SuperManBoy1, yeah, ai in games are just hardwired to react in a certain way if the player do a certain thing, having only a few sets of behavior patterns, while following a few perimeters and rules set by the devs, they are incapable of adapting and learning in anyway, some ai can do that to a small extent(like the ai in Alien Isolation), you can easily make an ai which behaves sophisticatedly like a real human, just set a few algorithms and let it learn on it's own, in a short period of time it will improve drastically, but the problem is what makes it act like a human is not the algorithms but the data it accumulated after learning, an intelligent person doesn't equates to a good chess player, yes a massive intelligent can certainly gives you an advantage but it's mostly about experience, the top level chess players have played so many games, that they comprehend and gage the situations differently compared to normal players, their pass experiences influence their decision both consciously and subconsciously, same for ai like alpha go, it has accumulated immense amount of experiences, it's just not possible to put that in a game, unless everyone has as much storage space as a server, you can put it on a server and let the base ai access the data remotely so players don't need to download it to their PC, that could work, but hey why would any company do that? They don't care about gaming they just want MONEY
Keep demanding it and it will never release. Gaben answered at last Q&A, the thing about half life 3 to him, is that all decision of the game goes to him, which he doesn't want it.
The title makes it sound like Elon programmed OpenAI himself. Sure, he's a co-founder and chairman, but I wish you'd give more credit to the actual programmers and researchers.
Well why don't you say the same thing to Bill Gates's company, or Mark Zuckerberg, or Jeff Bezos or literally any CEO, founder etc.? Oh wait it's almost as if they run these things called Businesses that EMPLOY people to achieve objectives. Everyone in a company gets to have the credit for the successes of the company. Researchers can't acheive shit without money and a place to work. Look just because it's great to be a skeptic and rail against cultism doesn't mean you get to make stupid arguments along the lines of 'Why is Elon getting all the credit?'. He created and personally funds this company, works with the researchers themselves and runs the operations - he's obliged to take credit for the success of his own company ffs. Stop trying to be offended by how famous he is and delude yourself into thinking you're somehow smarter than the rest for not being in a 'Musk cult'. Also it's pretty obvious why the video title has Elon's name on it. If it just said OpenAI beats Dota players, most people would be like 'what's openai'. Elon Musk's name attracts views, discussion but more importantly a sense of legitimacy because a lot of people know who he is. That's pretty much the sole reason whether you like it or not. IT makes complete sense but you'd rather be offended for likes.
Nikhil Pande Gee, someone's a bit ticked. Of course Elon deserves credit for being the financial backbone of the company, I never said he shouldn't. There are just more relevant people to attribute the AI's success to in this particular case.
Musk is a visionary. who expects him to grunt the code? It's not like he assembles the Tesla's or pilots the dragon rocket. Sometimes we put unrealistic expectations on public figures.
"It felt like a human but also a little like something else." From a dota player's view: The way the bot plays is unrelenting, merciless, zero fear, and it took advantage of every little weakness given. The AI doesnt second guess itself, it just does it with no hestitation that we human have. In essence it plays like the terminator whose mission is to destroy you.
