RPG AI and DX12

[Ripper]

http://www.extremetech.com/gaming/2...how-directx-12-could-change-it-for-the-better

Quite interesting, and a change of pace from "Witcher 3 rulz, Fallout 4 sux, now **** off."

 

[joxer]

Apart from chess simulators, I've seen horrible to decent AI in games.
But memorable AI that I'd use as example everywhere how spectacular it is? Never.

Even the best (but again decent) AI I can remember were actually "cheating" bosses who could pull out powers that by core rules of the game (meaning everything you know about it) shouldn't be possible.

DX12 won't change anything.
AI will still cast dispel or counters randomly instead of when it's the most fruitful/effective moment.

 

[Ripper]

The point is that DX12 will facilitate the use of dedicated AI threads on separate cores, which, as the article explains, could potentially be very helpful.

 

[Moriendor]

Well, there's always been so much speculation in the last couple of decades how this tech or that tech could help take power off the CPU for other things (like great AI) but the end result is we still have pretty shitty AI in games today.

The obvious (IMO) reason is that programming decent AI takes time and resources but is a feature that is very hard to properly advertise. It's also pretty complex so you need people with considerably higher skills than the average gaming programmer, i.e. AI specialists --> specialists = higher pay --> higher financial investment in an area that is difficult to pitch to marketing.

So this is another nice article theorizing about the potential benefits but as long as the suits are calling the shots with regard to the resources and how they are distributed across a project, well, no... nothing will change. And indies will be hesitant, too, because good AI is very time consuming and not really an option on a tight budget.

I remember that Swen Vincke of Larian Studios was a pretty huge AI geek in the days of Divine Divinity but even Larian Studios seem to only aim a little above average AI these days. Tight budget, tight resources, limited personnel, huge asset creation requirements (time & people in art) --> little effort in AI.

And then even if you would pour more resources into AI, it would also increase the QA requirements (more code always means more bugs automatically) so you're opening more than one can of worms with AI improvements and at the end of the day... is it worth it? Will the players really appreciate it? Do the returns justify the investment? Looks like most developers/publishers have found the answer to be 'no' and so they invest in other areas of their game instead.

 

[joxer]

The point is that DX12 will facilitate the use of dedicated AI threads on separate cores, which, as the article explains, could potentially be very helpful.

Erm…
Assassin Creed 4. Recent "modern" mainstream bestselling game.
Buy it on same sale if you don't believe me, recently it's available for some pocketchange somewhere.

Fire it up. Check CPU useage.
Core 1 - 100%
Core 2 - 30 or less %
Core 3 - 0%
Core 4 - 0%

Okay it's not RPG, but still.

DX12 may make things easier. Will developers "exploit" the possibility? I doubt it.

 

[GothicGothicness]

I am actually working quite some with AI innovation in my game... yes it still exists

Will it be successful.. well that remains to be seen

 

[joxer]

LOL
I don't think it's success will depend on AI.

If you want a successful product without millions to invest in make it phonegame without any AI and full of microtransactions - those sold the best in recent years.

 

[GothicGothicness]

LOL
I don't think it's success will depend on AI.

If you want a successful product without millions to invest in make it phonegame without any AI and full of microtransactions - those sold the best in recent years.

Well, my main goal is certainly not to earn as much money as possible.

 

[Ripper]

Well, my crystal ball is on the fritz again, but I think the point of the article is just to highlight a technical improvement that removes one of the barriers. What people do with it - who knows?

 

[joxer]

Well, my main goal is certainly not to earn as much money as possible.

No need to specify that, we all know already you're not EA's employee.

Well, my crystal ball is on the fritz again, but I think the point of the article is just to highlight a technical improvement that removes one of the barriers. What people do with it - who knows?

People will do what they're tasked to do.
It would be awsome if some CEO does bless investing into it, but I'm kinda not optimistic.
I mentioned EA? Sims 4 still has horrible pathfinding problems, even after they dropped openworld design thinking that'll help making things easier. And if anything AI should do without problems, it's pathfinding.

 

[Ripper]

It's also possible that by farming out the AI to its own core, some complexity can be removed from the main thread.

