Videogames Killed the TV Star

[Werthead]

34 minutes ago, Fez said:

Games are fully capable of having better AI (well, action-y games at least; strategy games are maybe a different story) but there's been a ton of audience testing showing that most players don't actually want better AI. They say they do, but if they encounter it they just get frustrated quickly. It's a shame, but there it is. So the big studios don't invest in it. I wish more of them invested more in story and writing, but that's harder to show off in trailers than particle effects and realistic lighting.

As for new experiences, they do exist, but they're buried under the overwhelming quantity of other stuff coming out every single day. Most AAA games have settled into the same handful of templates that they use over and over again, but the indie scene absolutely puts out new experiences. You just have to avoid the mountain of metroidvanias, '90s-style shooters, and crafting games.

Yup. There are games where they've done interesting things with the AI, but the problem is that it makes it far too tough to beat. In Age of Empires II on the toughest difficulty level, the AI can launch 3-4 simultaneous attacks on you with the same faction, micro-managing units with precision that's impossible for a human to match, but obviously that's unfair as even the best human can't fight three battles simultaneously, so they have various ways of easing the challenge down.

I think there's also the issue that most gamers - meaning casual people who only buy 3 games a year, if that - want games that are relaxing power fantasies to chill with after work, not gruelling difficulty walls you have to study hard and practice to overcome. It's why multiplayer is such a thing, because everyone is initially on a level playing field.

"Hard but fair AI" is the absolute Holy Grail for modern AI programming and it's so hard to achieve and so easy to fuck up, developers don't really bother with it. It's why the Dark Souls trilogy and Bloodborne are so highly regarded, because it's something FROM seem to have mastered when almost no one else can.

[Darryk]

2 hours ago, Centrist Simon Steele said:

I feel like games have hit a wall. Big companies all do the same thing and do nothing to really push gaming into new directions (what about NPC AI for example?), and smaller companies are providing us fun retro experiences, but nothing really new is happening there.

Well personally I'm never going to recover from Bioware and Blizzard becoming soulless husks, they were my two favorite developers by far.

[Proudfeet]

25 minutes ago, Werthead said:

Yup. There are games where they've done interesting things with the AI, but the problem is that it makes it far too tough to beat. In Age of Empires II on the toughest difficulty level, the AI can launch 3-4 simultaneous attacks on you with the same faction, micro-managing units with precision that's impossible for a human to match, but obviously that's unfair as even the best human can't fight three battles simultaneously, so they have various ways of easing the challenge down.

I don't think that is smarter AI, just faster and tougher. 

As far as I can tell, at least for AoE II, the Hard difficulty is capped at 80 villagers and unless you fall very far behind, only attack when they reach Castle age. For Harder and Hardest, they cap at around 130 villagers and will attack you earlier. That's it. Maybe they are more responsive, but the AI is not any smarter. The same tactic of attacking them early, baiting them into chasing you and then into your reinforcements to get wiped, and using that advantage to snowball the game works all the time and the same through the difficulties.

Smarter AI would be them actually making tactical decisions or at the very least having the sense to realise that the villagers are stuck and moving out of the way instead of eternally bumping into each other, or at least putting them idle if they haven't moved on the map despite being in motion.

[Kalbear]

15 minutes ago, Proudfeet said:

I don't think that is smarter AI, just faster and tougher. 

No, it's really smarter AI. We can make AI that knows how to hit you perfectly in shooters while avoiding 99% of your shots. We can make AI that can easily manage all of its resource gathering AND multiple attacks AND be preparing for counterattacks. We can make AI in things like DOTA that would beat expert players every time because it can do so many more actions per second and can look at all the board at once - even with just the board that it knows about. I know we can do this because we had this capability 20 years ago. Heck, I wrote AI that played FreeCiv, and it was very clear that even as a grad project it could easily crush people in certain ways - though holy crap did we never get boating right.

This isn't faster and tougher - this is much smarter. The challenge is not to make AI that is really good at playing games (or more accurately, playing games as a character in a game) - it's to make AI that is slightly better than basic scripted garbage. 

And if you mess it up? Your game is unplayable by most human standards. So people err rightly on the side of power fantasies come to life, because the alternative of making AI that crushes you is just not fun. 

15 minutes ago, Proudfeet said:

Smarter AI would be them actually making tactical decisions or at the very least having the sense to realise that the villagers are stuck and moving out of the way instead of eternally bumping into each other, or at least putting them idle if they haven't moved on the map despite being in motion.

That would be a nice, incremental upgrade that they probably can do. But it's also really easy to mess up doing tactical AI so that it makes it incredibly stupid early on.

[Proudfeet]

23 minutes ago, Kalsandra said:

No, it's really smarter AI. We can make AI that knows how to hit you perfectly in shooters while avoiding 99% of your shots. We can make AI that can easily manage all of its resource gathering AND multiple attacks AND be preparing for counterattacks. We can make AI in things like DOTA that would beat expert players every time because it can do so many more actions per second and can look at all the board at once - even with just the board that it knows about. I know we can do this because we had this capability 20 years ago. Heck, I wrote AI that played FreeCiv, and it was very clear that even as a grad project it could easily crush people in certain ways - though holy crap did we never get boating right.

This isn't faster and tougher - this is much smarter. The challenge is not to make AI that is really good at playing games (or more accurately, playing games as a character in a game) - it's to make AI that is slightly better than basic scripted garbage. 

Bolded looks faster and tougher to me.  Maybe if you limit them to the same APM and fog of war that humans do?

For the record, I don't think the AI of AoE II uses exploits fog of war. They are always much easier when you kill the starting scout.

26 minutes ago, Kalsandra said:

That would be a nice, incremental upgrade that they probably can do. But it's also really easy to mess up doing tactical AI so that it makes it incredibly stupid early on.

I mean, villagers bumping into each other and getting stuck is an obvious problem. That they haven't solved it yet suggests it might be harder than it seems? I have no idea, but I think that it is an actual example of being smart. Not just being able to do more, but to do it correctly.

[Fez]

The genre I wish had better AI, and I'm surprised players don't want there either, is turn-based tactics games. Because there it's not a question of reaction speed, it's entirely about making the right decision among a very limited number of variables. But inevitably you get enemies that only do basic attacks, or move out of cover in really dumb ways, or focus on my tankier characters even before attacks of opportunity/zones of control come into play.

[horangi]

10 minutes ago, Fez said:

The genre I wish had better AI, and I'm surprised players don't want there either, is turn-based tactics games. Because there it's not a question of reaction speed, it's entirely about making the right decision among a very limited number of variables. But inevitably you get enemies that only do basic attacks, or move out of cover in really dumb ways, or focus on my tankier characters even before attacks of opportunity/zones of control come into play.

I was going to comment on this exactly- its turn based games where the tactical/strategic AI shines (since its not about just the incredible multi-tasking capabilities of a computer).   I wish Civ6 had better water based AI for example (noting Kal's issue with ships!).  However there are some examples like Galactic Civ2 from startdock which had really competent AI.  I remember playing it the first time and being shocked I had lost on a difficulty with no bonuses to the AI.

edit: just to note- I realized after the fact you were more focused on just the tactical games, but I think the 4x types can also shine with better AI.