(06-14-2022, 05:55 PM)jagibelieve Wrote: The simple answer is, anything "AI" is a computer that is programmed.  A computer only understands a "1 or a 0" for the most part.  Everything is absolute.

A computer can theoretically "learn" from the input that it receives, but the output is always going to be absolute.

Contrary to some segments of the population, a binary machine is always going to be that.  Either true or false, 1 or 0, + or -, it doesn't really matter.  It's a machine and will do only what the programmer tells it to do.

In the traditional imperative paradigm of programming, that's mostly true. A given input gives deterministic results (there are exceptions if bugs creep in or for certain kinds of data structures and the algorithms used to implement them). Even "random" generators aren't REALLY random.

AI system's are stochastic which means those systems aren't designed to produce deterministic output. Developers usually have no idea what results their systems will produce for a given input.

How can non-deterministic behavior arise from a system based on binary storage? That's the trick of the analytical models that the AI developers use. Systems like neural networks are highly sophisticated and designed to emulate biological networks. There's a school of thought that any sufficiently complex system has the potential for self-awareness. Our brains host consciousness (well except for Bills and tack fans) that arise out of individual neurons exchanging electrons. The phenomena of self-awareness isn't a factor of two neurons exchanging an electron but in the relationships of all the trillions of neurons to each other as a whole and the ability of the system to form new relationships between neurons. The debate is if the AI systems are anywhere close to the complexity needed to cross that boundary. Personally, I don't think we're anywhere close to that.

---