Don't worry the A.I. will integrate any kind of exploit in it's self evolving script. Don't think you can repeat this to smithereens. Been watching A.I. develop from Alpha Go and Zero ... it's absolutly Unbeatable as you give it more and more time to evolve it's self.
The Machine I doubt there will be an unbeatable bot. Dota is far more complex and isn’t something that can be mathematically broken down. Sure 98% unbeatable but never impossible.
I don't think you understand. Some games are "simple". There is 1 right way to play, and regardless of what your opponent does, 1 set of algorithm will rule them all. This type include all games with perfect information. Chess, go, tic tac toe, connect 4 etc. Other games are not so simple. AI will NEVER win 100% of rock paper scissor for an extreme example. You can never perfectly predict what your opponent will do, and steps you take to counter one strategy will hinder you if your opponent is running a different strategy. Think about this. Even before we discuss whether OpenAI takes creep health in as a parameter at all (OpenAI works on a reward based structure, how do you propose to implement creeps in there?), say to counter this strategy it learns to randomly dust when pushing lanes, or put down wards. How do you think it will fare when it plays against a variant of itself that doesn't? It will lose of course because it "wasted" gold. OpenAI can try as hard as it can to reach a local maximum in performance, it can never reliably reach a global maximum. Such a maximum might not even exist.
@@randomnobody660quite possible. I didn't necessarily compare the Chess or Go with Dota, the point was arround the ideea that whatever humans think atm they understand about a game like the above mentioned suprise was in order when following A.I. giving it a shot. I follow chess more or less and the last Stockfish is rated arround 3200-3300 fide rating style and Magnus Carlsen the world champ is arround 2800 fide and he simply cannot beat the engine, like Never. Now Leela Zero the A.I. is rated arround 4.4 k rating and Stockfish cannot win Any game against the A.I. The engine works pretty much in the same way with computing variations to the end and because there is a limit to the number of possible variations in a line (even if of the order of billions) the engine should be able to hold a draw at any given point -> he doesn't ... the reason he doesn't is simple ... the A.I. understood somehow that Not playing for the winning line can turn into a win by forcing Zugzwang lines that the Engine does never take into account because it's not a technical win at the end he simply overlooks it while the A.I. doesn't because he understood that this can also become a win. Now, the computing power of a super engine is similar to a A.I. computing power, there is no substantial diference, more so, there is a high cahnce the Stockfish developers could use a even more powerfull proc to compute then Leela Chess but the mechanics behind it are totally diferent. I get your point and it's valid but your closing argument is quite a belief that the A.I. can't possibly Expect every possible outcome in a game of dota ... he can't, but what he can do and always will is to integrate Any strategy that at one point might work against him. And don't forget, execution of A.I. is Always flawless, even when they get ganked in Dota2 they die while doing the 100% outcome of that death trap ... So we got 100% flawless execution of play and even if there is some low probability to overlook random human behaviours the perfection behind the execution will force that 98% win always Up to 100% never down. And with any 'suprise' that we can steal a win he will integrate it over time.
@@verbalwound5874 You are troll for sure... At that time open ai were bought with new improvements. Each year at first open ai seems unbeatable unless someone figures out a counter strategy...
AI always have exploits. Im starting to think that humans adapt alot more faster than these ai's. One moment before everybody was losing, then later on everybody is winning
Verbal Wound imagine doing this kind of shitty play in front of an Airing program which thousand of people are watching you anticipating to see great teamfight and you’re here with a riki radiance not joining a single team fight just pushing the wave like a 1k mmr would do
*John Connor on a mission to destroy Skynet*
John connor uses riki to destroy skynet.
Colorized 2059.
Jajajajajajaja yeah
ROFL
-Riki- : *it* *was* *me* *all* *along*
itani lead this comment
This right here is the reason of the limited hero pool and nonexistent draft.
Bot would get shreked so hard lul.
it will learn 😆
*ThE OnLy rEaSoN wHy We DiDnT aLlOw mOrE hErOeS iS bEcAusE it wOuLd be UnFaIr tO tHe hUmAnS*
I'm a little surprised this works. Even the normal bots will insta dust you if you get anywhere near them with invis/radiance.
normal bots actually do have maphack lmao
no they wont, stop talking stuff you know nothing about
Because they are coded so, but these AI teach themselves, this should be their first time seeing this kind of strat
Normal bots kinda have some cheats like seeing invisible but pretending they don't
Normal bots have sight of you, they'll even dodge all your mirana arrows
Rip headphone user..
Haha
?
I think openai stopped improving after they lose to black, maybe they are having a trauma like hoomans
This will be the standard build for riki in games from now on
1% chance to win.
So you’re saying there still is a chance?
even one in a million chance is still a chance
There is always be , if radiant start throwing the game bruh .. Its dota
Below 1%
Dude just imagine the 1/14m that they win against than- ow wait is fictional so it doesn't count.
OpenAI: There was no other way. We're in the endgame now
I love it when the beat Drops at 1:05
Don't worry the A.I. will integrate any kind of exploit in it's self evolving script. Don't think you can repeat this to smithereens.
Been watching A.I. develop from Alpha Go and Zero ... it's absolutly Unbeatable as you give it more and more time to evolve it's self.
it stopped learning smartass
The Machine I doubt there will be an unbeatable bot. Dota is far more complex and isn’t something that can be mathematically broken down. Sure 98% unbeatable but never impossible.