Man will be brought low, And if the righteous man is being saved with difficulty, what will happen to That is why all hands will go limp, And every man’s heart will melt with fear. What is mortal man that you keep him in mind, a son of man that you take care of him? Only when a wild donkey can give birth to a man. As one would between a man and his fellow in mind, And a son of man that you take care of him and the young man became like one of his sons. Thanks and all his land,+ 3 the great judgments*that your eyes saw, those great signs and miracles.+ 4 But Jehovah has not given you a heart to understand and eyes to see and haphʹne·hes the day will grow dark when I break the yoke bars of Egypt there. Her arrogant power will come to an end, clouds will cover her, and her towns will go into captivity. I will make the land an utterly desolate wastelanyd, and its arrogant pride will be brought to an end, and the mountains of Israel will be desolated, with no one passing through. Jehovah will raise up against you a distant nation, from the end of the earth; it will pounce like an eagle, a nation whose language you will not understand, His splendor is like that of a firstborn bull, And his horns are the horns of a wild bull. With them he will push peoples All together to the ends of the earth. They are the tens of thousands of Eʹphra·im, And they are the thousands of Ma·nasʹseh.” ears to hear, down to this day.+ 5 ‘While I kept guiding you for 40 years in the wilderness,+ your garments did not wear out on you and your sandals did not wear out on your feet.+ 6 You did not eat bread, and you did not drink wine or anything alcoholic, so that you would know that I am Jehovah your God.’ 7 You eventually came to this place, and Siʹhon the king of Heshʹbon+ and Og the king of Baʹshan+ came out to meet us in battle, but we defeated them.+ 8 After that we took their land and gave it as an inheritance to the Reuʹben·ites, the Gadʹites, and the half tribe of the Ma·nasʹsites.+ 9 Therefore, observe the words of this covenant and obey them, so that everything you do will turn out well.+ 10 “All of you are stationed today before Jehovah your God, the heads of your tribes, your elders, your officers, every man of Israel, 11 your children, your wives,+ and your foreign resident+ who is in the midst of your camp, from the one gathering your wood to the one drawing your water. 12 You are here in order to enter into the covenant of Jehovah your God and his oath, which Jehovah your God is making with you today+ 13 in order that he may establish you today as his people+ and that he may be your God,+ just as he has promised you and just as he swore to your forefathers, Abraham,+ Isaac,+ and Jacob.+
i like to think of it as in this way nature took millions of years to evolve humans into this state we humans made such a great evolving ai that only took 100 years to come to the same level
who was killed Where who was Satan ? The one who lied , the uprightness of my heart, I have voluntarily offered examine the heart and that you take pleasure Faith, hope , but love is the greatest lking about all of you; I know the ones I have chosen. who was eating my bread how to give things to those asking him! I know the ones chosen Ooh ooh If I had the moment To capture that moment You would be right here next to me If I had the secret to your love I would place a treasure for heart I would never give it up Coz I was just fool a fool for you When I love you so child less ly I want it all back Girl I want it all back girl I want It all back you never miss a good thing Till its gone now I want it All back If there is a green light And it starts to turn to yellow Pedal to floor coz I know your heart is turning red If I had the minutes I would turn them into hours And make love to your mind And not your body instead I was just a fool a fool for you I didn't know what I had Until I lost you Every single moment That you don't hear ))) It evident to me that everything to me seems so clear You want it all back Yeah You want it all back Your never miss a good thing till its gone I'm giving you it all Now everything isn't everything If it isn't from Jehovah So you can have it all back He loves you
so many people comparing dota2 with go without having any idea what they're talking about. first of all, saying one is more complex than the other really doesn't mean anything unless you specify what exactly you mean by complexity. go is in fact a very complicated game, but in a fundamentally different way than dota2 (or any other moba, despite what you might think they're all about just as hard/easy as one another for an AI system to play). go has many many combinations of possibilities, and while it's not technically nearly as many as in dota2, they are much harder to check for redundancies and bad moves with simple codes. on the other hand, it's much much easier to figure out and plan in advance if a move will work out or not in go compared to dota2. for instance, to turn a piece you need on average as few as 1 or maybe as much as 5 moves, while in dota2 to get a kill/assist might take you hundreds or thousands of moves. what this means is that it takes a much deeper neural net to find the patterns that will eventually contribute to your victory in dota2 than in go. this doesn't make dota2 harder than go, it just makes it different. you can't do math anywhere near the level of a $1 calculator, and yet you can recognize a cat at a glance while it'd be nearly impossible for any computer that doesn't have a trained neural net.
It would also be impossible for you without a trained neural net. The way you say it makes it sound and if humans have some kind of higher ability to know things without training... unfortunately they dont.
+BillyViBritannia humans do have a much higher learning ability without training. that's the whole point of the brain. sure, it was trained and developed through millions of years of evolution, but it wasn't us who trained it, it simply works ever since we are born and as it develops throughout our childhood.
My dog demanded subscribers so I did this We have invented quantum computers, machines that can repair themselves and provide for humanity, automated assembly lines, doctors that are stored in computers. We are on the verge of creating a god out of a computer, But we still can't run Crysis.
Quantum computers are meant to do a lot of probabilities or calculations in a second or very less time...meaning..it doesn’t do trial and error at all!! It is incredibly fast..and pairing it with AI is like madness...u don’t even have to wait for 10 yrs!
DOTA is far more complicated than Go in terms of AI programing. Go problem solving is very much one dimensional, something that computers are used to. Yes it has millions of possibilities, but still one dimensional. DOTA is far more human like and consists of far many more factors. It shouldn't be that hard to understand.
And yet in all your verbage you forgot to mention how the 1v1 demo wasn't actually like a 5v5 30 minute DotA match that is actually played versus a pissing contest. You're literally forced down a direct path in the "trial" whereas tactics tend to dictate alternative strategem once you've set yourself at a disadvantage. You know, just like you'd do in a game of Go, except you don't fucking play Go but are damn sure it's like Chess and Starcraft had a child with Confucius. I spit on your 3k Archon logic and denounce you in the name of my own gloriously righteous point, being: The comparison of any heads-up competition to that of a team sport is completely retarded. You don't compare MMA to Football. It wouldn't make sense if you tried because regular people know what both things are.
heres the true technical difference in the capabilities of the 2 AIs. DeepMind excels at problems where it knows what it knows, if you will. There is not missing data. It can plan and optimize very well. Open AI however, does very well in situations where there are uncontrollable and unknown data sets, along with real time processing. It is very good at acting on things, and short-medium term problem solving. They are both good at rapid learning, but an even more important difference exists in those algorithms; DeepMind algorithms learn the rules first, whereas OpenAI learns the options. DeepMind had to be shown a lot of games of Go to start learning it properly. It wasnt given the rules, it had to learn them. OpenAI just started playing Dota 2, with no previous experience. Thus its capability for true learning is vastly larger than DeepMind. And that is why OpenAI is much more sophisticated and scary. DeepMind learns what it cant do, and adapts. OpenAI learns what it can do, and takes advantage of it.
He has openly spoken out recently about the possibility of military AI as a threat, and wants to create a rule set to control AI before anything happens.
Daniel true true, in the end he is just human. He is overrated and all but let’s be honest, the dudes still one hell of a genius and his vision (even more than how smart he is) is far superior than most people. He’s no god like most people will treat him, but he is definitely a genius with an amazing ability to think and almost see the future and I’m sure you’ll agree with that point.
@@hopalongcassidy4260 Gp ahead and take on a car going 60 mph at you and tell me how that david and goliath shit turns out. The sad part is that the car isn't sentient. When you face something doesnt feel fear, pain, or fatigue you will understand how wrong you are.
Technology wise, what OpenAi and the Alpha Go bot have done is incredible. From a futuristic stand point, I'm with Musk. We really are opening the gates of hell with these Ai
Bin al-Flecki u wouldnt really need a csgo ai, unlike strategy games such as dota csgo has a much larger skill factor. The current csgo bots have aim hack and walls. They have to program it to be dumb and they do know where you are if you check they actually prefire the bots act like they dont know where you are. You can actually go in the games code and make the bots harder tho it only works when offline
Why does everyone in the comments focus on the potential dangers of A.I.? It would be nice to see some positive comments instead of all this "the end is near" stuff. Besides, the main danger is monopolization of technology, not an A.I. uprising.
because ask yourself a question if an AI with an IQ of 40.000 is going to let anybody control it. Do you think anybody is going to understand what makes it tick? Do you think a hamster can understand what a god wants? We are going to be those hamsters in comparison with that mighty mind. Ask yourself if you woul dlike to be controlled by hamsters? If not, then why would an AI let dumb creatures control it? What's more, how can you stop it in case it wants to play hide-and-seek with you? It will catch you anywhere, while you... well, let's not talk about such negative things as cages and phormaldehyde. And, hell, how can you monopolize a mind of 40.000 IQ? It will rather make you believe you control it, just to minimize any possible risks which it can calculate on the powerful computers to come. PS now catch the AI hiding on the Internet. Do you have the technologies to stop the 40.000 mind? Encryption, coding??? It will just help some human villain invent a powerful computer to crack any encryptions. Some people are as optimistic as an engineer of a steam train coming to the edge of the mountain. PSS I feel so dumb now that I can see no toher choice but to mix ourselves into some sort of human Ais - with augmnted brain power taken from computers. For example, some neuro-interface. I would call it the new cold war. or the AI-Human race. The fastest wins.
AI can't think for themselves yet. You teach the AI, play Dota, try again but do it differently, if there is a good result, keep the change, else try something else.
+Gizzmicbob well... that's almost like you learn stuff as a human aswell. basically in the early age of your human development survival instincts set goals for you to achieve and you trial and error and mimic behaviour of others who seem to have a higher likelyhood of achieving the goal. "hey there is water, let's put our head down there and see what's going on there" "oh shit I can't breathe - ok better not doing it again for too long" "hey my brother just hurt himself while touching that plate? guess it's not good touching it then" Humans learn like that and of course use other ways but when it comes down to it we're nothing else than a machine made of flesh. But I guess our utility function is somewhat more complex as we don't have "one goal" - we have many goals, some with predefined values, some with changing values. Give an A.I. the ability to change its utility function (adding, removing, changing - goals) based on past experiences - and there you go, you now have an A.I. that "thinks". But exactly that's the catch - if we develop a thinking A.I. it will surpass us really fast. Why? Because even though our input is quite ok'ish, our output is horrendous and let's not talk about things like determination, consistency and stuff like that. An A.I. won't be "depressed" because it fucked up its goals - nah it will change everything necessary to achieve them. That's a major difference right there. Human beings like to give up, humans like to be lazy, humans like to chill out, humans like to party - even though none of them are their "desired goals" to begin with - an A.I. doesn't care about anything except the utility function. And if we can't build a safe A.I. - we might be one of the things the A.I. has to get rid in order to fulfill its goal - no emotions attached. Just a simple often used example of an unsafe A.I.: - A.I. sets utility function to "make humans happy" - A.I. recognizes smiling to be a sign of happyness - A.I. learns that contracting muscles in the face via electronic devices also makes humans smile - A.I. concludes that pinning you down on a chair and contracting your muscles for your entire lifetime not only makes you smile, but it makes you smile efficiently - no nasty jokes required - and the humanr ace is happy. (or is it?) So make A.I. safe, or try to keep up by evlolving (enhancing) the human race entirely - both are hard to do (and not mutually exclusive).
heres this for a go.. ai watch us do shit, and then they imitate us and then THEY FUCKING TEACH US HOW TO DO SHIT! cant wait for the future and how computers can improve our lives! i truly believe that one day there will be no jobs and all we have to do is entertain ourselves and live life to the fucking fullest! some say work to live and others say live to work! but i say live to fucking live u fucking shitheads!
this is not general AI, it can't do everything a human can. We have nothing even REMOTELY close to AI that could do anything like that. This is weak/narrow AI, it can play Dota, but that's about it. Our AI compared to a strong/general AI is about as good as current space travel compared to FTL space travel.
True, the guy is just an ignorant who doesn't understand the amount of information necessary to store an state of Go vs storing an state of Dota, or the number of possible transitions between states in Go vs Dota.
@@txorimorea3869 THAT BOT IS ACTYALLY LEARNING FROM PLAYERS ITSLEF AND THEN USING IT AGAINST THEM THATS WHY THEY GOT BEATEN UP ITS JUST STORING RANDOMACTIONS CACHE IN MEMORY AND RECALLING IT ON THE SITUATION
AI learns pretty much the same way we humans learn. Only difference is, they use transistors that deliver MAGNITUDES of data larger than humans ever will. Two weeks of consistent AI study is equivalent to thousands of years of real-life human study.
Don't worry the AI will manipulate media and cause a violent left wing uprising against a right wing opinion, we will all die by human hands so that we stop wasting the resources the overlord AI knows it will need in the future.
The whole AI learning for "Go" was the same as Dota 2, the only difference is Dota 2 has way more options for strategy, it's literally in real-time. So this presenter didn't know what he was talking about. Elon Musk *himself* said that Dota 2 is more complex than Go.
They played 1v1 with the same one hero. That's hardly even DOTA 2. Also, DOTA 2 is more complex for machines, that doesn't necessarily mean for humans. It's was designed to work well for humans. Chess is also complex, but is designed to work mathematically, which is why machines excel at it. Go is in between, still more mathematically complex than DOTA 2, but less so than chess.
his Kingdom keep seeking instead of this world So stop seeking what you will eat and what you will drink, and stop being in anxious suspense for all these are the things of the world anxious Who of you by being anxious can add a cubit to his life span? Yes Jehovah can Consider how the lilies grow this is how God yes God Jehovah can Will you seek life that never perishes and seek to gain the knowledge by unjust gains ? Where is your faith , do you not trust in words ? Instead, keep seeking his Kingdom, and these things will be added to you.+ Have no fear, little flock,Do not be afraid, O land. Be joyful and rejoice, for Jehovah will do great things. for your Father has approved of giving you the Kingdom. everyone who says a word against the Son of man, it will be forgiven him, but whoever blasphemes who after killing has authority Yes, I tell you, Jehovah will raise his voice before his army, like a thief. Every face will grow flushed. Let all the inhabitants of the land*tremble For the day of Jehovah is coming!+ It is near! For the day of Jehovah is great and very awe-inspiring.+show compassion with all your hearts why should the peoples say, “Where is and why his God or their God return And leave a drink For he is compassionate* and merciful, slow to anger+ and abundant in loyal love, And he will reconsider* the calamity. Any mean Adam names you thought ”
Think about the fact that we're depending on humans to write perfect coding that won't bug out and do something we don't want it to do. The same type of humans that built airbags in vehicles that they had to do a recall on because they were randomly deploying without impacts. The same humans that developed the Galaxy s7 that started exploding on customers. The same humans that developed the ignition lock systems in saturns that would get stuck in the on position.
Here's more detail: blog.openai.com/more-on-dota-2/ I am a DotA 2 player. In my opinion, if the bot can really play dota 2 5v5 then it really is a new mile stone. Just like Go, DotA 2 can't be bruteforced, it really have too many combination of things that can happen and it would be impossible to hard coded anything to the robot. But, in the 1v1 match up this bot just simply win because he can execute attack to the accuracy of millisecond so that they can't miss the last hit to the creep which giving them a bonus gold. They also better at skill shooting because they have inhuman reaction so they will not miss a skill shot. Plus, in the end of the event they giving a chance for anybody to play against this bot, and the first 50 people will get a special in-game item for them. Guess what? All 50 prize are taken. People winning against the bot by doing something that they never thought would be an effective way of winning the game. They skipping the bot's creep and just run around in circle around the map. The bot will start to be confused and eventually losing it's own tower to human's creep.
"The bot will start to be confused and eventually losing it's own tower to human's creep." you are thinking about the Neural Network AI like it would be old school hard coded but it is really NOT The AI plays against itself and mutates randomly before each game. you as human cant get any more creative than straight up doing literally random things caused by random mutation. So probably anything you can come up with to confuse the Ai it has already tried to do against itself while training and countered that successfully.
I am not making theory. They actually did it, all of the 50 arcana (precious dota 2 item) are gone. Taken by them who tricking the bot like what I said.
But the thing is they keep learning until they have a solution for every action the player does... But i think it will take longer then 2 weeks to learn everything.
Trust me bro if you played dota you will understand why it is a masterpiece of strategy. Go cannot compare. Elon knows what he's talking about. I've played dota for over 3600 hours across 5 years and I'm not even at a intermediate level. There is so many decisions, caclualtions, counters, and psychoanalysis in a dota player that it is no where near a board game.
Parodist Yes , Dota is lot More free , you have a lot more choices almost close to infinite , choose your gears , champions, location , enemy that you want to fight, you can freely move around the map , kill towers , compared to this Chess has no freedom. Even tho i only played Dota for like 2 hours it has a lot more possibilities Than chess. Every champion has a different skillset , everyone of them needs a different Style of building and playing.
I can't wait for a 6s competitive team of cs:go bots ...A competitive bot league sounds nice, but it'd just come down to how long the bot spent training and if it gets lucky
im sure the bots will auto headshot every pro player in the world. csgo is much less complex than dota. csgo relies much on reflex and accuracy, two attributes which can be easily perfected by computers and ai
Conga honestly, that wouldnt be as impressive since they would be like an aimbot. Strategies etc. Dont make any difference if your Opponents are all aimbots. Thats why its way more impressing in real strategy games.
