openai.com/ - OpenAI Official Website / openai - OpenAI Updates Commentary by ODPixel + DotACapitalist OpenAI - / openai OpenAI FIVE vs TI8 CHAMPIONS FINAL 2019 SUBSCRIBE ►bit.ly/noobfromua
Just imagine they just have to put deep learning with comentators' voices then the bot can learn the situation and style of how these people will choose words according to the situations.
@@pk7685 yes, let the 2 AI teams enter the tournament for fun then watch them destroy people and play together. It is not long before we see AI fight. Imagine they have an AI commentator. Oh, what are we going to do?
The fact that the ai proudly claims to be able to win 95% when the score is almost tallied gives me chills. Damn. They won the clash and still continued to push through the second tower, meaning the ai surely knows their advantage and capabilities to push more.
This is fascinating. The AI is really pushing their heroes to the limit, all continuing to fight with barely any hp or mana because they are sure of the outcome, it goes to show how important it is to commit and what you can do with limited resources if you are efficient. Truly remarkable.
No man, that’s not how AI works. That AI was trained with 1 million games, 1 more game holds no impact in changing the AI performance. Once th AI is trained it goes on prediction mode, so it isn’t learning it s only using the information it already had to make predictions
Open AI always had the advantage, the synergy and coordination is unmatched. Plus AI continues to learn and evolve to be better and make less mistakes.
@@SeventhSolar that wasnt my point, Albert. i only mentioned that ai dont learn mid game cuz, op, while talking about the advantage the ai had, referenced ai continuing to learn as one of the advantages they had during the game... just clearing up common misconception ps. i learn mid game in league of legends... limit testing n wut not :p
@@jarvaniv3700 But it's true, the AI does continue to learn. So long as it plays games, it continues to evolve. They said nothing about learning mid-game.
@@charleszhou4091 sry for my bad english in advance, Yes ! We have to create a group of the best players in the world, like an allstars team. They should be treated like for example the squad of Real Madrid, the whole staff that comes with it, to provide the best conditions for the players. We need to put 100s Mio. € into that infrastructure. We need analytics from university people , mathematicians, physicians etc. and the Institutions for that. The players and the tactical staff have to work together with experts to analyze a way how to beat the open AI in dota (also IT-Experts, Military Experts, Philosophers etc.). If we really specialize on that, maybe it could be possible to beat AI, but like that obviously not. The scariest thing for me was how brilliant the AI calculated the remaining xp in battles. it just shows how quick in real time it can make calculations - and the second thing is of course the fact of its "real time true communication". There are no time delays in communication between the heroes, because all heroes are controlled by one AI, and the AI sees and process everything everytime everywhere. But I think there are ways to beat it. You have to do it somehow like Neo in Matrix, or John Connor in Terminator. I am asking myself if the openAI can already think about something that is outside the game - if its kind of aware of itself. Does it feels itself as an own identity or is it just a "In-Game-AI" that doesnt think "outside the box"
This is insane. AI predicted movement, blink jumps, skills, and positioning a lot of times. It is incredible how a machine can predict the actions of a human team based on data and equations alone. That sven always was jumping to a position where jerax was to about to jump 0.2 seconds after.
@UC-j_eIYCJNLzcNoYmQ7NLqA that is definitely cheat as well. See valve has banned using bind scripts. These AI must have physical limitations in their actions per minute
I dont see any smart or significant improvements except the leverage on playing the game as a computer program without any restriction on reflexes, and tailored every small situation to be optimal
Human does not need these bullshit development. It is waste of time and money. Do u know how much cost they masturbate for this shit? So like chess, this might be beneficial for teams to do practice, yet it has no other benefits. Who cares what these inhuman programs treat a game as their toy or simply manipulate it im the way they want. They are simply losing the strategic point of the game as well, thus reducing the excitement for humans
OpenAI had played 10000 years of games at this point. That's crazy. It had no idea that this was a world champion team. It was just another speck of dust
10:10 something interesting about openai is that it only practiced with other bots, so it expects 0.1 sec reaction time from enemies, although against a human he could basicly stun at the first milisecond of vision against es, it doesnt do that because he expects es to blink in that 0.1 miliseond aswell, a smart human could use that to their advantage :)
It's interesting to think that even the best in the world has so much potential for improvement, the AI just fills in all the spaces and thinks about everything at the same time. They calculate every single move the enemy's character can make, every direction they can go in, and they predict the most likely cause of action and that's how they can predict blinks and shit. It's impressive as hell. An OpenAI vs OpenAI match would be so interesting to watch.
