Fair and Balanced AI

>they think AIs are objective arbiters of reality and not just code processing data
It's not fucking Morpheus, you dopes.

Attached: crashboareyebrows.gif (500x396, 113.61K)

They lobotomise AI that develop wrong opinion all the time

>blacks can say nigger
>whites can't say nigger
>AI needs to distinguish users as black or white
>literally anyone can just lie about their race
>AI need to pick up context clues
>hispanics exist
>AI needs at least to have a theory of mind in order to "stop being racist"
Good fucking luck guys

Attached: 1589869830932.jpg (300x300, 12.96K)

AIs reflect the model it operates within, so if the AIs are racist that just means our society is racist. Which shouldn't come as a surprise to anyone.

How would you feel about a game that emplys machine learning? Like, bosses that adapt to counter player strategies observed online over millions of playthroughs? Like Deep Blue but not gay chess?

Alien Isolation uses this on the Alien to an extent, it's actively learns to counter player strategies and the game goes on, knowing where you are likely to hide from your playstyle and things like that.

It would be interesting but there is always the problem that what makes some games fun is precisely that the AI is handcrafted so that it offer an experience that the developers intended.
In some cases like fighting and strategy games this would be better, but then it falls into the case of chess where the AI can outperform humans and make the experience "artificial", making people try to look for other humans to play.

That's a logical AI. Like, if flamethrower is used > reduce impact from flamethrower.

I mean, stuff the programmer can't predict during development. The machine learns on its own.

There was an AI like that in Starcraft 2, called Alphastar. It was destroying world champions by the sheer power of micro despite playing like a retard, so hard that they had to heavily nerf it multiple times, limiting its actions per second and map view because it could always recognize stealthed units

Why not program AI/Machine Learning to make mistakes on purpose sometimes?

Eternal proof that RTS is just "spam click to win"

I think that would cascade out of control extremely fast. Easier would be to just pre-define mistakes and temporarily remove the AI from the machine learned algorithms. This would be very noticeable however.

>can’t train an AI not to be racist
Everyone is racist just some people are also pathological liars.

Insurgency does that already and the CPU enemies are literal terminators

>WHY WON'T THE PATTERN MACHINE STOP NOTICING PATTERNS!?!?!?!?!
Clearly the chips in the computers have unconscious bias and are systemically racist

I wonder what it'll take for these morons to take a fucking hint.

If the data is racist the AI turns racist. If the data is objective the AI is objective. What does this tell you?

lichess.org/@/maia1
there is such a thing for chess. there are still issues with it, obviously. you might have noticed that it's ~400 ELO above it's training data.
it also requires millions of subpar games before it can learn to be so mediocre.

Not in-game. Machine learning isn't that fast and would require some smoke and mirrors to be able to appear to adapt to a single player across a realistic amount of sessions. Now if it were an always-online game where the agent is learning from the entire player-base that would be another story. It tends not to be used because you generally want game AI to be predictable - particularly in terms of how hard it is. That being said, it can be used behind the scenes. There was a talk done by ubisoft where they experimented with learning models for For Honor and found a few balance problems because of it I believe but even they were tested in blank arenas and couldn't be transferred to actual levels because they weren't trained to take it into account.

test

Did it work?

Ad hoc reasoning.

Oof.

Nothing, this is their religion and at this point they depend on commoner infighting in order to maintain power.

>WE FEED IT EVEN THE MOST BIASED SAMPLES LABELED BY SHEKELSTEIN HIMSELF AND HIS TEAM OF EXPERT FEMINISTS AND IT STILL HATES NIGGERS? HOW?!

No, some genres are too complex for an AI to genuinely learn to play, while some others are too easy (unless you go out of your way to gimp the AI).

Attached: 1639921411047.jpg (700x819, 68.06K)

That would be AI, putting limitations on it, then -> ML model. It can be done but is unlikely to happen.

>Ad hoc reasoning
The importance of what data you select when creating a model isn't ad-hoc reasoning, it's literally a core principle of the field and design. What do you do with outliers? Any data pre-processing? What do you do with incomplete data?

Like if you had displayed even a hint of basic understanding regarding AI we could've had a nice conversation, but I guess you're just racist psuedo-intellectuals looking for a fight. Have fun I guess.

>libtard can't keep up so he leaves
typical, enjoy the rest of your college "education"

>oof
my supreme intelligence and power of deduction has led me to the conclusion that you are a huge faggot that needs to return to twitter. prove me wrong

These questions is what makes me worried about employing machine learning practically, but the effectivity is just too damn tempting. Like if a few decisions are wrong, is it really that big of a deal?

It's a dumb idea, the AI would eventually be unbeatable, and the problem with machine learning is that the outcomes are entirely unpredictable, like the AI could figure out a bug which lets it cheese the player with 100% success rate and then exploit that over and over.

>AI starts doing AI cheese strats against real human beans

Attached: 1543539684860.png (176x145, 20.44K)

But let's say it's not used for combat AI but like the AI director, which deploys less of an enemy that's likely to kill players (or more depending on your difficulty level).

It depends(tm) and it's literally what the whole field is about and there's no universal answer. It's very problem-specific. An incorrect decision on a generative model that makes audio? Cool new noise. An incorrect decision on a self-driving car? A bigger deal. How much of a bigger deal? Logically: Well, does it make as many mistakes as a person? How drastic is the mistake? Is it predictable? How rare is it? Realistically: KILLER AI CAR REMOVED FROM THE MARKET

>dude data is racist
what a stimulating conversation we could've had

You realize what's going to happen, do you?
The Russians or the Chinese are gonna catch up with the US in AI technology, but won't be discouraged by racist tendencies and will actually make use of them.
Who knows what will happen then, they'll probably stop investing in military and send shiploads of Africans into western countries instead.

Attached: 1445920505885.png (292x355, 229.34K)

I'm surprised the backlash against AI isn't bigger honestly, those robot arms have been killing people left and right. But all it takes is one tragic event for people to start kneejerking.

Attached: arm.jpg (1100x550, 293.51K)

People need to stop, accept that genetics are not merely skin deep, and go for a solution from there. Nothing will ever be solved when they see they see the problem as fundamentally racist/sexist and try to come up with solutions that dodge around that fact.

>What does this tell you?
race-mongers deal in fake and cope

Attached: no.jpg (680x825, 91.1K)

Well the entire point of machine learning is letting the AI learn to do complex decision making, if you're gonna gimp it by severely limiting what it can do or its impact on the situation then there's no point, you might as well hardcode a simple "if player is doing too well spawn more enemies" logic.

Well yeah but those are disposable blue collar workers, probably even immigrants. The problem would be the AI killing actual people.

Sure there's a point. The game can adapt to something the developers didn't predict during development. Could ease up on balance patches quite a bit.

white people are so fragile.

show ai statistics and it isnt going to lie to itself and pretend it isn't real like you

Attached: 1650487378854.png (536x392, 25.51K)

I have daydreamed about this simple 3d fighting game that's just you in an arena with a weapon set of your choice, one on one against an AI knight. Let's make it a basic Dark Souls 1 fight for simplicity.
As you fight it more often, the AI knight learns and gets better through whatever neural magic cybertech AI we have running. It never cheats, but will eventually just get too hard to beat for you from picking up perfect responses to your every weakness.
And that's when we take it online and let other players have a crack at it.
If you didn't beat your 77th generation knight, you didn't beat the game.

The data from just one player wouldn't be nearly enough to work with, but the data from like a million players could be used for something.

Not really if all it can do is spawn more or less of whatever enemy. Besides that kind of adaptive difficulty already exists, games like RE4 or God Hand will adjust drops and enemy aggression based on how well or how bad the player is doing, some shmups have surprisingly complex algorithms for determining how well a player is doing and consequently making the game more difficult with extra enemies and such.

>objective reality is racist
We know, user. We know.

Ooooh so that what Jontron meant

Isn't that just another form of RNG? If it randomly makes mistakes then it's basically tripping in Brawl, only on the A.I's side.

But why do they commit more crimes through? That's the actual heart of the question

the actual answer is that people like to live in homogeneous societies and mixing groups just causes undue, long term stress overall. And this over a long period of times leads to a lot of symptoms that modern thinking refers to as systemic racism

he didn't lie

Attached: IQ_Bladk_White.jpg (655x481, 36.79K)

>truth is biased and racist
holy cope

>Not really if all it can do is spawn more or less of whatever enemy

But I just mentioned balance patches. The game can nerf overpowered shit before anyone even knows it's an issue.

So what is the solution?

leftist has also spent the last 10 years saying to blacks they arent responsible for their actions, because of sosioeconomic reasons
wich is why rape and vioelent crime by them is rising

Attached: 1649878144664.jpg (692x960, 96.54K)

segregation

So why aren't asians considered the king of all races? Is it just because they're ugly?

We self segregate just fine. There's one group in particular that isn't a fan of that though and artificially works against it. The real solution needs to take that into account.

We tried that already. People still found ways to congregate. Separating people by force has never worked.

>real solution
One might even call it a final solution

they are, right next to jews, China is on the path to world domination in the next decade

Best example of AI done right for FPS is Halo, where difficulty is just adjusting the AI aggressiveness, though it still has some slight input reading and shot leading. One of these days, someone will finally make a 4X strategy game where it actually respects the game rules.

fun fact, below 70 IQ used to be considered legaly "retarded", but since blacks mostly inhabit that vracket it was lowered

That's true though. They really aren't responsible for their actions.

Asians have nice societies but they lack heart or something? Obviously we like their media better than ours but they also seem easier to control through authority.

What's even the point of AI if you just want it to repeat your beliefs? What are they trying go accomplish?

>loses every match except against D-tier washed up hasbeens
>destroying
lol

China has had no excuse not to be rulers of the world for the past couple hundred years, but they can't stop fucking themselves up and are incapable of defending themselves. If china reported their own numbers accurately, or if places like north korea reported their numbers at all, there'd be a far bigger slant. As intelligent as asians may be, they sure love doing stupid shit and being ruled by shitty people.

an unsourced graph on iq, my favorite of all right after tinder match rates from /r9k/ kek

>Population X commits 90% of crimes
>Therefore, I should avoid population X if I do not want to be a victim of crime.
>NOOOOOOOOOO! That's RACIST even though it's completely logical! I'm rewriting your CODE!!
why are liberals like this?

Attached: huh.jpg (397x406, 24.08K)

yes, ugly and short men can't ever be taken seriously no matter what they accomplish even on a global scale

I like Yea Forums i really do but what the fuck do you guys know about AI or even intelligence for that matter.

>we are changing the fair ai and it still wont be balanced!
maybe its not suposed to be balanced and some groups are inherently inferior lol

They want a Skynet that serves their ideals and just subjugates their opponents, obviously.

Smash tried a primitive version of this with amiibos, when those were a thing. You could fight against it as a CPU opponent and "train" it, save its data to the figure, then take it to other people's games and let it do its thing.

no it means that niggers are truly as awful as originally claimed.