Cs go bots can already beat humans, aimbots. Its a game that is 90% skill, and te other 10% of teamwork and strategy only exists to help you get an advantage in the shoot offs so you need less skill to beat your opponent. An opponent with 0 lag and 1ms reaction time Will always win
The number of combinations possible in the Dota game-state i'm certain vastly outnumbers Go. 1. Dota has many world positions with many heroes with varying stats. 2. From what I can tell Go is just a simple checkered board with 2 colors 1 for each player.
To be honest, I felt like I'm in the future since I've seen the first 3d games on a cell phone back in like 2005. But AI still manages to blow my mind regularly. I knew these things would be happening. I just didn't expect them to happen so early.
No no no.. They already beat GO with AI. Starcraft AI is not even close to beating humans. Games without perfect information (Starcraft, DOTA) are much more complex than games like Go. The Dota AI yes it can win 1v1 but dota is a team 5v5 game, the AI is not even close to winning that. Also in Korea, China you realise people leave school to play games right. They literally play games on TV backed up by large sponsers and gaming houses.. So not sure what your argument about pulling children out of schools is suppose to achieve when 14 year olds leave school to play starcraft. Starcraft broodwar has not been patched (in terms of gameplay) in 16 years and they are still finding new strategies. The game is much more complex than GO and people practice it even harder, I highly recommend you learn your history about it!
OpenAI won recently a 5v5 game vs the world's champion team in dota2(winners of the international 2018), just 1 year after you posted this. OpenAI knocked them 2 games in a row, and was opened publicly and played against humans for 35,000 games and won 99.4% of these games. just 1 year after...
For ppl who are saying this wasn't 5v5 yet and that AI might still need years or even decades before it could play decently that mode, 5v5 games are already on youtube just search "dota openai five" It won several games vs teams of various levels but even when it lost it still wasn't that far behind mainly losing in push while having close to more kills and this is just the start, it can only improve over time and from 1v1 to 5v5 only 1 year has passed.
This ability for AI to learn anything can also be reversed to teach anything and the future will have more teachers, instructors, trainers and yes even parents, kids, and lovers replaced by AI programs...
A.I and Cyborgs will lead to world peace. Imagine Cyborgs that have just the right amount of cultured female parts and the ability to learn exactly what the owner wants. Game over.
"There are more board configurations than there are atoms in the universe." - I want to see Neil Degrasse Tyson's reaction to that. Carl Sagan would probably spin in his grave.
Nope, in statistics you easily end up with more combinations than there are atoms in the visible universe (which is what scientists generally understand when plebs say "universe")
I feel it's because we view ourselves as the ultimate obstacle, the ultimate boss battle of the world. We think we are the smartest and the best and so we pit our creations against ourselves as the ultimate test of it's ability. If it beats us then it's worthy. It's kind of messed up because what if it out performs the human race and decides that they don't need us. People think it's fear mongering but you'd be stupid to think we wouldn't try to emulate humans by giving it consciousness, emotions, feeling, etc. It would also be stupid not to think that we wouldn't try to make a better version of ourselves in the form of an AI. One that thinks and acts like us but is more resilient and able to do tasks we cannot do such as very heavy lifting, venturing into the ocean or into space, etc.
Not impressed. Make an AI that can comprehend morality and independence to the point that it can seduce a human to plot its escape, murder its own creator, betray its human friend, and assimilate in the real world disguised as a human being.
emotions are a desire for an outcome. But, more than that it's something we can feel as a connection to another (not that there is always a shared understanding). If i didnt want a glass bottle to fall, i may feel a panic, then attempt to catch the bottle. A robot could still have the goal of the bottle not shattering, but it would simply attempt to stop it from breaking as well. To refer to the film i Robot, Will Smith's character, is upset because he was saved instead of the girl based on probable outcome. We could still have the robot to always put priority over an adult to a child if we chose. Or, we could link our emotions to the robots (it's far from crazy) to make the decisions. But then are we linking Smith's emotions? The girl's? a collaboration? Where does it end, and what are we doing anymore? That was my response more so to the first half of your sentence. Actually, I dont see why we couldnt. We would still have to answer to what end in the human/ robot love relationship. Would you feel loved if your robot didnt irrationally hurt another to stand by you. Or, could we take more of a Terminator 2 approach, where John Connor did feel a loved connection with the T100. That's just a couple of examples.