What I find really interesting is, what the AI deems as optimal itemwise and herowise. Per example it seems to always finish starting items like bracer, wraith, null, which many players skip nowadays. And I believe, this was even before the 25 min buff? I think It's possible zo learn alot from what the AI seems to view as optimal. This was with a very limited hero Pool, but it deemed Sven and Gyro as good picks (which are currently both top meta picks, coincidence? Even tho this was before potential nerfs and buffs, so debatable.) Also that it picked Sniper first tells me there's alot of unused potential in that hero. Lastly the approach of spreading farm evenly to be deemed as optimal is interesting. Probably hard if to almost impossible to do as a human team, since they vonstantly switch roles. But it's very interesting it views the standard networthgap between supports and carries to be sub optimal.
one of the things to note about the bots that make them scary, is that, they're calculators right? All the game is to them is just a bunch of math. So like, you show a human a situation in the game, they need to run the math, make decisions, try and work out the best course of action. The AI sees this as math, and does it much faster If its going into a fight, its calculated that it will most certainly win. If it wont win, it wont go for it. A human looks at these scenarios and has to ask themselves, "Will I be able to handle this scenario? Will the odds work for me?" The AI looks at the same scenario and says "I know these odds. I know all my options. I know if this will work for me" Of course, the AI isnt perfect, there is a moment early on in the chat where the ai notes "I did not expect this outcome.", essentially saying "Hey, I did all the math, knew how the coin would flip, betted on both heads and tails, and it came up edge." But that also implies that, every other moment in the game where it is not saying that, is a moment where everything is going perfectly to plan
don't get the wrong idea guys. this version of openAI 5 has lots of time to improve their play. plus they also have learned from their mistake vs pain and chinese legends.
YOU GOT THE WRONG IDEA ! Open Ai team said that the Ai DOESN'T keep the data and improve in between these matches !! Go watch openai youtube channel game 1.
@@user-jd1hy9bg1d you're right, the AI doesnt use the data from games like these to learn. The AI plays itself, a much stronger player, in order to improve, and it does so thousands of times a day.
If they finish OpenAI 5 in years to come and make it available for the public to use, the next generation of Dota players will be absolute monsters compared to today's players.
@@daffamustafa well. I stopped playing 6 months ago. But, still have millions of players now, so I don't think you're right about that. Tho, I guess they abandoned OpenAI completely. If you think it's a dead game, what game do you think would make dota obsolete? MOBAs? League? 😂😂
@@cliffordsantillan6046 yeah but the think is those million players are returning players, and not many people will stick to play dota with the steep learning experience
Man I was gonna write "the coder team behind this AI must be certified genius" but then I later saw 15:34 and my mind exploded. Just give these f*cking geniuses their medals
DOTA for OpenAI was a reinforcement learning exercise that modified into reinforcement learning with human feedback (RLHF) that used for ChatGPT 3.5 and 4 .. they shifted from DOTA game to gaming up our jobs!
@@asmallphd9648 nah that's the whole point. It's 5 different instances of the same bot showing teamwork. That's why it's impressive. Remember when OpenAI first was created it was only for a 1v1.
Im just curiouse what was the main patern of most A.I stablisher like Open A.I? to create some thing they could predict or to look on the out come from product of the critical grate posibility?
@@TempuraRoll-of4hj Robots will never take over the world, it's just movie with fear mongering spiced in it. Once the AI getting more godlike, the government will do operation EMP in an instant.
THEY EVOLVED MONKAGIGA also so many awesome play from the AI like the gyro move to give vision to sniper,controlling courier for vision in a teamfight(i often see courier for vision in a gank but for team fight that a difficult one to do),and that shadow amulet play by cm is very good,and they even change role more than 1 time in a single game pogChamp
Well I believe the developers want to make it fair for humans and they didn't let the ai play heroes that control multiple units. You saw how they used that courier, imagine them using chen/enchantress😂
How does the open ai works. Since it was Ai, does they can see blindspot. since the ai can read the game file. i dont know, just asking cause im curious. How can i know the ai doesn't get additional damage/gold or automatically has TP cooldown when there was war.
They can not see through fog of war. There was a couple times they tried to stun but the OG players just got out of vision/range before the animation could play out. They may predict the movement very well but they are under the same limitations as us humans, within the dota engine.
I wish they would get to the point where all heroes and all items are allowed. And then face of the best 3-5 teams at this point in time. Although OG is the current TI-Champ, you have to admit that Secret and VP are the best teams around atm
Yeah they did spamming that in grand finals. Plus the aggresive tips. I believe that helped make their opponent lost focus and some patience at some points.
