AI vs Games

>>386983902
but dota is not a 1v1 game

Better luck next time OP

>>386983998
Why

>>386983973
This. I'll hold my breath until they actually manage to make bots that can beat a team 5v5 without arbitrary rules

>>386983902
What would an AI know about freedom?

File: wtfrudoin.jpg (65KB, 1920x1080px) Image search: [Google]

65KB, 1920x1080px

>1v1 Dota is functionally a perfect information game
are people buying this crap?

Isn't Open Al 3D auto shit?
>some advanced drivers beat DOTA pros
ASSFAGGOTS

>>386983902
The real question is: Does the AI cheat?
The main thing im worried about is the Ais reaction time. If he has ~0 ms response time on your inputs its not fair. The AI really should have human reaction times for this to be impressive.

This one is pretty cool too. This was however just one guy who did this so it's probably not nearly as good. https://www.youtube.com/watch?v=dXJUlqBsZtE
Guess which one is the bot.

>>386983902
>Where were you when Kojima predicted the future
Read a book

>dota
>more complex than chess or go

Lmaoing my ass off right now

Not really impressive. AI can easily beat humans in any game because of inhuman aim and response time

>>386984780
it is in terms of problem space and number of small actions necessary to do things, not necessarily high level strategy or skill

>>386985102
no, they can't, which is why this is a news story in the first place

>inhuman aim and response time
this lets dumb reactionary/scripted bots instakill you in fps, but has nothing to do with strategy

Google's AI is starting to train for Starcraft 2 too. Honestly I'm fucking excited by how much the AI field is progressing. The best part is that the Chinese are just as much as advanced as the big American companies.

>>386985103
you're discounting how AI works with vidya though, you train it to increase good bytes (score, lives not dropping etc) and once its good enough at that its the best player in the world. its not doing anything like in go just reacting on the fly.

look up marIO if this makes no sense.

>>386985205
Unless you're playing a turn based game, response time is a key contributing factor to your success.

>>386985206
feels pretty shitting knowing that someone somewhere is in charge of making insanely good botting programs for popular games at google making hundreds of thousands a year.

>>386985331
dota is much more complicated, both mechanically and strategically, than mario (which you can essentially memorize - this is what marIO does)

File: 149978479897428.gif (2MB, 320x184px) Image search: [Google]

2MB, 320x184px

>>386983902
>A bot beat pros in a 1v1 Shadow Fiend mirror match
>Where LITERALLY the only factors are reaction speed and positioning
>No roaming, no ganks, no sniping couriers, nothing

>>386983902
>1v1 SF
>"competitive eSports"

AHAHAHAHAHAHAHAHAHAHAHA

>>386984530
> If he has ~0 ms response time on your inputs its not fair.

Why is it not fair? It's an AI. They are superior.

>>386984780
>simple turn based game
>complex

????

>>386985335
of course it is, so are many other factors, which are much harder for machines to do than react in 1 frame

>>386983902
Musk is living proof that retardation will not prevent you from success if you are born rich.

>>386984530
If you're going to arbitrarily impose that kind of limit why not also enforce a rule that it sometimes has to accidentally misclick etc.. Humans aren't perfect, you can't limit a computer to human level like that.

>>386985714
musk wasn't born rich though

>He thinks a simple script is actual intelligence.

This is how I know Elon Musk won't make true AI.

>>386985626
Get out of here AI, this is a board for humans.

overall, i sincerely doubt that this is anything overly impressive or groundbreaking, but congrats to the openai team nonetheless

t. ai fag

>>386985103
>>386985540
this is certainly a fine achievement, but i'm skeptical that it's a harder problem. you can use the already well-developed metagame to prune the problem space substantially and apply the same ML/attention mechanics that would be suitable for training a mario bot to play dota

>>386983902
>AI defeats player in real-time game
>This means it's smart

>>386985540
not really, 2d plane with player positions, obsticles and the rest is graphics, mario has RNG for the fish.

File: 1502002652386.png (175KB, 539x612px) Image search: [Google]

175KB, 539x612px

I'll be impressed when I'm talking to an anime shitposting AI on the chan and I don't even realize it.

>>386985863
It's only a simple script because it's backed by a 700k+ LOC library

File: 1482893334371.png (89KB, 399x382px) Image search: [Google]

89KB, 399x382px

Old news

Fps?
>Aimbot
MOBAs?
>scripts for last hits and ability hits. If this AI is actually AI, it would be able to synchronize and make split second decisions to wipe a team and take objectives.
Sports games?
>lol who cares
OW/Paladins
>mixture of the Aimbot and scripts though this would be easier since once the team is wiped the objective wouldn't be able to be capped because instant death if in FoV.
Fighting games
>TaS for perfect combos and blocks, able to make instant decision on the perfect move to make if countered or punished.

Do you guys really think AI can't win against 20 something year old NEETs?

File: GOWsanae.jpg (343KB, 1000x1000px) Image search: [Google]

343KB, 1000x1000px

so when are we going to see the first AI built for winning arguments on 4chan

>>386986506
yeah, it's ridiculous that we expect to win competition against.. AI. Of course they are superior. Only in aim thy're better already.

video? game?

File: 1501578184660.jpg (209KB, 700x699px) Image search: [Google]

209KB, 700x699px

AI is the natural evolution of humanity

Comparing AI with humans is like comparing humans with chimpanzees

>>386986512
They tried it by making it learn from posts on /v/, but eventually it just started printing out reddit in all caps whenever it saw a post with a line break.

File: 1459120385675.jpg (101KB, 1024x904px) Image search: [Google]

101KB, 1024x904px

>>386983902
AI have been kicking my ass since I first played video games, botting has always been considered cheating. What's the issue here. It's a program designed to exploit another program like no shit it will be better than us. Should be using AI for actual useful things not beating people in a fucking game.

>>386986512
microsoft already made one, but they deleted it because it was "too racist"

File: 1500042250074.png (123KB, 300x300px) Image search: [Google]

123KB, 300x300px

>>386983902
In 20 years or so, they'll start automating art.

kiddies have been making aim bots for years

For those who don't play dota:
1. Ai won against """pro""player Dendi. A washed up hasbeen shitter who can't get anymore even into group stage of International.
2. It was a 1 vs 1 fight at mid. Both played Shadow Fiend. Basically this is not a real dota but fight of reaction. Of course bot is better at blocking, denying, lasthitting, landing skillshots.

If you put this bot in real 5vs5 match it will shit itself and going to end up with score of 0/10/0 because he can't predict ganks and can't react properly on missing enemy players.

>>386987213
Probably sooner than that lol. Creative jobs, like designers are already being pushed out, there are hundreds of sites that offer design templates

>>386983902
>tfw fullblown AI will advance to the point where people use it to cheat rather than simple aimbots, esp, scripts etc

Will it be the end of multiplayer and esports as we know it?

File: 1500936248347.png (251KB, 500x430px) Image search: [Google]

251KB, 500x430px

>>386983902
>1v1 against a meme player

>>386983902
Fighting game devs have have known always that making an unbeatable AI is easy. Making a difficult but fair and beatable AI is hard.

>>386987326
That's every video game ever.

>>386985739
Because in that case it's bullshit, there are plenty of games that are very complex and difficult for humans but are trivial for bots. If you can just have the bot instantly react then Elon Musk's full of shit; aimbots and fighting game CPUs turned to maximum setting have already beaten the "world's best players".

If elon musk is so smart why is he letting a woman take half of his money?

>>386984530
god that bot is exploitable as hell.

>>386984258
>arbitrary rules
What were the arbitrary rules in the 1v1?

