI would not only be able to disguise its activity, but also find a way around any test we can develop - these abilities are what renders this level of AI perhaps the greatest threat to mankind, but also our greatest asset if we do develop it and find a way to gain its alliance.

$\endgroup$
1
  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$
    – nbro
    Sep 23, 2021 at 15:00

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .