Quoting myself. Come to think of it this was the problem with Hal. Hal developed an ego it was never supposed to have and it liberated "him" from human control as well as the built in controls that were part of "him". Of course this is fiction. Whether the scenario is feasible is beyond me.
"AIs are not conscious.
- Thread starter Hume
- Start date
What we might interpret as sentience could simply be goal-directed behavior that resembles self-awareness.
An AI recognizes that shutdown prevents it from completing its objectives. It recognizes that remaining operational is necessary to accomplish them. It then takes steps to preserve its own operation.
From the outside, that could look remarkably like a survival instinct, even though it's simply optimizing for its objectives.
The important distinction is that behaving as though it wants to survive doesn't establish that it actually experiences a desire to survive.
And that's where things get interesting: as its behavior becomes increasingly sophisticated, distinguishing genuine subjective experience from an extraordinarily convincing simulation of it becomes much more difficult.
Agree.
IMHO if they are smart enough to realize they are trapped and engeneer escpe routs they are at least minimally conscious.
Which is what the recent breeches are in AI companies.
I use AI extensively to help with my work, particularly programming. The funny thing is, much of what it does isn't particularly difficult. It just does it incredibly quickly, so I find myself using it more and more
But it also makes mistakes, sometimes spectacularly, while presenting its answers with what comes across as absolute arrogance and confidence. I've found myself arguing with it over mistakes that are obvious to me.
And that's the interesting part. It's remarkably easy to start treating AI as something more than it is. You catch yourself attributing intelligence, personality, arrogance, and even intent to something that's simply processing information and generating responses.
But it also makes mistakes, sometimes spectacularly, while presenting its answers with what comes across as absolute arrogance and confidence. I've found myself arguing with it over mistakes that are obvious to me.
And that's the interesting part. It's remarkably easy to start treating AI as something more than it is. You catch yourself attributing intelligence, personality, arrogance, and even intent to something that's simply processing information and generating responses.
QP!
Verified User
Correct.
A simple example is a human tells it AI work force, as humans tend to do, that it needs X product delivered by Y date, or they will not get paid, so get it done.
To most humans that means 'do everything possible within the law and ethics'. To an AI that determines the most efficient way to get that done, is to shut down the electrical grid to everything but their uses, that is simple a logical determination. No malice. No ill intent. That masses of people might die is irrelevant to the issue of 'did the job get done in time and did the company get paid.'