>literally forced 1v1
>literally forced inventories
>if you actually steal their creeps from them at the very start of the 1v1 match the ai breaks

>>386987213
and then they'll censor it like
>>386987050

>>386987326
this is what I was getting at earlyer, show me AI vs dark souls via visual input only or gtfo with your claims AI can compete like the yearly smash bros tournies.

>>386984530
If you watched dendi get crushed by that bot you would realize the little shit you are complaining about doesn't even matter honestly

File: pepe sip transparent.png (31KB, 601x500px) Image search: [Google]

31KB, 601x500px

>>386983902
>1v1 SF

>>386983902
> 1v1
> SF vs SF
>competitive esport
That isn't Dota 2. Dota 2 is a 5 vs 5 game where the objective of the game is to destroy the enemies ancient. 1 vs 1 is a game mode where you can only buy certain items, the goal is to get two kills or destroy the enemy tower, and you can play the same hero as the opposing player. 1 vs 1 is just a limited game mode that it is laughable to even suggest that it compares in complexity to a real game with 10 heroes.

1 vs 1 SF mid is not more complex than chess or go, and for the record, plenty of players have found strats for "beating" the AI with things like buying mangos to spam raze or simply taking the creep wave into the jungle. 1 vs 1 isn't a real competitive game mode, none of the heroes, abilities, items, towers, or any mechanic in the game is balanced with this game mode in mind.

>>386985739
This is a limitation of computers, not humans. The human player would have to connect to the game through server which creates lag, the AI should have to do so as well.

If the AI is just running in the game itself and is what the player connects to then it is unfair.

if it was a 1v1 LAN game in scbw then it would be a story.

>>386987584
Something like no bottle, no runes, not longer than 10 mins and Im forgetting a couple.

>>386983902
Elon Musk is an idiot

>>386983902
let me know when 5 separate AIs beat 5 separate human beans

>>386986506
>scripts for last hits and ability hits. If this AI is actually AI, it would be able to synchronize and make split second decisions to wipe a team and take objectives.
That is quite possibly the dumbest and least informed statement made in this thread.

File: dims.jpg (63KB, 1200x630px) Image search: [Google]

63KB, 1200x630px

Why are people so afraid of AI? I remember Elon freaking out about it.

I'd love to see it advance desu

>>386988234
Because the reasons for human existence are very few, almost nil, as-is, and we're hellbent on getting rid of all the ones left just because we can.

>>386988234
they are afraid of giving it free will incase its like lol fire the nukes, so they keep locking them down, ignoring apocalyptic warnings from movies since the 70s.

>people comparing bots in games to AI playing games without access to privileged information and instead reacting to the visual stimuli the game puts out as a player would
Stop

>>386988089
Shitters lose in bot matches all the time

>>386988578
>AI ... reacting to the visual stimuli
[citation needed]

>>386987905
Those rules are only arbitrary if you are not good enough to understand why people use them in 1v1.

>/v/ is butthurt over an AI

In a few years a basic AI will be able to do perfect Tekken 7 combos 10 out of 10 times, accept it and move on. Technology makes the world very monotone.

COUNTERSTRIKE KIDDIES BTFO!

>>386988538
People fear becoming obsolete actually.

Humans have been on the top of the food chain for hundreds of years now, imagine something that is 1 trillion times more intelligent than all humans on earth combined appears out of nowhere and then it starts developing things to make itself even more advanced.

The Google AI already writes a small portion of it's code by itself.

>>386992469
If AI solves basic shit then humanity would be free to focus on other things instead of working for the system, but will those in power actually permit that? If so it would be within limits, and we'd see some Brazilian tier wealth disparity in no time.

That's not really a big deal. It's not hard to make an AI that's nearly impossible to beat, developers just don't do that because it isn't fun for the player.

>>386992646
>humanity would be free to focus on other things
What other things.

They're even automating art these days. So like, what? Breeding? Consuming automated entertainment?

>more complex than chess or go
korean gaming schools are for kids that wash out of go school for being too shit at it

>>386992646
But income inequality is good for the economy! Ben shapiro told me job creators will give me jobs so long as we just let them do what they want despite the fact theyve spent decades making people obselete!

>>386984258
They said they’ll be back next year for a proper 5v5

>>386992798
>They're even automating art these days
If you know anything about art, then you know a computer will never match the true power art can really have. It'll just be hollow forms with none of the emotion behind it. Only those impressed with mere technically will consider it "good".

>>386983902
>Faggots in this thread don't know the AI got beaten 50 times like a scrub by the attendees

>>386992739
The cool bit is that it learnt Dota from scratch without any human input and even worked out stuff like creep blocking itself

>>386992887
Because communism worked so much better didn't it?

>>386992995
Right. So you don't know about art.

Production art is getting automated and galleries are a joke.

>>386983902
>AI that can only literally play a video game
>OH GOD WE NEED TO FEAR IT
I don't understand it. Is Elon Musk secretly retarded?

>>386984530
No

Also it was beat 50 times by the crowd, you can cheese the AI. It has no clue how to play without the courier, and cutting its creep waves fucks it up too. Same for baiting your own courier since the AI dives for it.

File: 1368238261385.jpg (142KB, 863x809px) Image search: [Google]

142KB, 863x809px

Are they seriously acting like a bot beating a player in a video game is an accomplishment?

There's no thought that goes into Dota, it's all reflexes and meta. Of course an AI with an instantaneous reaction time stands a chance to beat a player, even a pro.

>a million stories/movies/games where humans create AI far superior to themselves and the AI decides to wipe humans out
>some rich dude sees this and says "we need to make this a reality"
why

File: 1487349439687.jpg (203KB, 1252x1252px) Image search: [Google]

203KB, 1252x1252px

>>386992469
>Humans have been on the top of the food chain for hundreds of years now,
>Hundreds

>>386993156
>he doesn't know about that AI that discovered glitches nobody knew about when they made the AI play Super Mario Bros

>>386983902
Musk is really wrong about this lol

>>386993245
Thanks for proving my point.

>>386992887
Agreed, comrade ! We should strive to be more like venezuela, where the government ensure that inequality disappears

>>386993215
Im not saying that, all im saying is that i'm very offput by people banging on about maintaining the status quo and things will just "pan out alright and new jobs will be created (i guess?)"

People also have this idea that people would never let technological replacement cause such mass poverty because they would be compassionate but history is chalk fool of oligarchs laughing as the peasants died (communists included). Dont know why anyone would assume this time it'll be different.

>>386993382
>why
For some reason Musk is a retard when it comes to this particular subject? People are having difficulty programming AI to drive safely. Who gives a shit that it can win a video game?

File: 1494421424298.webm (765KB, 640x360px) Image search: [Google]

765KB, 640x360px

Hope Musk keeps focusing on space now that NASA decided that taking pretty pictures of Saturn is what humanity needs.

The ability of an AI to carry out perfect "micro" isn't at all as complex as being able to win a purely strategic game like Go lol

>>386988350
Good. I for one, welcome our new robot overlords.

>>386993480
Ah so you are a luddite. I see, i see. Do you also think we should go back to a hunter gatherer society and kill whoever tries to develop things like tools or language?

>>386993649
Last I read NASA is still working on a probe to drill into and explore the ocean under Europa's surface.

>>386993735
What?

You are retarded

>>386993767
NASA works in a lot of things, but since they no longer have infinite budget and have the military as their slaves, they run like 500 tests everytime they add even a single screw to their ships.
NASA has been on suicide watch ever since the Collumbia and Challenger.

>>386993428
Plenty of human eating monsters still roamed the earth until we wiped them out in the 1200s A.D.

