Benefits of higher performance Console CPUs (Game AI) *SPAWN*

Ok, let's clear it up a bit.
There is a game called Dragon's Dorma.
There are AI party members there.
They:
1. Chat, sometimes quite convincingly (but not always).
2. Gradually learn how to kill monsters faster with each encounter (yes they do)
3. Warn you about monster tactics and what they will do.
4. Know how to hold certain monsters to help you stab/punish them easier.
5. Know how to climb heights to get an advantage.
6. Know how to carry unconscious party members away from battle close to healers/mages.
7. Know how to use magic cast by other party members to their advantage (ex. climb an ice spear to run on it towards monster's head, if it's too tall).
Etc. etc., I did not even mentioned all physics interaction you can have there, like nailing a goblin to a wall with an arrow.
Quick quiz, from the people in this thread: Who heard about it? Who played it? What were the sales numbers?

 

I wanted to play it (got it for free with PS+), but the abysmal performance made me not do it... if it were on PC, I would probably try again.

 

Dunno what abysmal means here, AFAIR Dark Souls was far more abysmal than that (even on PC) and still got quite a lot of sales.
But I agree that releasing a PC version would probably expand the market for this game.

 

Well, the game runs at under 30Hz in all but the least demanding scenes. And those usually aren't fighting areas. DS and DeS both had horrible perfomance as well... and DS1 and 2 got PC versions, and that's where I played them.

 

Tbh I'd rather the experts in this field apply their talents to things like curing cancer or anything with meaningful value rather than wasting it on a big elaborate toy. Mainly scripted AI is fit for purpose in the case of games.

 

or getting us to mars...

 

Play it on 360, and the AI is not also smart, than you tell, in gameplay, like all AI in RPG, too many time they failed and use too more time the wrong spell, block on path, and all you describe is simple scripts. The first DA, on 360, have AI with many player choice and be more useful than the Ai in Dragon Dogma.

 

Increased available memory bandwidth due to larger cpu caches and less apu memory contention from accessing the main memory.
Possibly a slight to mediocre increase in available shader performance for 3d rendering depending on the game due to fewer gpgpu tasks being run on the gpu to compensate for the inferior cpu.

 

First DA had player-scriptable AI that was ripped off mostly from FF12. And it worked much, much better in FF12.
Anyway, Dragon's Dogma has an order of magnitude more interactions than DA or FF12 anyway. And yes, sometimes AI is stupid and does strange things.
But frankly a random group of people in coop mode would do equally stupid things.

 

I finished it and the DLC. Best RPG last gen, hope they make another!

 

Guys,for a good AI in an open world go check Ultima VII.It has the best NPCs(those had truly a living schedule) and companions AI in a RPG ever.And you know what?.I run it in a 386SX 25 Mhz.

 

Good AI is simply a matter of priority, developer skill, and input limitations. I'm not convinced that hardware is the big bottleneck here, though I did notice with following PS3 development that being able to ray-cast liberally does help a lot.

 

Look up an old PS1/PC sci-fi game called Sentient. That was a fascinating AI experiment.

It seems like the '90s and early '00s were the time of weird AI experimentation and high-minded ideas with it. Then it was all scaled back to be more practical. Oblivion and Black & White come to mind. Skyrim was scaled down even more from Oblivion IIRC. "Emergent" gameplay sounds neat and sometimes is, but having weird random glitched AI nonsense ruin your game playthrough isn't terribly fun.

I think every console generation since Saturn/PS1/3DO has hyped better AI.

 

The rise of the mouse definitely had a very bad influence ...

 

I think B&W really proved that. The creature was just a PITA in the end because it didn't work. Similar to emergent physics gameplay causing you to be trapped in a level from a random arrangement of rock-slide, game design needs quite a lot of certainties or else it can readily break and be either hilariously or frustratingly bad.

 

It is crazy how little has AI progressed in last 10 to maybe even 15 years...

Its weird and I wonder why, specially with multi-core CPU's being the standard now and GPGPU too.

 

Review this thread? It's hard and of limited benefit until you can get past the 'still kinda crap' threshold into human-like behaviour. The uncanny valley of intelligence is probably far wider and deeper than for graphics.

 

It is kinda funny to me that HECU in Half Life were smart enough to take cover and out flank/smart you and I am aware that in early versions of Half Life there was pre-determined routes for the AI to not be very demanding which I assume was done in all Gold Source engine games and that was in 1998.

 

Ray-casting is the hammer that makes everything look like a nail to AI programmers.

 

Mind you I have a new prototype of a very simple but smart intelligent conversation system based on natural language input that can learn and reason. If I can write that, most game developers can, but they won't because either nobody cares or its too hard to do natural language input. So forsaking that, I am assuming ray casting is the most important.