LogFAQs > #910428680

OhhhJa posted...

A.I. only gets better as time goes on and I dont see why that progress would halt any time soon.

Two reasons are pretty blatant:

1) Humans are incredibly paranoid about the idea of creating something that eventually turns on us. Even if we ever achieved a level of understanding where we COULD create fully sapient AI, odds are we'd go out of our way to NOT do it for decades, at the very least. And even if we ever did, we'd slam so many safeties and failsafes into it that it would be almost impossible to achieve any sort of self-reinforcing escalation.

(See also, the same reason why we'll likely never have Von Neumann machines or unrestricted nanotechnology even if they become possible, unless our species as a whole undergoes a radical paradigm shift - which isn't going to happen over decades, and likely not even over centuries.)

2) Diminishing returns are a thing, and at some point things start running up against concepts like the uncertainty principle. It's why, in spite of the fact that we've spent almost 50 years constantly inventing newer and newer hard drives that can hold more and more data, we're starting to push up against the limits of what's possible to compact into a small enough space. To counter it we're considering radically different means of storage (most of which are currently only hypothetical), but there's still a built-in wall to the constant improvements and refinements of development. In the same way, we may easily find that we reach a certain point beyond which we cannot proceed, for reasons that we don't even know enough to predict at our current level.

(In the same way that it's been suggested it would be literally impossible to create a perfect simulation of the entire universe because it would require more power than exists IN the universe to run it, it's entirely possible that we'll never be able to model fully sapient and free-willed AI on par with the human mind because we can never anything more complex than our own)

Most assumptions that AI development will not only continue apace but run up against no blockages of any kind, and ultimately reach a point of self-sufficiency, are effectively wishful thinking.

Which is not to say that it CAN'T happen that way, but there's also no guarantee that it ever will. And plenty of reasons to suggest that it won't.

It's also worth noting that, no matter how often American schools teach the idea, progress is not an infinitely increasing arc that always advances and improves. Modern phone technology shows us that we can keep shrinking the tech, but we may eventually reach a point where we don't WANT to make it go any smaller (because we go past the point of convenience in the opposite direction). In a similar vein, we've seen that advances to one technology may render other technologies obsolete - we basically stopped trying to improve pagers and fax machines because they've mostly been superceded by alternate tech. Having reached the moon, we basically shifted our entire space program to the point where we were literally no longer capable of reaching the moon without radical redevelopment.

And hell, we keep trying (and failing) to come up with new versions of 3D and VR every other decade or so, with tons of stutter-steps and complete re-evaluations of praxis between iterations. And the tech often being pushed before it's actually ready does more to set it back than it does to encourage it.

The fact that AI research exists at a certain point now doesn't necessarily mean it will automatically be more advanced in the future, or that it will advance in ways that are easy for us to predict today. Any number of other factors or outside circumstances could easily derail the entire projected arc and divert it down different paths.


---

"Wall of Text'D!" --- oldskoolplayr76
"POwned again." --- blight family