Call of Duty WWII

[Leisher]

FPS that features you as a fresh recruit from Texas getting ready to see your first action on D-Day.

The action is pretty intense, and the only flaw that bugs me is the Germans will focus on you from time to time. That will hurt your immersion. It got me wondering about game design, but I'll hit that at the end of this...

One of the cool things is that the type of game you're playing changes quite often. You'll go from FPS to stealth game to driving game and so on. That's kind of neat. There's only one section that sucks, the airplanes. They just don't fly well. Thankfully, it's really short.

The game is very linear. That can be good and bad. In this case it works because it's very story heavy, and it's a good story.

The game isn't super long with only...maybe 10 levels? However, some levels can take a bit to get through if you're actually playing it and not doing a speed run.

For <$5 it'd be a good purchase. Anything more and it wouldn't.


Ok, discussing AI...In war epics like this one where you're part of a squad or platoon, how hard would it be to train the AI to only shoot at you IF it notices you or on a more random basis? Point being, if you've got 20 guys alongside you, shouldn't there be a realistic chance that you go "unnoticed" or simply don't draw fire if you're keeping your head down? I'd like to be able to get some shots off without being targeted by half the German army.

Would that be difficult to program?

[thibodeaux]

random = rand(1,20)
bool shootPlayer = (random == 1)

Done

[thibodeaux]

And I'm not even joking: all modern "AI" is based on ML, which is (in theory) about modeling things with probability distributions.

Also, the SJW bedwetters get mad when the AI comes up with a probability distribution that isn't COMPLETELY UNBIASED BIGOT!

[Leisher]

random = rand(1,20)
bool shootPlayer = (random == 1)

Done

I'm sure it's not that easy.

We'd want them to act realistically. So, a Nazi under fire would obviously attempt to shoot the people shooting at him...IF he knows where they're shooting from. A Nazi sniper might take shots of opportunity or the biggest threat to his allies. Point being, in a massive firefight, I don't think the AI should be focusing on the player at all times or impossibly spot him among a group of allies.

I mean, in real life, I'm certain a lot of people were successful or survived because they simply were never noticed. In games, that can NEVER happen due to how the AI is coded. (Obviously, stealth games are the exception.)

And I'm not even joking: all modern "AI" is based on ML, which is (in theory) about modeling things with probability distributions.

Also, the SJW bedwetters get mad when the AI comes up with a probability distribution that isn't COMPLETELY UNBIASED BIGOT!

'Splain to my non-programmer ass. What's ML? I get the probability distributions and how monumentally stupid people are who think that makes it racist.

If you're making AI, probability seems like the most obvious way to make it act, and probability being based on real life statistics seems like it would be as accurate as possible.

[thibodeaux]

ML is "machine learning," which is what most people mean when they say "AI" now. Although probably video game enthusiasts actually don't, since they actually have a continuous tradition of rules-based AI, tree-search, path-planning etc.

Whereas everybody else's vision of AI is very much based on the hype-cycle of impressive achievements in multi-layer neural networks. I.e., deep learning or ML.

[TheCatt]

ML is just fancy correlation

[thibodeaux]

Linear regression is like correlation.

But I like to explain ML to programmers as: it's a function. And instead of you writing code, you give it several examples of (input, output).