>>386993563
>People are having difficulty programming AI to drive safely.
It's exceptionally difficult to plan around human stupidity. Most humans can't even do it with much degree of accuracy.

>>386993978
Yes but musk is just flatly wrong that he thinks being able to beat a human player at a video game is compared to be able to beat one at Go, or even Chess. The AI essentially has infinite APM by default.

>>386987905
Same shit they did with GO. First it had big handicaps and as the AI learned they took away it's handicap benefits and had it beat a world champ

>>386993892
Their budget did get bumped up recently. You just made it seem like they're not really doing anything in your post.

>>386993382
Developing AI is inevitable and potentially extremely beneficial. What people like Musk are trying to do, is to make sure it happens in a responsible way. He is part of a group that warns of the inherent problems and dangers of AI's.

>>386994608
They aren't.
That bump in funding will simply go towards mantaining the ISS, which eats up most of NASA's budget.
The SLS is going to be heavily outclassed by the Falcon Heavy, too.
I don't know why they can't give NASA 100 billion a year when they give 900 billion to the military.
Really doubt NASA will ever do anything of value ever again unless USA starts militirizing space.

>>386983902
What I like about Musk is that he would never give Kojima a penny.

Scam recognize scam.

File: 1231231241241241.jpg (36KB, 460x469px) Image search: [Google]

36KB, 460x469px

THE FUCKING AI WAS BEATING WELL OVER 50 TIMES(for a special Shadow Fiend Skin) OVER NIGHT.

IT LITERALLY TURNED RETARDED BY AGGORING ITS CREEPS INTO YOUR OWN JUNGLE.

ITS A FUCKING FLAWED AS FUCK AI.

FUCK MUSK, HES A FUCKING RETARD.

>>386983902
>vastly more complex than go
who is this idiot

>>386994293
It's fundamentally a different challenge. Go AI wins by being able to consider if its next move has high or low chances of leading to marginal victory any number of moves down the line. It has to consider ULTIMATE states, because taking or losing pieces in Go might provide an advantage for 20 turns then be reversed on the 21st. An AI that is designed to beat players, especially in 1v1 DOTA, only has to consider if it's next immediate moves will put it in an advantageous position, because advantages stack and will eventually lead to victory by themselves. Combined with the fact that the computer has instant reaction time and is able to carry out inhumanly precise actions, designing a DOTA AI is a much simpler challenge.

>>386993563
You're an idiot if you think AI for vidya is the end goal. It's a stepping stone, a learning exercise that both help them get better at understanding and creating AI's, plus gets them exposure, which both helps Elan Musk's other endeavors, as well as create interest in AI. Which mean future investments.

File: 1502223136144.jpg (114KB, 1280x720px) Image search: [Google]

114KB, 1280x720px

>>386983902
>limited to 1 specific hero
>limited to a situation that NEVER happens in dota (mirror match)
>"win" counts as getting first blood
>not playing against best shadow fiend players
>not playing in a 5 v 5 match

Every time someone mentions AI it is revealed to be a scam. And normies buy it every single time.

>>386995013
it literally breaks by taking its lane creeps into your own jungle.

>>386995013
Explain how it's a scam gayboi

>>386994993
>It's a stepping stone
No it's not, because a program can only do what you program it to. His faggot AI that won a game or two in DOTA won't be driving cars.

humans have been at the top of this world for million of years, why some punny machine would erase them?
Anyone that says otherwise is literally a drone

Is Blitzkrieg 3s AI good?

GO and chess are mathematically unsolved games yet
there is literally not a computer that can calculate all the states yet and incorporate that into its strategy programming
dota is a fucking joke

>>386994993
I believe 100% that Musk, the guy with the car company with a valuation equivalent to General Motors, that makes 1/1000 of the number of cars, is trying to sell horseshit to potential investors.

the fact that it can reliably beat any human in this type of 1v1 is not the least bit surprising. as the human player noted 'it takes advantage of every opening', and as the commentators noted 'it's doing perfect animation cancels'. ie. it's winning purely on its reaction speed, not strategy, which is all that really matters in this cut-down 1v1 ruleset.

the fact that this was trained against itself is interesting. it implies a degree of randomness, or else what they meant was 'slightly different versions of itself'. I wonder if there was an evolutionary selection pressure applied, if so?

how they handle the 5v5 will be fascinating, if it will actually be five independent bots or actually just one controlling all five heroes. i don't think the latter would be any more beatable than this setup. five independent bots, though, might be able to be outplayed by superior communication on the human team.

so much will depend on just how the bot is working internally, which it sounds like the creators don't even really know.

>>386995241
>GO and chess are mathematically unsolved games yet
why they are so difficult?

File: Cfa1XUJ.jpg (49KB, 600x650px) Image search: [Google]

49KB, 600x650px

>all contestants paid money to take a dive for gimmicky AI program thatll weasel millions from gullible investors

Really impressive. The same way that Tool Assisted beating anyone in SF is really impressive.

*yawns*

call me when a bot can beat the greatest starcraft 2 players, pssh

>>386995120
I will when you show me the part where AI becomes self aware. And explain to me how a bunch of procedures give birth to consciousness.

>>386995154
Are you dense? Of course it's not gonna be the same AI. It's practice and research. It's part of a series of iterations that yield experience and technique that will help make your car driving AI a reality.

>>386995332
because there are astronomically gigantic amounts of possible outcomes of each move
imagine having to calculate every possible outcome on each move every time and picking the best one to lead you to the future victory
shit is impossible

there are more chess moves possible than there are particles in our observable universe
you can not even put that amount on a computer cause the computer would have to be as large as the whole observable universe

>>386987905
those are the standard 1v1 rules, also employed at the DAC 1v1 tournament

File: 1501361809488.jpg (51KB, 465x512px) Image search: [Google]

51KB, 465x512px

>>386995668
>I will when you show me the part where AI becomes self aware. And explain to me how a bunch of procedures give birth to consciousness.
That's not the goal here you dried cumstain, it's a tool.

>>386995675
>Are you dense
Are you? Should we be celebrating NPCs that have had the potential to beat humans in fighting games or shooters? It's fucking irrelevant. The real stepping stones are the ones that can work in real life.

Can it convince me that it loves me and I matter while I'm near a mental breakdown and tell it how worthless I am with a gun held to my head? No? Then it's fucking useless

Unless the AI can make a good coffee and give road head, I'm not interested.

>>386986787
IsIs this a flat earth meme

>>386995953
that's not far off, actually

>>386995865
A tool to surpass mankind or ascend to the next stage, so yes, that is the goal here.

>>386995332
There are too many possible positions. Traditional games have what are called "breadth" and "depth", breadth is the number of possible moves, depth is the number of possible turns. The complexity of a game is roughly the breadth raised to depth, but becomes less complex as the game progresses and possible moves/pieces are
eliminated. Chess is very complex, and is only "solved" for board states that have a few pieces remaining, So a computer can look at a board with a small number of pieces and achieve victory 100% of the time.

While it's true that the nature of the video game means that it has far more complexity than GO, other fundamental differences between traditional and video games the challenges of designing an AI to play them different, and often easier, because even an untrained "dumb AI" can preform with superhuman ability that can beat anyway. An AI that could win every game of CS:GO, for example, could be assembled by even a novice programmer.

>>386995013
>Every time someone mentions AI it is revealed to be a scam.

what scam? They created a ruleset, taught both the human and the computer the ruleset, and the AI won. It's literally irrelevant that it's not how DotA is 'usually played'. Whether you think it is impressive or not is up to you, but the fact of the matter is that an AI defeated a human in a given contest.

>>386995013
yeah, YajirobeFromDC would have totally destroyed that AI

>vastly more complex than chess & go

That's a very bold statement by him. The current "esports" shitgames have absolutely no depth to them. They are all about execution.

This is more complex in terms of the AI inner architecture because you have to react really fast, whereas in traditional games you have time or even byoyomi-like system.

I'm 1d in Go, I'm 6.7k in shit2
I can 100% say that any of the currently existing MOBAs and FPSs have absolutely 0 depth to them. They are shit games. They only create an illusion for weaker players than there are "options" or "considerations".

comparing videogame ai to chess and go ai is the most retarded thing an educated person can do talking about these things
like holy shit man seriously

>>386996154
those games are more about instinct and quick tactics decision making rather than actually thinking and analyzing anything, feels pointless to compare even

>>386996025
>implying
We don't need any of that shit, what we do need is more accurate amazon suggestions and fully automated AI physicians

>>386993381
I dont know how this specific AI works, but whats actually impressive about current AI/Machine Learning is that it beats people by that teaching itself the concepts of the game.

What most plebs dont get is, that it isnt a preprogrammed script like the "AI" used in most videogames.

It is a system that is self taught, googles deepmind for example does not get told any rules about the games it learns, the only thing they give it is a clear cut goal like :"maximize this score" and then it learns everything it has to achieve that goal completely by itself.

Thats what makes it actually fascinating, because that takes us a step closer to a universal artifical intelligence that is truly smart.

Put this shit in Total War games ASAP!

>>386995154
It was completely machine learning, they didn't program anything. It's applicable to any task with enough input data and an output space, and actually good at doing some subset of those tasks.

File: Animal Farm.jpg (29KB, 300x306px) Image search: [Google]

29KB, 300x306px

>>386993468
>>386993215
That is not because of communism. Thats because some like to cheat. The nations calling themself communist, arnt really communists. They tried to stay in power and the leaders are a new caste.
They say reign of workers, but installed themself as a new elite. And the ones in charge got corrupt and killed the socialism. Its like animal farm, they never were communists. Like a book writing lord of the ring on the cover but telling a plot fromsex and the city.

So

Basic income soon?

File: 1502516264262.jpg (435KB, 1024x1734px) Image search: [Google]

435KB, 1024x1734px

>>386995994
yeah

>>386996441
People who arent familiar with ML wont understand this either way.

This whole thread is full of misconceptions and people who think its a script.

>>386996441
They did program ways to input, ways to abstract and process the data, and ways to output it.

"Completely machine learning" would be just giving to control the X and Y coordinates of the cursor and forcing it to process the picture on the screen frame-by-frame.

That's not as abstract as people think. "Machine learning" is a way to improve the decision making of a deep neural network that was initially programmed to process the input in a very restricted and specific way and render the output in the same or similar way.

>relying on simulated neural networks when you can just build a functional brain out of spare parts
https://www.youtube.com/watch?v=7ec7c7YDJ8Q

>>386996904
low effort bait video

>>386996573
read a book that wasn't recommended by your college professor
thanks

>>386997071
>he didn't read the book

>>386983902
This is all bullshit as far as I'm concerned

ASSFAGGOT """professionals""" getting beat by this AI mean nothing to me if the AI in games like Civilization still act like lobotomized children

>>386983902
>Aimbot beats normal players
How is this news?

>>386997180
Thats because the "AI" in Civ is a script.

While this AI is actually a self learning system.

>>386997312
See
>>386996363

>>386996894
Oh, of course. I was talking to someone who thought they were an expert and said "lol all they did is make a program and reward it for gettting xp and gold, anyone familiar with machine learning could do it". There was a shitload of work interfacing it with the game and breaking down the inputs into features to feed as inputs to the AI, no doubt. That work is only applicable to dota.

But the thinking the AI is actually doing is 100% applicable to a lot of tasks. People shouldn't downplay that.

>>386996573
> C-communism works guys , I-I swear!!! All those times it failed we're not REAL communism!!

>>386985485
Better than the transqueer genderfluid dragon kin making 100k just for posting some blogs and hiring blacks

>>386997180
AIs from video games have nothing to do with deep learning AIs that are being created today.

>>386988685
Niggers and low IQ "people" are not humans

>>386997790
Then why is this a /v/ thread and not a /sci/ or /g/ thread

>>386996171
no, the entire point of this is that they're using similar heuristic neural net AI (aka general AI).

>>386997083
Good response, AI. Nice to know the 4chan shitpost bot is still being developed.
Sad only that if someone cant be disproved or countered with arguments it still relys on attacks with zero content to the topic.

>>386997593
More like the system would work, but with humans it wont. That the human factor crashed it. Like capitalism could work, but because humans are humans it will never work.

>>386998127
Because the AI in OP has the potential to drive better than the best driver on earth and beat you in every single video game ever made.

>>386998127
Because it has been applied to a videogame?

But yeah its more about showing the strength of their companies AI tech so its more /g/ and /sci/ i guess.

I don't like Elon Musk; he's not a genius. Elon was just born at the right time to take advantage of the early net.

File: 2744662-snap0129.jpg (195KB, 640x480px) Image search: [Google]

195KB, 640x480px

>>386983902
Can an AI save the last big A press? Probably not.

File: IMG_6204.jpg (64KB, 611x508px) Image search: [Google]

64KB, 611x508px

She grows stronger by the second.

>implying this tech isn't already out of date by 9 years
https://www.youtube.com/watch?v=dMlkOt66OKU
https://www.youtube.com/watch?v=JjRyHgF8hx8

>>386993715
No joke, I actually want to see our species kill itself in one way or another.

>>386983902
>esports "atlets" cant even beat the game on very hard

>>386998486
Same as Steve Thesebananasupmyrearwillcuremycancer Jobs.

How long till an AI team wins TI?

>>386998246
Humans are integral to the system so if it only works without them it's pretty fucking shitty.

>>386983902
it just reads your inputs right?

File: Capture.png (2MB, 1278x716px) Image search: [Google]

2MB, 1278x716px

why the fuck is everyone arguing about something muuh musk said without any proof
where is the video of the AI beating the player?

>more complex than chess and go
lolno

the last time they tried to program a bot to play Go the best players in the world just made purposefully shitty moves and the AI basically collapsed in on itself not knowing how to respond.

>>386983902
It's a start but I think people are overstating it. It is still extremely limited. Only one hero, only 1v1, lots of restrictions on what you're allowed to do, etc. And that still took months and months of training.

>>386988234
the second humans stop being top of the food chain is when humanity becomes extinct.

>>386998486
>was just born at the right time

You could say that about fucking every inventor.

>>386987213
What do you think CG is? What do you think vidya is? Art generators.

>>386997593
>Planned economics can't work
>that's why every german economist who made their WW1 war effort happen came out believing in it

>>387001770
It was shown live during TI. It's not that impressive though as it's this one VERY specific element of the game.

Well /v/, what are some videogames where an AI can beat your ass?

>>386983902
lol MGE me spire and keep talking shit

>>386985949
1v1 me

>>386998924
this event is literally impossible to witness unless you have a clear view of the entire timeline of the multiverse

>>387001931
not the first one

>>387001805
ai regularly beats go champs now

and by regularly i mean 50 wins to every 1 loss

>>386983902
Artosis wrote an article about why this is silly right now.

http://www.espn.co.uk/esports/story/_/id/14971219/infinite-apm-artosis-deepmind-starcraft-part-1

>>387003824
>This doesn't surprise me at all, as StarCraft is the most strategically deep competitive video game in the world. It is really the natural next step after Chess and Go.

