THere is one overrideing limitation of AI.Wizard22 wrote: ↑Thu Feb 15, 2024 12:56 pm I think you're missing some crucial factors of the current capabilities of AI.
Contemporary AI, right now, is "designed" to "make choices" in almost the same way a parent "designs" his/her child to "make choices". In both circumstances, we are less and less predictive of the outcomes. We hope that the child, or the AI, makes the "correct" choices and outcomes to what we desire, but Choice is finnicky, and imperfect. Sometimes the outcomes are far better than we expect, sometimes far worse. Our Education improves, based on the improvement of better outcomes. We want more better results; and we want better quality results. The better the education (programming), the better the successes. But those are never guaranteed.
I do remember my education/indoctrination/training/programming, very well in fact. Philosophy helped me "un-learn" most of that bullshit. But it also taught me, it's not all bad. SOME discipline is required. SOME fantastical beliefs are worth holding onto. SOME of the fairy-tales, have practical and pragmatic value. SOME traditions are worth reaffirming and continuing. You don't need to believe in Santa Clause, to value family getting together during December and exchanging gifts. It's a nice tradition, without any metaphysical or mystical mumbo-jumbo.
As it applies to AI, obviously its product and consequences, are going to be mostly 'determined' by the programmers, the code, the sophistication and complexity put into it...
ChatGPTs are already passing the Turing Test. So what comes next?
It has no interest, volition, purpose. I t cannot recognise a result in its own favour as it has nothing upon which to base an advantage.
When a child choses to put its hand in a fire it learns about heat and pain.
AI can only respond to things it has been instructed to repsond to.
It cannot undestand what it is doing in ANY sense.