Also, if we are talking about market forces dictating the adoption (or not) of better AI, bear in mind that a great deal of game development is essentially outsourced to middleware. Many games will use a third-party solution for AI, such as Kynapse and PathEngine.

The market logic is surely that the developers will choose the best AI middleware product (which may ultimately garner them better reviews), and so the AI vendors will have a strong incentive to improve their offerings.

 

[joxer]

Now that's an interesting thought there you have. Good point.
You've actually added +1 to my hope something could change.

 

[Couchpotato]

Sorry hate to burst your optimism bubbles guys, but ever time a new tech innovation happens for the betterment of PC gaming the tech is never fully used later on.

The various excuses given why is due to the hardware needed to run it.

 

[Fnord]

And then even if you would pour more resources into AI, it would also increase the QA requirements (more code always means more bugs automatically) so you're opening more than one can of worms with AI improvements and at the end of the day… is it worth it? Will the players really appreciate it? Do the returns justify the investment? Looks like most developers/publishers have found the answer to be 'no' and so they invest in other areas of their game instead.

If you create a good AI foundation, then there is the potential for less scripting to be needed for each level, which in turn means that you'll have to spend less time testing your levels. I suspect that this will in particular be the case for friendly AI, which can result in a real headache when it does not work well. We don't care much if an enemy gets stuck in a door, we're going to kill it anyway (unless the game is thief...), but if a friendly gets stuck in a door, or can't navigate the terrain, then that's ground for a lot more unhappy players. And this is something that reviewers routinely point out is a problem, and something that lowers review scores.

 

[azarhal]

Sorry hate to burst your optimism bubbles guys, but ever time a new tech innovation happens for the betterment of PC gaming the tech is never fully used later on.

The various excuses given why is due to the hardware needed to run it.

The article is not about tech, it's about CPU load and multi-core making AI hard to program because they don't have CPU cycles available easily. DX12 reduce the CPU load of games and has better multi-threading support.

Most DX11 compatible video cards, which is basically any video card released after 2009-2010, are compatible with DX12. All people will need is Windows 10 to get the benefit of it.

 

[Couchpotato]

Yeah but we'll see as they say.

How do I know well in my twenty five years of experience in playing games I have seen many new software, and tech innovations wasted, or never fully implemented.

 

[azarhal]

Yeah but we'll see as they say.

How do I know well in my twenty five years of experience in playing games I have seen many new software, and tech innovations wasted, or never fully implemented.

Well, DX12 is already here and some games are already using it and from what I've seen from developers, they really like it because it's not that different from using DX11.

If they are going to use the freed CPU load because of it is another story, but that has nothing to do with tech advance.

 

[Morrandir]

When it comes to AI I'm really curious about what Warhorse Studios will do in KC as they're working together with some experts from Charles University in Prague. In a published scientific paper they talk about

specific design decisions we have made that let us achieve a level of behavioral fidelity unmatched by existing games.

 

[ChienAboyeur]

An article to sell the idea that what should have happened already is going to happen soon tomorrow.

Graphics are a sore technology based excuse as devs put them aside when it suits them: not every game displays state of the art graphics, yet when is the AI?

The PC scene is about graphics, improvement in graphics is all what the PC scene has.

The road is set: 4k. 4k matters much as it is required for VR. The VR headset market is too small though, the mobile cell phone market will have to take charge of it.
The thing is that on that type of screen, only a few humans in the world might be able to see a difference in the resolution, all the others will buy a product that makes no difference for them but must be if the PC scene feels entitled with VR.

Once VR is set on track, it will be a huge sight of relief for devs as it allows them to reboot and postpone deliveries on AI.

In parallel, there will be panoramic 4k (for the immersion), and later the 8k revolution...

All about graphics...

 

[Zloth]

Do we want RPG AI to improve? I do actually want to win my games in the end.

What excites me more is the possibility for MORE AIs. Give everyone in town a decent AI instead of just a simple script to follow day and night (and, if it's a Witcher game, when it rains). Give all the little critters in the woods around you an actual AI.

Unfortunately, console CPUs aren't real impressive. Even if DX frees up CPU power, there's not that much power there. Still, something is better than nothing.