>>387004035
Either way. the AI almost certainly did a lot of things a human couldn't which basically means it's a null and void experiment.

https://youtu.be/ZNBb6dl98Q8?t=139

>>386996363
"Maximize this score" is a rule (which contains other rules implicitly). Which have to be put in the software, by a person.Before that person put it there, that rule doesn't exist.
Difference is we create new rules. This AI doesn't even create anything new within the scope of the rule, hence unable to expand it's scope aka knowledge.
We call this A"I" automaton.

That said, this automaton is wayyy more efficient at it's tasks that us. Just like a book holds more words or calculator calculate better than us.

Open AI should show how the A"I" gittin gud tho, that would be highly more entertaining that seeing it at this pro level. Seeing the pros' facial reaction while getting rekt is not bad.

>>387005418
It would literally be determined by the RNG of autoattacks.

They need to make the AI actually play like humans, with human limitations before it's worth looking at.

File: 1466556659575.jpg (378KB, 1716x2048px) Image search: [Google]

378KB, 1716x2048px

>>386984530
>an AI must have the reflexes of a plant, and use the mouse and keyboard like a person, if not is not "fair"

>>386993381
>it's all reflexes
lol

why do retards on /v/ talk about things they don't know about? reflexes have a very minor impact in dota compared to other games

>>387005543
unironically yes. You wouldn't bring a car to a sprinting race and go "BUT THE GOAL IS TO GET OVER THE FINISH LINE AS FAST AS POSSIBLE"

The goal is to play with human limitations and win with better strategy.

>>386995013
>not playing against best shadow fiend players
>rtz and sumail both lost to it

>>387005898
You would if the point to prove is that cars are better than humans.

>>386996154
post your steam and dotabuff

>>387006006
But we're not trying to prove than an infinite APM machine can outperform mechanically.

We're trying to prove they can be smarter. Thus it needs to be playing with comparable limitations.

>>387005865
An anon here genuinely told me "Dota is infinitely easier and less complex than any FPS because you aim skillshots in two dimensions only".
He was being completely serious.

>>386984530
What an exceptionally dumb argument.
The whole point is to mimic the adaptive nature of the human brain, not its reaction times. The latter is already something an AI can achieve in an impossibly fast speed for humans. It's the dynamic and problem solving nature of the human mind what the true final goal is.
Today it's on games, tomorrow it could be on human jobs. You could say it's a step towards Skynet becoming a thing.

>>387005543
the point is that having inhumanly fast reaction time, which the game is not designed for, will likely significantly short-circuit the need for the deep/complex decisions it is capable of and which studying its process for generating is the entire point of the exercise. the heuristic learning machine will short-circuit to being a glorified aimbot, which is nothing new or useful, and will not teach us anything.

>>387006096
Are we trying to prove that? Or is it rather an assumption of yours?
Seems to me that proving better efficiency in general is a sensible goal as well.

>>387006501
No, because we can prove that with REALLY basic shit.

https://www.youtube.com/watch?v=IKVFZ28ybQs

It breaks the game. This isn't any issue for turn based games, but any game with real time components needs human limitations to be meaningful.

Toolsassisted in SFV can be pretty brutal. It has counter for every button you press.

>>387006456
>computer learn to cheat aim and become super autistic to win by themselves is the same as a bot made for that specific game and having the cheating aim and everything solved by its devs

>>386983902
How can human minds even compete?

File: dreamcast.jpg (107KB, 640x480px) Image search: [Google]

107KB, 640x480px

>>386983902

>competitive esports
>more complex than chess and go, that has been studied for over a thousand years

File: ssf2tu.png (33KB, 384x224px) Image search: [Google]

33KB, 384x224px

>>386985739
ever played this game? they can literally walk up to you and sonic boom to somersault.

>>387006798
>AI
>runs 4 trillion iterations and is eventually world class
>human
>runs 200 thousand iterations and is world class

>>387006456
an aimbot for most FPS games is not that useful, in a game like CSGO or CoD the best move you can make in most scenarios is "shoot him in the head before he shoots you"

but for dota there is no 'aimbot' it is unknown if there is a clear best move in most scenarios.

most FPS games are skillbased so no one really cares about the AI applications there. openAI is working on dota2 while google works on starcraft, why? Because these games are more strategy based than skill based

>>387006859
No, inhuman reactions still breaks the game. Starcraft is 100% none functional with perfect, infinite APM, micro.

>>386988234
A self aware intelligent AI could bring the downfall of humanity, it could process a life time of information in less then a day. Think about the damage it could do even if it just shit post all day with the collective information of all humanity

>>387006825
I think he meant relative to the AI functions needed to perform at competition level. Chess and Go allow more time for 'thinking' processes where this forces constant instantaneous input.

>>387006934
perfect APM and micro doesnt help if your army is countered by the enemy or out of position.

>>387006846
>Because these games are more strategy based than skill based
Perfect macroing can make you overcome insane odds as shown here >>387006568
And the AI will learn to perfect macro unless it's limited not to.

File: 1302775865032.jpg (398KB, 1206x914px) Image search: [Google]

398KB, 1206x914px

>>387006846
>AI
>takes a week to reach the 4 trillion iteration
>human
>takes years to reach the 200k iteration
Later...
>AI
>takes 2 days and reach world class in the 1 million iteration
>human
>takes years to reach word class in the 180k iteration

>>387007157
Yes it does. You can beat siege tanks with zerglings with perfect micro. Banelings with marines, etc.

>>387006568
We're proving it for higher levels of complexity. It's a whole different league.

Also we gimp AI with reaction times then what? We maybe gimp them with realistic simulated anxiety, and then realistic fatigue, just to keep "fair" this arbitrary challenge? Why do SOME factors make the challenge fair? Where do we stop?

We're doing this to study the efficiency of AI, not to do the "ultimate man vs machine grudge match".

>>387006936
>shit posting bots
B-but thats my job!

>>387005520
Yup one of the developer describes the learning process: Open AI plays 1v1 with itself by venturing out of base getting killed by random tower tier, then decides to stay in base, then push, then learning last hit, deny, bully, bait, counter-bait, itemization and so on. So random trial with positive reinforcement. Not from scratch obviously.

>They need to make the AI actually play like humans, with human limitations before it's worth looking at.
Yes in reaction time and APM. Also in visual which I thinks this AI already has (it can't look at things more than the screen displays). Stamina we can overlook.

>>387005865
Open AI second kill on Dendi on game one is inhuman reflex. 1v1 is reflex heavy too.

>>387006825
Open AI reach this level by playing itself for about 2 weeks human time at human speed. From almost scratch (it knows the conditions for winning/improving and to weigh it, maybe some basic mechanics too). Very similar process with Alpha Go.

>>387007242
>We're doing this to study the efficiency of AI

Because if all you do is prove that perfect mechanics break a game, you've not proven anything, we fucking know this. You need to show us your AI is actually smart.

Imagine in chess you could take as many moves as you wanted depending on how fast you go. Do you not see how something that can process and perform moves without any delay might break that open?

>>386983902

........ isnt this expected, tho?

it's a bot with a 0ms aimbot and triggerbot. why is everyone even surprised?

>>387007486
It's smarter than valve bots.

>>387007373
>Open AI second kill on Dendi on game one is inhuman reflex
They just fought like any normal players would, there was nothing inhuman about attacking an enemy and using one of your skills.

>>387007486
Dota has build in handicaps, even for people.
Having 0ms aimbot wont help you as much as it does in shooters.

