Re: What's up with Oblivion? [Spoilers alert]

i own a yacht said...
Mark Morrison <drdpikeuk@xxxxxxx> wrote:

Give PC games another 10 years, and we might have convincing AI -
until, just make do with what we have.

why do we need another 10 years though?

Absolutely. In the limited domain of Oblivion, or any other "real
world" sim, most of the problems were solved years ago - if not decades
ago.

Before someone jumps in and demands two bots talking to each other on a
variety of topics in an intelligible way, you might like to investigate
where we're at already. Thus, limiting the domain to a few dozen domains
- and a few hundred topics - isn't outside the scope of a modern game.

Pathfinding is trivial. Why a NPC selects a route through water, rather
than taking the next street is silly. Presumably the devs have simply
used the shortest path algorithm without any mods. That's nuts.

A basic technique for searching nodes is to use recursion. It's a hammer
to crack a nut, but it works. Recursion is taught in comp science before
university these days, so it's a basic skill. A typical task to
demonstrate a grasp of this knowledge would be to ask students to find
the shortest path, say, for a knight to traverse all squares of a
chessboard without landing on the same square more than once.

As a bonus question, you'd remove, say, four squares, and ask them to
solve the new problem.

My point here is really just to let folk know that the "AI" domains
within Oblivion are not quantum mechanics. We gamers should be demanding
- or, at least, asking nicely - for better things, and this is an area
where games could be vastly improved for a lot less effort than, I
suspect, folk realise.

the only thing preventing
developers from delivering good AI right now is how much they value
it. and it's obvious the majority of game developers don't value good AI
very much at all.

I suspect that it's that old geek chestnut: not invented here. What
they've produced, in terms of AI, is an average attempt for an undergrad
project, whereas if they'd employed an AI PhD geek, they'd have been at
the stage they got to at the end of the first day :-)

i mean, a lot of them talk about it and then produce a
game with AI as shitty as any other game. a good example is half-life
2. AI was supposed to be one of its advances, and it's horrible. barely
an improvement over doom. say what you want about far cry but the AI is
without question better than almost any other fps out there. crytek take
AI seriously, that's the difference. they respect the difference between
good AI and passable AI.

I don't enjoy FPSs - never have - but I bought Far Cry when I built a
PC, because it was the most testing game out there at the time I built
the machine. I loved the game - and look: no story :-) As you suggest,
the AI was vastly superior to Oblivion.

It's also worth looking at the amount of audio work they put into that
game; that's another weakness in Oblivion, but then I do a fair bit of
post production work, so I'm going to notice those holes.

and far cry's AI isn't even that amazing. that's how bad most AI is. i'm
sure most gamers are perfectly happy with passable AI anyway. just look
at any review of oblivion. i'm tired of it though. i don't need better
graphics. i want some friggin' decent AI, for frig's sake!~

If this is the general state of gaming AI, then someone in those
companies needs to do half a day's homework.

--
Best,
Marc
.