Hmm, what do you think?

Hmm, what do you think?

Attached: ML_Game_Development.png (800x800, 44K)

Other urls found in this thread:

thiswaifudoesnotexist.net/
blog.openai.com/better-language-models/
twitter.com/NSFWRedditImage

Oh wow potentially infinite worlds, I bet a survival crafting game would work really well in one of those!

>potential to be unlimited in size
Yeah because fuck causality right? Generating an infinite world on the fly is easy, I wanna see how you're going to store its infinite states when actors start interacting with it and modifying it.

I was playing AC Origins, with it's endless Ubisoft map barf of question marks, and thought about how I'd make a ML code to solve the traveling salesman problem and walk from place to place for me.

Attached: TravelingSalesmanProblem_1000.gif (507x268, 4K)

fuck every single poseur trying to sell common procedural algorithms as some kind of breakthrough in AI

also these are all just cheap, shitty alternatives to hand-tuning by a human developer. these companies are basically trying to steal other people's jobs while resulting in inferior products for the consumer

>Textures and objects will render dynamically as you get closer

This.
This is a sign that this was made by a retard; someone who just sorta tosses out ideas and screams IT'S GONNA HAPPEN; someone who knows nothing of actual current development and likely is seething about his inability to participate in it.

What kinda nigga doesn't know what fucking LODs are in 2019.

how about they develop an ai to make games fun again

here's your game bro

Attached: nms.jpg (259x194, 11K)

When will Machine Learning fix steam's recommendations?

Attached: 1539403594609.png (1049x233, 265K)

A number ML techniques out perform other methods.

Rising storm is not that bad.

>see artists making millions on Patreon
>decide to learn how to draw
>suddenly see /g/ talking about AI spitting out tons of random waifus
>decide to stop learning how to draw

Attached: 1550430384357.jpg (1024x576, 83K)

If implemented correctly you could have a Rune Factory game with unlimited dialogues. Too bad it will never happen.
thiswaifudoesnotexist.net/

NLP models have gotten pretty good at generating text.
Issue is how you could link text to behavior of NPC or whatever.
You could generate loads of random books for TES I guess.

A lot of it is just hyping ML too much and shoving it in everything, but AI would make procedural generated worlds work a lot better, and there are a lot of things it would enhance especially when it comes to things like physic engines.

Or even just AI companions could vastly improve with ML and not in the sense of "omg this AI is totally real" but by determining how useful the companion is and how much damage its taking it could change its actions to better fit the individual player's play style.

How would AI improve physics engines?

>vague noncommittal statement
this is exactly the kind of dialogue i'd expect from an ML bot

Look at image recognition type stuff.
ML methods BTFO other methods. You can hand tune some algorithm to try and extract features or use CNN type thing that will do it faster and better.

meant graphics, but i suppose it could maybe offload some of the processing done with the physics engine just with prediction analysis. not sure if that would actually decrease the overall workload though.

Just imagine it. You'd never be able to make it to the center of the universe.

I'd rather explore a well crafted small area than a giant unlimited random world filled with random shit. Give me Deus Ex or Arcane Studios games.

when you start using the ignore feature. only then it starts learning your preferences.

I WANT TO FUCK NARMAYA SOOOOOOOO BAAAAAAD

AI creating higher res backgrounds of old prerendered backgrounds has me the most excited. Some shit still looks bad, but considering how new the technology is and how people are still tinkering, this will be the best way to preserve and play these older games.

Attached: vbo0Emm[1].png (2048x960, 2.49M)

Actually working with ML. As in, paid for it.

NPCs:
It could work and help a lot. But devs who don't give a fuck about AI won't suddenly start making the significant changes required to have ML AI in the game.

Modelling systems:
Not in this decade. It's doable, but it's purely academic at this point.

ML rendering:
Fuck no, not any time soon. Could produce new textures statically, though.

NPC dialogue:
Fuck no, twice. Current NLP is fragile and only works in specific environment.
Plus, AAA jews won't want dialogue without voices or synthesized speech.
Maybe in 2 decades or in hipster games.

Procedural generation:
ML is better at procedural generation than most devs already. But the user will only see the less shitty open world, not the generation process.

Mobileshit:
Utter gimmicks yeah. Actual game mechanics unlikely.

tl;dr When it comes to vidya, ML is AI first and foremost. And it will stay that way for a while.

Attached: 1532255221625168924.jpg (525x517, 21K)

It would have benefits for sure but it would still require a ton of work to craft an interesting world.
What would obviously happen is extreme repetitivity and doing the same shit over and over with very little differences.
Seems like people think throwing more processing power at the problem will solve everything.

>Fuck no, twice. Current NLP is fragile and only works in specific environment.
blog.openai.com/better-language-models/

>ML is AI first and foremost

>implying its not used for large scale data analytics to better market games

you fail to realize those waifus will never have their naughty bits exposed, that's why people make so much money from patereon

This will bring my waifu to life

Check out the results. They're getting 70% on tasks that humans get 95% on.

It's gonna be years until it's useful enough to have as a chatbot, then years more before it's mature enough for vidya, then years before devs realize that and convince the suits.

>AIs start getting extremely politically incorrect within a day
Haven't they learned?

>decided to finally put my old talent and odd likes to use
>get a drawing tablet
>spend the last couple years getting use to the difference between it and traditional art
>mostly practice by anonymous drawthreads
>finally feel comfortable enough to start doing smut with a name 2 months ago
>I have one patron and make $4 a month after fees

Its a start and I greatly appreciate that person for their support, but fuck this is a lot to manage.

its fun but incredibly frustrating at the same time...