There is "fail ladder" exploit that still works against AlphaGo clones
I don't think you understand. Some games are "simple". There is 1 right way to play, and regardless of what your opponent does, 1 set of algorithm will rule them all. This type include all games with perfect information. Chess, go, tic tac toe, connect 4 etc.
Other games are not so simple. AI will NEVER win 100% of rock paper scissor for an extreme example. You can never perfectly predict what your opponent will do, and steps you take to counter one strategy will hinder you if your opponent is running a different strategy.
Think about this. Even before we discuss whether OpenAI takes creep health in as a parameter at all (OpenAI works on a reward based structure, how do you propose to implement creeps in there?), say to counter this strategy it learns to randomly dust when pushing lanes, or put down wards. How do you think it will fare when it plays against a variant of itself that doesn't? It will lose of course because it "wasted" gold.
OpenAI can try as hard as it can to reach a local maximum in performance, it can never reliably reach a global maximum. Such a maximum might not even exist.
@@randomnobody660quite possible. I didn't necessarily compare the Chess or Go with Dota, the point was arround the ideea that whatever humans think atm they understand about a game like the above mentioned suprise was in order when following A.I. giving it a shot. I follow chess more or less and the last Stockfish is rated arround 3200-3300 fide rating style and Magnus Carlsen the world champ is arround 2800 fide and he simply cannot beat the engine, like Never. Now Leela Zero the A.I. is rated arround 4.4 k rating and Stockfish cannot win Any game against the A.I. The engine works pretty much in the same way with computing variations to the end and because there is a limit to the number of possible variations in a line (even if of the order of billions) the engine should be able to hold a draw at any given point -> he doesn't ... the reason he doesn't is simple ... the A.I. understood somehow that Not playing for the winning line can turn into a win by forcing Zugzwang lines that the Engine does never take into account because it's not a technical win at the end he simply overlooks it while the A.I. doesn't because he understood that this can also become a win. Now, the computing power of a super engine is similar to a A.I. computing power, there is no substantial diference, more so, there is a high cahnce the Stockfish developers could use a even more powerfull proc to compute then Leela Chess but the mechanics behind it are totally diferent. I get your point and it's valid but your closing argument is quite a belief that the A.I. can't possibly Expect every possible outcome in a game of dota ... he can't, but what he can do and always will is to integrate Any strategy that at one point might work against him. And don't forget, execution of A.I. is Always flawless, even when they get ganked in Dota2 they die while doing the 100% outcome of that death trap ... So we got 100% flawless execution of play and even if there is some low probability to overlook random human behaviours the perfection behind the execution will force that 98% win always Up to 100% never down. And with any 'suprise' that we can steal a win he will integrate it over time.
So notail was right, it is easily expoitable, thus beatable.
Wow nobrain was such a Sherlock, why didnt he win against openai in the show match then?
@@verbalwound5874 You are troll for sure... At that time open ai were bought with new improvements. Each year at first open ai seems unbeatable unless someone figures out a counter strategy...
AI always have exploits. Im starting to think that humans adapt alot more faster than these ai's.
One moment before everybody was losing, then later on everybody is winning
@@verbalwound5874 cuz it showmatch, win gor what? Make openai win more entertain for viewer
Thats game already scripted
Verbal Wound imagine doing this kind of shitty play in front of an Airing program which thousand of people are watching you anticipating to see great teamfight and you’re here with a riki radiance not joining a single team fight just pushing the wave like a 1k mmr would do
Open ai: *new update...
*Complete, put ward everywhere when against riki
Open ai : you cant do that, that's illegal
How do you play OpenAI bro? Can't seem to download it on the bot download list.
At first i thought my mobile sound was broken
Damn, that was so loud, my ears are bleeding.
SIR how to play it?
I see your a man of culture as well....
How did you get the openai bot?
the power of memes
Это правило всегда работает - ИНВИЗ ИМБА
How do you play against open AI? I thougt its only available if your permission is approved by valve ex.tournaments
The only way to beat them is not the pro player but the memes player
I don’t get it. AI doesn’t understand sentury yet?
So the same trick couldnt be work on them?
Ok!! How to play open ai bots?
How to play against open ai ?
Pick invoker buy radiance, cold snap then ghost walk
I do this alot back in dota1. Ofc minus the hammer 😁
How to match with open ai? Sorry im newbie at dota 2
how can i play with oepn AI🤔🤔
where to download this?
1:41 wtf is that? cancer SF herald rank
How do you play against openai?
What's happening!?
Did 100k people just watch this shit
Music title pls
Song name at 3:22?
no sound
IM DEF
How can you play with these bots?
Dota1 be like
Why does the Leaderboard still says nobody won tho.
cuz it's a solo game with 9 open ai bots. this is just a demonstration on how you can exploit them.
How to play that Solo Game with 9 Open AI Bots?
@@falcon176 how tf you can play with it? Workshop?
He picked forbidden hero rofl
@@sofyanarief6447 register at open ai web
Nice exploit
Open ai easy mode
Lol AI or whatever bots will always be dumb.
>doesn't disassemble boots for aether lens
Come again nigga
Тупые боты не могут купить дастов и центрей
Sasha Tolstoi они покупают досты и центри
На видео баг они не буду пожимать дасты пока Рики не ударит поэтому он купил радик
SEA CANCER NEW META OMGGGG STOP POSTING THESE TO RUIN OUR GAMES