Time Perception in Synthetic Intelligence: Why It's Important

By Jason M. Pittman, Sc.D
 

I previously claimed that a synthetic intelligence will not perceive time or, at least, not tell time in the same manner as humans. Further, I noted that if my claim turns out to be true, there was a measure of import to the consequence.

Before I outline that consequence, as well as why I think the lack of time-perception in synthetic intelligence is important, we ought to know the context, right? Certainly you have read or heard about the existential risk an artificial general intelligence (AGI) may pose to humanity. But you may not have heard about the work going on to develop a means of containing or boxing such an entity, thus rendering it safe(r).

Suffice it to say, containment is an inordinately difficult problem. Some have proposed leveraging cybersecurity technologies. Others have suggested that keeping an artificial general intelligence boxed is a pipe dream. I disagree with both positions.

If you want a strong example of how containment needs to function, watch The Truman Show. The safety of the main character -- Truman -- as well as those external to Truman's world depends on Truman not coming to know he is in a simulation. Spoiler alert: exactly the opposite occurs. Truman breaks out so to speak. The reason Truman breaks out is that technological containment can fail. Given enough time, given enough interaction between Truman and what he takes to be reality, the illusion will inevitably collapse. To quote the great hero, Dr. Malcolm: life finds a way.

Now, imagine the same scenario with an AGI and assume, like Truman, AGI will find a way. It would be a massive understatement to say that there are more undesirable outcomes associated with an uncontained AGI than positive outcomes. Further, I would argue that a perception of safety through technological containment is miles worse than no containment at all. If we assume containment is untenable, we can choose to not have AGI in the first place. On the other hand, once we have AGI in a box, the basis of safety rests upon the continued integrity of the box. That, in my opinion, is pure folly. Should we abandon AGI altogether then?

No. I think time may be an answer. Let me explain.

Truman broke free because the illusion of his reality was based on technological controls that perfectly mirrored true reality outside. However, what if Truman's perception of time, his ability to tell time, was altered in such a way that he was always n + 1 behind true reality? Stated another way, imagine if the world around you passed one hour faster than you were capable of perceiving. Simply, you'd always, forever be locked in the past, relative to the true reality, by exactly one hour.

In this way, not only are you contained but any attempt to circumvent containment is a past event relative to us on the outside of the time bubble. I don't think such containment is invulnerable but I think the advantage engendered by time is unconquerable. Thus, time is the perfect containment mechanism and that is why the lack of time-perception in synthetic intelligence or AGI is so important.