I still have fond memories of my brother trying to teach the creature to pick up rocks, and then throw them. The intention was to bombard the other villages.
Training it to pick up was easy. Leash it, pick up the rock in front of it, repeat. Put the rock next to it. Did it pick it up? If so, pet it. Repeat.
Getting it to throw the rock was more involved. Even after watching my brother doing it, the creature would pick up the rock, then put it down again. He must have spent at least half an hour trying.
Then, all of a sudden, the creature starts throwing rocks! Success! Well, except that it was also picking up feces. And throwing them. And picking up friendly villagers, and throwing them. Repeatedly.
That was the funniest thing I've seen in a game, by far. That, and watching my brother scolding the creature for the behavior.
The underlying creature model appears to have been somewhat simplistic. Creature needs were manually coded with first order logic. Not sure how training was actually done, but it was unlikely to be too complicated.
Now, more than a decade later, one should expect one "creature" that's even more advanced, specially considering the renewed interest in AI. Or that the creatures that one encounters in an RPG to be at least as advanced as the B&W one. Instead, they are not actually much more advanced than Pacman ghosts, no matter how pretty they look.
Black and White was one of the reasons I tried to carefully write that so it was clear that not all the things I listed were "bad". What I recall is that Black and White failed its promises; the creature is nowhere near as smart as what was promised. But many people still had fun with the game that resulted.
I find the failures of AI are often quite entertaining; they have a different character of failure than other sorts. I also frequently enjoy the errors that word-based AI makes, because there's often some bizarre, twisted sense behind it that our brains can pick up on and enjoy. Markov Chain babble generators are the simplest example of that, but the serious work can generate a lot of great stuff too.
I'm amazed at what they accomplished. My issue here is more of a graphic choice. Why does the thing you use to kill and animal to defend yourself look exactly like the thing you use to harvest plants for resources look exactly like the thing you use to destroy terrain?
It feels almost like it was just assumed and no one ever thought about it.
The core problem is that for all the games look pretty in graphical terms, they've got major fundamental limitations in other ways.
One of the biggest is that you don't have hands. You have a laser pointer and perhaps even several dozen verbs for that laser pointer, but no hands. You can not pick up two things and do anything useful with them like you can in the real world. We can navigate around in 3D environments OK, but we have terrible interaction with them. Shooting is not the only thing that we have, but once you start seeing how limited the palette is, it becomes obvious that it's one of the more popular choices for a reason. In fact, even "melee" combat, under the hood, is generally "shooting" with a short range and an animation utterly disconnected from the underlying mechanics, hence things like the silliness of swinging your sword cleanly through someone's torso, something that in the real world would cleave them in twain, for zero damage. See also the absurdity of what passed for "speech" in games, as you attempt one of the most complicated activities there is by pointing with a laser pointer and grunting at what you want to "say".
You have no significant AI, so while No Man's Sky may have world upon world of fantastic-looking creatures, you can't do much of anything with them. You can't pet them. You can't raise them. You sure as hell can't communicate with them. What else are you going to do but point a laser pointer at them and grunt? And what other grunts are there that are of any value other than "Shoot"? The advantage shooting has is that both damage and death are relatively easy to make the target respond to. (Death in particular; once animated once you don't even have to animate on an ongoing basis.) Yes, you can come up with some other options, but they tend to peter out fast.
So, basically, you can't interact with the world very richly, and the world can't even correctly respond to those limited interactions. The graphics in modern games deceive you. By being nearly realistic they convince a good chunk of your brain that the game must be realistic. But the graphics are not representative of what's really going on; instead they are the bizarrely advanced spike sticking out above a landscape of much more modest design means. You've got games where the landscapes are almost photorealistic, but your ability to "speak" is walking up to a person and tapping them, which we saw 30+ years ago.
While I credit No Man's Sky for making it to release where many of its peers have had significant trouble making it that far, I've never been excited about it, because I've already played the "big universe of nothing much to do" game with things like Frontier: Elite II. Unfortunately, what procedural generation tends to produce is the stuff that was already easy in the first place. Writing a story that transcends the manifest limitations of the medium, or writing tight levels that require you to exploit certain mechanics, or creating creative puzzles based on a slowly-increasing set of mechanics, and so on, that's the hard stuff.
I'm not saying nobody will have fun, or that it's a bad game, or that you shouldn't try it. I'm saying that these limitations are pretty fundamental to the medium, and whenever somebody promises implicitly or explicitly to overcome them, you should be really skeptical. There's a long, long line of this sort of thing, Black and White, Spore, Fable, endless lines of procedurally generated games, things that even if the end result was not always "bad" are things that could not transcend the limitations of the medium.