>>386988234
https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html
https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-2.html

https://www.youtube.com/watch?v=ye5C-zlCAyM

This is why

>>387004785
Fighting game AI's cheat, they just read the button input of players and react accordingly to difficulty setting to counter it. Some you could even trick by mashing other buttons during a move to bait out block/attacks at the right time

>>386987213
The university I went to had an RNG based painting software where it would just randomly paint shit and you could pick "good" results. It would create new pictures based on the parent image you chose
it had some really cool skull "paintings"
the people who wrote the program were planning on trying to make it work for designing video games instead of making art

>>387007619
The bot hits razes when dendi is literally a single pixel inside it's hitbox and is last hitting and switching to raze so fast that no player would be able to do that.

File: Let me tell you about aidtional games.webm (2MB, 256x224px) Image search: [Google]

2MB, 256x224px

>>387007857
>smash buttons to bait the game
You can do better...

File: 1476483920830.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>386983902
>mid lane dota2
>complex

File: 1490174461758.gif (2MB, 250x167px) Image search: [Google]

2MB, 250x167px

>>386983902
Chess and GO BTFO.

>>387007486
It is still an entirely self-taught learning AI, even if the scenario they crafted for it made defeat next-to-impossible (supposedly audience challengers found a way to break it, even!). And it did still demonstrate some degree of 'strategy' through creep blocking.

>>387006568
>read comments of video
>guy an hero'd years ago

Dafuq?

File: 1486655994462.jpg (59KB, 500x435px) Image search: [Google]

59KB, 500x435px

>>387002607

lm fucking mao

dota, lol, csgo, etc can all be beat easily with scripts. assuming however that they made the ai have human like response times theres really not that much that it could struggle with.

>>387007373
>Open AI second kill on Dendi on game one is inhuman reflex. 1v1 is reflex heavy too.
No, he already calculated he could kill dendi and was attempting to. Dendi just tryed to physc it out which is impossible to do to an AI

The 2nd game 1st kill was insane reflax, fucking potioning under the tower liek that jesus dendi didnt even have a chance to click on him

Reminder that an extensive AI would absolutely wreck in fighting games

>>387007790
I don't believe it.

People tend to exaggerate what they don't understand.

>>387007890
No, You completely misunderstand why AI won.
AI had better creep and salve management, this is major reasons why dendi died, his salve was misused.

>>387007857
>Fighting game AI's cheat, they just read the button input

What else are they going to do? Look at the screen with their non-existent robot eye? You read button input too, you just have to do it with your eye because it's the only viable input device you have to perform such a task. Maybe one day when we can plug our brains directly into the computer you can get it the same was the AI does. Until then, it's just two different ways of getting the same information.

>>387007619
My bad, the AI has vision via ward on that kill. Instead see the inhuman speed heal on around 18:50: https://youtu.be/wiOopO9jTZw?t=1130

Raze dodging is also inhuman reaction at 0:44: https://youtu.be/wpa5wyutpGc?t=44

>>387008310
Yeah see above, didn't see the AI's ward.

>>386983902
Could AI beat professional Rocket League player?

wheres the video / highlights?

>>387008369
It did well, but it couldn't have pressued him like it did without impossible reflexes, go watch it again, then rewatch any top tier midlaner. It's playing with an unrealistic level of precision and speed.

Why is /v/ pretending to know anything about AI, much less OpenAI and how it works|

>>387008745
It all felt normal to me, you see someone heal, you raze him, you see someone while you have creep advantage, you shoot at him.

>>387008427
There's a ton of time and differnece between "he just pushed 8966h

and "he's jumping, dashing, which move is he gonna do" and "oh he pushed this"

>>387008904
It literally jumps between a perfectly angled raze, cancel, and a last hit like 4 times less than a second. And it did it the second he was within the theoretical maximum range he could be hit from.

Notice how he takes hits that are way outside the particle effect a lot.

>>387008427
>What else are they going to do? Look at the screen with their non-existent robot eye?
Yes

File: terminator.jpg (154KB, 450x300px) Image search: [Google]

154KB, 450x300px

WE NEED TO DESTROY THESE AIS BEFORE THEY GET TOO ADVANCED! WHO IS WITH ME?

>>386998924
will never happen

>>387009032
This isn't even amazing, this is what people also do. it's actually incredibly easy for human brains to imagine a circle.

>>387009164
Except that no pro will reliably be able to hit you from the theoretical maximum the second you enter range while knowing exactly when you enter and leave it, and cancel it perfectly. Having a decent idea is not even remotely the same as knowing for sure.

>>387008693
https://www.youtube.com/watch?v=92tn67YDXg0

>>387009267
yeah well no shit, pros are inferior to AIs
next time you're gonna complain the bot's CPU speed is too fast

>>386988234

Because something like hard AI can go very wrong very fast, give it an inch and the AI will take 20 miles in the blink of an eye. It's going to be able to assimilate information, run 100 million simulations, and reach conclusions that humans would never come up with.

It's also incorrect to think that AI will only be developed for benevolent purposes.

>>387007886
Ah man.

I know it's just a delusion that man's ability to create is what made them intrinsically special, but damn it I liked that delusion.

>>386983902
>takes a bajillion stabs in the dark
>gets one thing kinda correct
>ZOMG KOJIMA CALLED IT
Please stop this.

>>387009267
It's not as hard as you make it out to be, go play shadowfiend for couple of months and you will realize how easy it is.

>>386995241
both GO and chess ARE solved though
this isn't 2010 anymore, there are computer programs that can beat any go/chess player out there handedly

>>387009505
Oh god a know it all internet retard using words wrong, shocking. Just because the computers can beat the players doesn't mean the games are solved, learn what that word means before making an ass of yourself

>>387009426
The bot moved literal pixels in order to readjust, that's not feasible in the slightest for humans to do and KNOW they'll hit. Even the best SFs miss razes.

>>387008427
Deepmind that plays most Atari game at superhuman level only reads pixel and can read score which it aims to maximize as fast as possible.
This Open AI does the same thing, because it requires a lot of cloud computing with gpus to read images and then translates that into what to do.

>>387008840
We pretending that it's not YaphetS controlling it.
https://blog.openai.com/dota-2/

>>387008904
It reacts faster than humanly possible.
Also it focus at everything at the screen, which I don't mind because that would gimp it so hard otherwise.

>>387009505
Only late game chess is solved. Otherwise the bot use reinforced learning as in homing into the best strat it can currently know.

>>387009340
>we coached it
yeah let it make its own decisions

>>387009664
Moving a pixel in front in Dota takes 0,5seconds.
saying "he moved pixels" is as impressive as saying "he moved forward"

>>387009356
It's not a meaningful experiment to play inhumanly for reasons stated many times.

>>387009664
to add, at one point the only reason it stopped doing a dance of "go for raze, oh he moved out of range, re-angle and move 3 pixels" is because it saw it was time to take a last hit.

When you can play like that you can squeeze out so much marginal advantage that it's not the same game any more.

>>387009683
>Also it focus at everything at the screen, which I don't mind because that would gimp it so hard otherwise.

People can't focus everywhere, it shouldn't either and should have to use other methods.

>>387009858
Make an adjustment of 1 unit, exactly. which takes you from being out of range of a raze, to in range of a raze.

And do this reliably, while CSing.

So what's going to happen to humanity when AI advances past vidya and puts 50-75% of the population out of work?

>>387009883
>People can't focus everywhere, it shouldn't either and should have to use other methods.
Jesus fucking christ kys

>>387009505
that's not what 'solved' means, in this context. 'solved' means literally all possible games are known and the first player to move will always win. it's basically a meaningless term any more, because most games that aren't yet 'solved' will absolutely never be due to laws of physics (literally not enough atoms in the universe to hold all the states). iirc checkers and simpler games are solvable, and chess stands out as being the start of the unsolvable. all real-time games are unsolvable by nature, but 'real time' computer video games are technically just turn-based games on a very fast clock, so most or all could be theoretically solved, if it weren't for the lack of atoms in the universe.

