Cyberpunk 2077 Pre-Release Thread [GAME RELEASED, GO TO NEW THREAD]

[Harthwain]

There's no money in good AI for gaming. And if you're a good AI engineer you work on things like Terminator-vision augmented reality programs and retire at 35-40. So we're left with FEAR and Half-Life as unsurpassed stand-out AI from centuries ago.

Good AI is very game-dependent and hard to get right.

Developers of Plague Tale: Innocence shared some interesting bits about how their swarms of rats worked. Interesting stuff: link.

 

[DalekFlay]

View attachment 13879

"Lord knows we've seen Bethesda struggle with melee combat, the main mechanic of their fucking game," says video game expert who probably gave Skyrim a 10 out of 10.

 

[Sarkile]

Shouldn't they be doing little beyond bug fixing at this stage? After such a long period of development?

 

[Harthwain]

Shouldn't they be doing little beyond bug fixing at this stage? After such a long period of development?

Bug fixing will be done after the release.

 

[Wesp5]

It's not a unsolvable problem to create a "perfect AI" but who would've liked to play against it?

Indeed! I remember the original Unreal Tournament where the final enemy began cheating after a time, like dodging the moment you pressed fire...

 

[karoliner]

Yeah it was a real asshole.

 

[Harthwain]

It's not a unsolvable problem to create a "perfect AI" but who would've liked to play against it?

Indeed! I remember the original Unreal Tournament where the final enemy began cheating after a time, like dodging the moment you pressed fire...

This is why AI shouldn't cheat. Instead it should stick to the rules, same as the player. That's the difference between good and bad AI.

 

[curry]

There's no money in good AI for gaming. And if you're a good AI engineer you work on things like Terminator-vision augmented reality programs and retire at 35-40. So we're left with FEAR and Half-Life as unsurpassed stand-out AI from centuries ago.

Rain World says hi. I'd say it has on of the most impressive AI in gaming

Ever heard of Dota 2 OpenAI? It was able to beat competitive teams. I played 20 games or so against an OpenAI team and didn't win a single game.

 

[Naraya]

There's no money in good AI for gaming. And if you're a good AI engineer you work on things like Terminator-vision augmented reality programs and retire at 35-40. So we're left with FEAR and Half-Life as unsurpassed stand-out AI from centuries ago.

Rain World says hi. I'd say it has on of the most impressive AI in gaming

Ever heard of Dota 2 OpenAI? It was able to beat competitive teams. I played 20 games or so against an OpenAI team and didn't win a single game.

Yes, I've heard about it. And watching it play was boring because it was so perfect.

 

[Riddler]

There's no money in good AI for gaming. And if you're a good AI engineer you work on things like Terminator-vision augmented reality programs and retire at 35-40. So we're left with FEAR and Half-Life as unsurpassed stand-out AI from centuries ago.

Rain World says hi. I'd say it has on of the most impressive AI in gaming

Ever heard of Dota 2 OpenAI? It was able to beat competitive teams. I played 20 games or so against an OpenAI team and didn't win a single game.

That was almost entirely due to them changing the rules to favor the AI and not letting the players practice using those rules. After it had been released for a week 4k teams easily beat it. Still very impressive though.

 

[DJOGamer PT]

AI in game depends on the kind of experience the game is trying to be.
Not every game benefits from a clever and dynamic AI
They should be crafted to suit that experience not to advance the actual science of making AI

So we're left with FEAR and Half-Life as unsurpassed stand-out AI from centuries ago.

FEAR's AI while effiecient and impressive at first sight, are heavily scrited to the level they appear in.
They are not like for example Metal Gear Solid's soldiers that can mantain a constant level of inteligence no matter what enviroment they are placed in.

 

[DalekFlay]

"Good AI" is always about the perception of intelligence and tailoring it to a good gameplay experience for the player. It's almost never about making it as smart or aware as it could be.

 

[Grotesque]

It's not a unsolvable problem to create a "perfect AI" but who would've liked to play against it?

Indeed! I remember the original Unreal Tournament where the final enemy began cheating after a time, like dodging the moment you pressed fire...

This is why AI shouldn't cheat. Instead it should stick to the rules, same as the player. That's the difference between good and bad AI.

A good AI is the AI "that is smart enough to be stupid".

"Making perfect AI is easy. Making fun AI is really hard"

"it would be fairly easy to make AI that could beat the player "

Anyone recognize whose word belong to?

Answer below:

Spoiler

https://youtu.be/aTVBSxnwiCo?t=1151

 

[compvet24]

Anyone recognize whose word belong to?

Answer below:

Spoiler

https://youtu.be/aTVBSxnwiCo?t=1151

It is really insane comparing Developers today to these legends. Fellas these days just don't match up at all in terms of creativity, effort and thought process.

 

[Zer0wing]

It is really insane comparing Developers today to these legends. Fellas these days just don't match up at all in terms of creativity, effort and thought process.

It's only indie market now where one has a chance to find creativity and effort.

 

[Ol' Willy]

So we're left with FEAR and Half-Life as unsurpassed stand-out AI from centuries ago.

STALKER AI wants to say hello. And, unlike previous examples, it works best in open environments.

 

[Hace El Oso]

STALKER AI wants to say hello. And, unlike previous examples, it works best in open environments.

FEAR and Half-Life weren’t meant to be an exhaustive list of all games that stand out for their AI, but the fact is that gaming AI on the level of FEAR(or Stalker, another good example) stands out a mile(the FEAR creator wrote a paper published by MIT on his AI work, for example) and has not generally been surpassed. It is noteworthy that there is seemingly no serious effort to do so, either.

My personal experience with schoolmates who went on to become serious AI engineers combined with what I see and read has led me to conclude that there is comparatively no perceived market(money, payoff) for more developed gaming AI. You get some one-offs, sure.

 

[Ol' Willy]

but the fact is that gaming AI on the level of FEAR(or Stalker, another good example) stands out a mile(the FEAR creator wrote a paper published by MIT on his AI work, for example) and has not generally been surpassed. It is noteworthy that there is seemingly no serious effort to do so, either.

I was saying the exact same thing in this very thread, so I see what you mean, china.

 

[Whisper]

Why they dont make good AI?

 

[Grotesque]

Why they dont make good AI?

because it's hard and the payoff is small

 

[Pegultagol]

Isn't there some AI middleware already that would solve a lot of this, at least pathfinding and generating immersive NPC chatter?

 

[Lambach]

View attachment 13879

"how do you do melee in a first-person perspective?"

Find the guys who worked on Dark Messiah of Might and Magic and ask them. Best first-person melee combat to date, even if the game's other aspects were meh.

 

[Grotesque]

View attachment 13879

"how do you do melee in a first-person perspective?"

Find the guys who worked on Dark Messiah of Might and Magic and ask them. Best first-person melee combat to date, even if the game's other aspects were meh.

It's overrated.



Look in the video at the timestamped instance where the player slashes that mage three times.

If Cyberpunk came with that footage, everybody would cry what shit the melee combat is.

 

[Lambach]

It's overrated.

It ain't perfect, but I can't think of a game with better first-person melee combat.

 

[just]

It's overrated.

It ain't perfect, but I can't think of a game with better first-person melee combat.

every japanese porn sim https://files.catbox.moe/7cb585.webm