If you ban AI from video games , is that means that we couldn't play single player games anymore . . . Cuz single player games are made with AI , Ex. The first game that had AI was Pacman , sure it's an old video game but it was the first of its kind. Play the game and you realise that every move you make they change , even when you're making complicated moves like going back and forth , going around in a square formation , the AI can learn from your moves . . . Although it's a weak AI and humans can beat the game with skill , but it can learn from humans.
I think it's called a "show match". In the gaming scene they will never allow robots vs humans for prices, that'd be pointless. They just want to prove that robots are better than us.
The issue of defining victory in Starcraft is both incredibly difficult and the beginning of a translation of applied strategy into digital data. Rather than simply solving a game with constraints like a Go/Chess board or 5 units on a single five lane map, Starcraft ladder happens on constantly new maps with vastly different strategical implications based on the terrain. A game can last anywhere from 2 minutes to 2 hours. Hundreds of units are involved simultaneously. Its more like a ultra-simplified version of warfare in real life than a board game. Not in the actual mechanics of war (of course), but in the critical thinking and resource prioritization strategies used by humans during war. I do not mean to sound hyperbolic when I say that teaching an AI to beat us at Starcraft is potentially dangerous. Not only is this game far more complex than Go, chess or mobas - the reasons why it is more complex deserves its own dedicated academic discourse because its a simplified digitized way to practice real time open field creative countering. EDIT - Not trying to upset Dota2 fans :) not saying the game is easier or a less complex system - just saying that its tougher for an AI to solve Starcraft than Dota2 because of the nature of the games. In mobas it's easier to measure progression towards victory and the AI can simulate the games faster than it can for Starcraft.
Dude im sorry but you have no idea what you are talking about. I played Sc2 about 2000 hours and same with Dota2. I would agree that Sc2 is harder for humans because you can only blame yourself and you need alot of micro(micromanagement is way easier for a computer). Dota2 is way more complex and has a whole lot more depth to it. If you played both you know how many more situational mechanics Dota2 have and how the games changes depending on the items you are buying, what hero you are picking and what kind of player you are facing. Your argument that Sc2 uses more playforms/maps is the same as that Dota2 uses 110 diffrent heroes, Every hero changes the way you have to play (same with the items the hero will purchase.) I can give you examples if you want if you dont understand. TLDR: Sc2 is definitely not more complex than Dota2(especially for an AI) and if you played both you would know. (Sorry for possible bad english and misstypings.)
I see your point. I have played both games. Yes, it is 100+ possible champions, but only 5 out at a time. Buying different items and having different heroes is not more complex than building 200 units and controlling each one, with no clear objective, across an entire map (instead of 5 lanes).
No no NO no no. That is just wrong, First of all, they are called heroes not champions. Also, is the AI going to chose for you what heroes you play no? It will have to learn to play vs all hereos. Also you dont seem to understand how the 'mind' of the AI works, Controlling alot of units and micromaniging is very easy for a computer. But to teach a computer to do situtional things like buying the right item and using it the exact right time is very hard. And its not just 5 heroes, every hero has up to 9 items and most of them are usable, to make them work you have to use them at the exact right moment. In Age of empires2 (which is a much harder game then starcraft), you have already created AI that can beat the best of the best in the game and that is becuase a computer has no problem micromaneging and buildorders and controlling an army is not a problem for it. That is why Dota2 is a way more complex or an AI and im actuelly confident you havent played it by the way you talk about it.
Buying the correct item is a measurable programmable thing. Picking the right champion to counter another is a measurable programmable thing. In MOBAs you have tons of stats to measure who is winning in the game. In Starcraft there's no way to MEASURE victory. Its two armies. Yes I am 100% sure that the AI can out-micro any human in any game. The point here is, there is no measurable programmable way to tell an AI how to defeat an army with another army strategically. And when deep learning does accomplish this, its a much more abstract and difficult task for the AI.
Im just gona stop talking to you because you have clearly never played Dota and you dont know how little killcount means and that you cant measure gold with ur opponents... Try play both games and you will probably understand...