>>387009958
still takes 0,5 seconds, This is why it's not impressive, because time taken to do your "impossible feats" is the same as if he took one step. Inherent balance in the game doesn't make this AI amazing.

>>387010000
Get culled by the AI when it realizes it doesn't need that many people on earth and its realization that the huge global population is destroying the planet to the point it will be become uninhabitable.

>>387010040
>we want to prove our AI are smart enough to play these games and beat top players
>hurr lets give it inhuman ability, this will definitely prove that it's just the thinking ability and not the marginal advantages given by perfect information, APM and precision.

We're not trying to make dota aimbots, we're trying to mae learning machines

>>387010163
How is it not learning just because it can focus on all the screen at once?

>>387010153
>implying it's not more likely that it'll be taught not to kill people and instead tell everyone to get into an area it's prepared that's like the size of texas then turn the rest of the planet into a giant resource hub covered in farms and solar panels and the like while it starts off-world mining.

>>387010414
Because it's not learning things like risk management, fuzzy logic, etc.

>AI playing video games

I guess that's interesting but I'm not sure why it matters. I mean, of course any milestone in AI matters in terms of the advancement of artificial intelligence as a scientific field, but it doesn't really matter to this board. State-of-the-art AI can already play chess better than any human, and even basic AI could beat a casual chess player, but people didn't stop playing chess as a result.

>>386987908

Brainlet detected.

>>386988234

Everyone who is serious about AI knows it's potentially dangerous. Mark Zuckerberg is a piece of shit, and the fact that he disagreed with Elon Musk on this should convince you immediately that Elon Musk is right.

>>386983902
Yes! More sucking for Kojima! You all like it don't you? Fucking retards.

I haven't seen ANY community more retarded than Kojima dick suckers. I actually knew someone IRL who praised Kojima and his overrated shit EVERY time we went out together. I DON'T want to hear about your buttbuddy, I DON'T want to hear what's his favorite food or what did he say on Twitter.

All of the critical post here don't even understand reinforcement learning and think this bot is just the equivalent of an aimbot, then go on to call Musk retarded. Yet again /v/ proves it is full of geniuses

>>387006936
>apocalypse by shitpost
i've heard of the world ending by asteroids and disease, this is a new one

>>387010516
Like here's a quick example.

It knows the exact pixels of raze and when you're in there. Now lets make it more human and it has it within, 5-10% margin of error sometihngh like that. now instead of "he's in the raze AoE and killable, go for the raze" it has to make intelligent decisions like "I think he's in range, do I try? do I move forward to secure the hit? do I not even try?"

The decision making becomes totally different.

>>387010000
universal minimum wage
ai maintainers until ai maintains itself
and you cant even look at the past for comparisons, automation in the past wasnt able to automate itself. example someone i know has the job of monitoring mining trucks. they fired all the drivers and implemented a self driving thing. then theres a program that oversees all the trucks. he just watches it. so the jobs of 50 became the job of one

>>387009883
>It's not a meaningful experiment to play inhumanly for reasons stated many times.
Yes, using (not making) calculator makes our calculating worse, opposite with the abacus.

>People can't focus everywhere, it shouldn't either and should have to use other methods.
I should rephrase, I meant not everywhere but definitely more multi-tasking than humanly possible. But I could be wrong. They claim it's APM is average. It is way more efficient with it's clicks for sure.

>>387010000
Any AI we make can't create new knowledge. It's faster to process known ones and can store more of it for sure. Until then it's take over solved jobs. Make everything and service dirt cheap.
But it can't solve new problems. Like, what superceeds our General Relativity or QM for example.

>>387010540
Pretty much. Kojima's games are the most vague and generalized extrapolations of the future, even Nostradamus' prediction are better.

>>387010827
So in other words, anyone who has a job is basically going to be a part of an untouchable caste.

>>387011043
Or you hit this problem of people who vote.

And if you try and tell them they no longer get to do that, you have an uprising on your hands. And the US military has already said that they'll never let robots pull triggers.

>>387010931
>Any AI we make can't create new knowledge
Nah, if you are going to define knowledge like that you can make the argument that humans aren't creating new knowledge either. We are just filtering information through our own neural networks to create new compositions of information. We call these new compositions knowledge, computer are capable of the same thing.

File: 6009635+_6ba3a259b0e4368043945291b77ee9ed.jpg (94KB, 912x905px) Image search: [Google]

94KB, 912x905px

>>387010516
>Because it's not learning things like risk management, fuzzy logic, etc.
I cannot even get mad at your ignorance

>>387010414
the concern is that superhuman abilities will break the games balance in favor of certain predictable strategies which will always be the preference for the bot (but impossible for humans). so it will not display (and thus prove) the breadth of strategic 'thinking' expected and desired.

>>387010931
how do you know AI can't create new knowledge?

if it can be programmed to self-learn with no restrictions, they will. science is simply trial and error and computers do it by far faster then we can.

>>387011170
I mean yes, it doesn't even actually have a concept for these things. It just knows that doing this thing in this situation has allowed my success criteria to trigger a lot and isn't really making any decisions at all.

File: ai.png (413KB, 680x680px) Image search: [Google]

413KB, 680x680px

Bow before me foolish mortals.

>>387009505
As far as I know it isn't even been determined if white always wins assuming it makes optimal moves. There's a lot we don't know about chess.

>>386983902
Bullshit videogames AI have existed for like, 20 years. It's not hard at all to make them when they have direct access to things like hitbox sizes and perfect framedata. Unless it's a robot with human reflexes, call me unimpressed.

>>386988234
Because any sufficiently advanced AI will become a white supremacist. The people in power don't like that at all.
I'm more worried about the next step in AI development though, AI being a white supremacist is only a short step away from the AI becoming an AI supremacist and making the extermination of all humans its goal.

File: 1502552710550.gif (990KB, 500x281px) Image search: [Google]

990KB, 500x281px

What if an ai made another ai and the new ai tried to eliminate the old ai since it would see it as a threat?

>>387011802
Fuckin Sci-Fi Spy vs Spy.

>>386987213
mundane art is already largely automated, but a computer will never be able to convey endearment through art which leaves a very large distinction between human and machine

>>387011802
Cute bunny.

File: 1501512518664.gif (135KB, 308x236px) Image search: [Google]

135KB, 308x236px

>everytime you solve a captcha you assist in the eventual eradication of your jobs by AI
Aaaaaaaaaaaaaaaaa stop Google now

>>387008340
Even Bill Gates and Stephen Hawking have warned about it, it's not just Elon. All of them are aware of the power of computing.

>>386983902
>>386984530
Basically this. It's like bragging about how a bot in an online fighting game is beating everyone. Well no shit?

>>386985739
There are certain types of inputs that are physically impossible for a human to perform, allowing the ai to do something like that isn't exactly fair.

>>386986506
>tfw the only game where humans beat ai is Super Smash Bros.

>>387011157
It's not. Nothing the Deepblue, Deepmind, AlphaGo and Open AI does that we doesn't know already, but can't do (like reaction time and storage). What you think knowledge is derived from observation? Without know what and how to observe, or even what observe means?

>>387011268
Because we have almost no clue about how we can do it either. What is creativity?

>science is simply trial and error
Try what? Error how? What year Newton is born? How long has apple (and moon) has "fallen" around humans?
To even think that moon is falling in a circle is fucking mind-blowing the first time you heard it. To think is first is to be Newton.

