NPC AI is just "get a random value, then take a random action based on that value" so it's technically 100% human (since they wrote the thing in the first place)
Even genetic algorithms are still human.

To me, the issue is mostly that what's being referred to, these days, as AI, for kicks and giggles and to get people all hyped about it (and to generally disregard the massive data mining that comes with obligatory accounts and all that jazz), is not actually AI.
Or it's to "Artificial Intelligence" what we now call VR is to "Virtual Reality": a far away approximation of one of the easiest aspect of what would actually be the real thing.

It would definitely be really creepy if it had any actual intelligence because it's learning from the internet and the last thing I'd want if I had a kid is to let them loose on the internet to educate themselves XD

That depends on your definition of artificial and human. I don't think that you can ask that kind of question without defining the terms and hope to get even somewhat relevant results.

Interesting for games would be AI in NPC dialogues as was showed off as a rudimentary proof of concept by Nvidia. Not that it's a new idea entirely as we had companion Vilja like one and a half decades ago already.

I didn't go thru the thread... if someone hasn't suggested, but maybe watching The Terminator will help? https://www.imdb.com/title/tt0088247... lol

Just here for reading comments and feeling intellectual.
This is a nice post man!

Definitions of intelligence can vary but according to how its most often applied current gen AI can be considered intelligence- we have to keep in mind there isnt just human intelligence. Biology recognizes and some scientists even make ranks of the levels of intelligence in the animal kingdom - and when weve been labeling insects to even some forms of primitive intelligence on that scale AI definetly fits the bill.

Heck the amount of neurons far surpasses some forms of life that we recognize with primitive intelligence.

I think it would help to make that distinction... life, natural. Current AI surpasses many animals but its whole artificial, man made.

By the word alone i would say 100% artificial BUT i dont think that was the intented use- if by artificial youre meaning 'fake intelligence' as just replicating/remixing data it was fed with... its currently at mostly 'fake', not 100% anymore

Ive been (trying) to keep up with the field, glanced at some of the papers but i mostly follow some researchers and content creators for their wider net... we have clear evidence of SOME level of modeling happening - wich is at least some models been developing some kind of model of the world or the enviroment, crucial to intelligence. Its nowhere near our level and probably bellow the models of mamals but its some model none the less - and heck scientists description of lowest forms of intelligence in the animal kingdom dont perceive then as having models

Im no speciallist but im quite confident the 'fake' pure data regurgitation fitted the earlier models only, the dumbest chat bots.
That doesnt mean AGI is upon us. A lot of the debate gets muddied where many take any indication of actual intelligence as the robots mastermind will be threatening us in 5 years...

But we could well be under our way (to AGI, not that it will be that threatening). Weve noticed some level of modeling, the way the digital neurons work in tanden is pretty much how organic neurons do and frankly AI wont have anywhere near our amount of neurons for awhile. Current gpt4 is like the amount of a squirrel, and some optimistic projections expect human amounts by 2030(but i wouldnt bet on it- could be later, with luck sooner).
What can make current AI seem smarter isnt neurons or intelligence its still the data - in a way its kinda of proving how much experience matters for most things in our fields- it doesnt need to be as smart or smarter then a human to pass a test or write an essay on human level when its memory contains 90% of all experts work on a subject. So in that sense its kinda 'fake' but... really?

What i mean is that most of what humans do the majority of time dont require that much of our intelligence to begin with. Our brains run on auto most of the time because the smartest part of our intelligence is extremely resource consuming- most of the time what our brains do is exactly what the earlier models of AI did. A teacher explaining a subject isnt analysing, making decisions or deducing anything just on auto regurgitating what he already knows well; When were chatting were not ever stopping to analyse or create what we will say, were responding with the same sort of neural process that is regurgitating our data, we just dynamically adapt the output without even thinking... wich is like what gpt do most of the time.

What we consider more logical thinking can happen in current AI but for now were only sure it happens when prompted for it. Researchers been testing prompting for slower and more deliberate step by step responses by AI and when prompted like that the models ponder much more and re-tread steps reaching for smarter responses, problem solvind and deductions... wich is how we do it. Right now it doesnt do it often or without prompt but the evidence of mental models show that the AI been doing a little bit of that on the background by itself so...