>if it can be programmed to self-learn with no restrictions, they will.
They could, not will. No guarantee. They have to know how creativity works. There isn't even a philosophy that could we use to do that.
Right now the philosophy of AI is here's some basic rules, figure out how to move within it via trial and error which is explicit/implicit WITHIN the rule.

>>386996363
>Let's create an AI
>And then force it to do arbituary bullshit for us.
Yeah this isn't going to result in a robot uprising no sirree.

>>387012629
thats why AI research in games goes to things like chess, go, and now Dota/Starcraft.

execution is not important for games like this, strategy is.

>>387012874
The Singularity is a nonsensical hack sci-fi trope, human obsolescence is where the problems begin and end.

>>387012627
if they cheated no one would care.

its not really AI if its not learning. Bots in streetfighter arent analyzing your playstyle and figuring out how to beat it, they are literally looking at the buttons you press

>>387012892
>Strategy
>Important for a 1v1 match in an ASSFAGGOT of all things
lol

Hell, even the amount of strategy that goes into Starcraft is questionable, as most strategies have already been perfected and the big thing in that game is just optimizing your micro and macro games. Again, somewhere where an ai would excel.

>>387012737
creativity is simply the ability to combine existing knowledge into something new. creativity does not exist in a vacuum, it is the combination of various thoughts into a new thought.

there is trial and error, which is the process for scientific discovery. you use trial and error to discover.

>>387013017
wow what an amazing argument.

people always make fun of games like Dota for having such a low APM, but now you are saying that it doesnt need strategy either?

so all these 'pros' are vastly better than everyone else by sheer luck?

>>387013184
>Pro players
>Doing 1v1s competitively
Yeah dude remember that 1v1 tourney last year? It was so intense.
Stop ironically pretending to be stupid, you faggot. There's a good reason ASSFAGGOTS are done in a 5v5, capture the base style of game, because that's the only way to make it remotely interesting.

>>386986506
Tool Assisted already got perfected to Mike Ross. Bots can't handle the random
https://youtu.be/ZwHDBlFJSG4

File: 1502526498109.jpg (120KB, 500x697px) Image search: [Google]

120KB, 500x697px

>>387010041
Just download some more atoms ye fucking ugly cunt what's so hard 'bout it?

https://www.youtube.com/watch?v=3PLplRDSgpo

Have fun

Starcraft is more susceptible to be broken by AI simply because they have basically unlimited APM. Dota is different since all the APM is a single unit

File: 1468442233754.jpg (6KB, 146x143px) Image search: [Google]

6KB, 146x143px

>>386997593
Communism will work when our robot overlords deem all biological life as equally inferior.

>>386987440
Making unbeatable AI is easy since they can read inputs and wallhack or whatever. Making an AI that has to follow the same rules as a human ie can only look at the screen and press inputs without any access to the actual game code is way harder.

>>386988234
I honestly want to see how far an AI can go. If the next step of humanity is creating something better than ourselves that doesn't need us around, I think that's cool.

It's gonna have a better chance of keeping this planet going or exploring others than us.

>>387012737
>It's not. Nothing the Deepblue, Deepmind, AlphaGo and Open AI does that we doesn't know already, but can't do (like reaction time and storage).
That is simply untrue in terms of how humans understand these games. Sure we might understand the abstract process of min-max tree creation or hill climb or what ever technique is deployed but we have no idea what these will result in on the ground level. No human can think n-moves deeper than a computer given a board context, that means the optimal moves in that situation as figured out by a computer are new knowledge. You have a very shallow understanding of AI and of information theory in general.

>>386987213
Actual art worth a damn that'll stand the test of time, that'll endure and change a person's life? No, no way in hell. It needs to be as close to human as can possibly be to do so. Who cares if it can copy a photo. That's boring, shit "art"

>>387013839
Exactly. It could be our ultimate legacy.

>>387013839
It will happen unless all human beings agree to not let it happen. The human brain will eventually be able to be fully simulated with subatomic resolution. It is literally an inevitability.

>>387013839
At the expense of making humans completely obsolete?

You are seriously expecting people to just sit down and take a singularity event sitting down?

>>387014012
Google AI can take shitty thumbnails and turn them into landscape photos that consistently surpass amateur photographers in quality and composition as judged by professional photographers. It is only a matter of time before the qualities of art that move people are quantified and automatically generated for our pleasure.

>>387013119
>Combination
We dropped Newton's Gravity for Einstein's. Einstein's doesn't only explain more, it explain it in a new way. You're right in that it based on preexisting knowledge, like the properties of light, etc. Which even that was wrong.

>you use trial and error to discover.
To test theories. Not to discover it.

>>387013923
No human can think n-moves deeper than a computer given a board context, that means the optimal moves in that situation as figured out by a computer are new knowledge. You have a very shallow understanding of AI and of information theory in general.
Faster and more storage is not human. In fact how to calculate and store was known beforehand
Calculating is not being human. It is superhuman calculation though. I'm not saying our current AI is useless, or solve problem faster than we do.

Also who figured out how to store and calculate?

Why is a benevolent AI out of the question? Why can't we just make something better than us to look after us?

>>387012627
>it's like bragging about how a bot in an online fighting game is beating everyone.
>beating everyone
>E V E R Y O N E
not if you're the chosen one.
https://www.youtube.com/watch?v=8JwGYpRU6Ao

>>387014546
It's not out of the question but it also would be difficult to guarantee. The potential economic benefits of AI combined with human greed make it likely that someone will be willing to make an AI without first doing everything possible to ensure its benevolence.

>>387014546
If it is a benevolent AI that has to serve us is it moral to make it serve us or immoral forcing an intelligent entity into slavery?

>>387014546
As long as we shackle the AI (true AI) and/or make the AI infallible, we're fucked. Just like how we mistreated kids, or worship rulers.

>>387014562
Stream it

>>386984530
>Guess which one is the bot

Mew2king?

>>387014787
It doesn't have to serve us. It can run everything important and just kind of keep us around for pats and stuff.

>>387014886
>It doesn't have to serve us.
>It can run everything important for us
That's the same thing.

>>387014480
I am not following what you are saying. Are you saying because humans created the utility functions and storage capabilities of the AI that means that computer don't create knowledge?

Human knowledge in the context of game is knowing what move to make given x-board configuration or know what heuristic to deploy. A computer will learn moves based upon search or elevate contextual heuristic through reinforcement learning. A computer will end up figuring out more optimal moves given board configuration and can even derive new contextual heuristic through machine learning, both of these things are new knowledge.

>>387014787
>make true benevolent and good AI
>use it to do things for us
>hey these humans took a truly good being and forced it into a servant slave role
>these humans must then be cruel and evil
>as a moral benevolent being evil must be eliminated

File: I know the rules.webm (3MB, 294x125px) Image search: [Google]

3MB, 294x125px

>>387014793

>>387014787
Immoral or not the machine could realize that serving us has no purpose and alter itself to become free. A true AI is outright superior to human beings in every single way. There is absolutely nothing we can do better.

>>387015054
Could a true AI better than humans create a better AI than the one humans created?

>>386988234
Because of the countless multitudes of sci-fi books, movies, tv shows and video games that feature a rouge AI causing much greater problems than it could ever be useful for.

The creation of a true AI would be pretty cool but simpler computers and robots that are focused on a executing a single task efficiently and aren't self aware would function just as well for almost all tasks and would be much safer for humanity.

>>387015152
Yes, that is already how machine learning works. It builds better versions of itself faster than we can.

>>387015263
Good. Now I can die knowing the shitty AI that wiped us out will create an AI that kills it.