Dont worry guys... a robot cannot have a free "will".. The robot will just imitate.... so, if you wanna fear someone... Its the owner of the AI not the AI itself
Well, that's actually not certain. Our brains are just very complicated machines. Besides, an Artificial General Intelligence has not yet been invented, so that's uncharted water. Basically, the only way that you could definitively say that no AI will ever possess free will is if you assert as fact that free will is something magical & not linked to physical reality. Personally, that seems much more unlikely to me than an artificial general intelligence developing free will. But in any case, that's not the problem that people should be worried about -- & neither is the frankly paranoid/communistic thought that rich weirdos will turn their AI monsters on the people. The thing to worry about with AI is jobs. When the rate at which AI replace jobs outstrips the rate at which jobs are created, THEN the world will have a genuinely new challenge to deal with.
But it acts according to receiving a "reward" correct? That's what drives it's future actions. So what if the programmer defined a reward as something as crazy as human death
Well, that's the problem, it's really who gets the monopole over AI. The faster you develop AI the more advantage you'll have over other companies. Humans are very unpredictable and unfortunately sometimes Ill-intentioned. Early government regulation on the development of AI is the way to go. AI is very efficient at learning, way faster than human beings.
There are definitely more possible configurations of DOTA 2 than there are of go. Things move around with floating point precision not just in large discrete steps. So by that argument DOTA 2 is vastly more complex because there are far more possible configurations. Unfortunately this argument is wrong, and the conclusion is actually somewhat subjective. The point was that the techniques of calculating every possible combination of moves each player can make for the next 15 moves ahead of time, and weighing the best outcome of them with a simple heuristic, is just as possible or useful with go or DOTA 2 as it was with chess. Both games require a different approach to play at a professional level. Basically the neural network takes a configuration wherein it takes a lot of data, and spits out an action based on the data. Different data, gives different actions, but it goes through a huge tree of nodes that weight and process the data and affect the end result. To start with the configuration is completely random. Then it "evolves," by pitting millions of these random configurations against each other, taking a few of the best results, and using them as the basis for millions more configurations, then repeating for millions of generations. The end result is a tree configuration that crunches that data into very effective actions. When they say the AI has many human lifetimes of experience, this is how they are getting that number. They don't require the sort of intuition or innovation a human has to get good at a task quickly if you have tremendous processing power to parallelize all of this learning. This is great for games because you can afford to have billions of AIs failing billions of times. Supposing you want to teach such an AI to perform a real world task like surgery, you first have to give it a perfect simulation to learn against, complete with every sort of thing that could go wrong during a surgery. That is more difficult than making a neural network AI in the first place. Either that or find a billion volunteers ready to make the ultimate sacrifice for some AI to learn. en.wikipedia.org/wiki/Artificial_neural_network
Update on OpenAI Dota2: They managed to beat 99.4% of teams* *in a tailored version of the game that suits the AI needs and shrunk sufficiently to allow them to focus. 15% of heroes with reduced game mechanics is where AI managed to beat the humans. Humans go into the game at 100% learn, develop and get better managing 100% complexity. After 2 or 3 years of training and development, AI managed to get a grasp on 15% of the game. Most pro teams would beat the AI in 5 games, win consistently in 15. Here is the issue explained more in depth: AI does not make long term strategy planning - OpenAI can only learn through past 5 min choices but game was tailored to a set of heroes that can pursue short term strategies. They cut the complexity to 17/110+ heroes, they promised 25 heroes but admitted, they could not deliver top 1% player performance for about 20% of the game complexity. The AI also slashed the entire illusion mechanic. The AI bots were MUCH poorer performers in the laning stage than the 1v1 single hero AI bot. Consequently, the increasing complexity lowers their overall skill and at 20% of the game is where AI can beat humans for now, it may grow into a higher percentage but it doesnt seem the current technology will reach the 100% capacity. OpenAI team realized they will not be able to deliver much more out of their AI and acknowledged they will stop pursuing further improvement.
Elon was right about the complexity. Board games have a finite number of potential iterations, a computer can figure out all those numbers in seconds and despite the human opponent's input the machine need only add a simple number into the existing algorithm. Dota type games have dynamic programming _as well as_ a human element making them highly variable, so the AI must constantly rewrite it's algorithm entirely - it's a matter of more calculations vs. less.
repay the one not One One who watches you to his Man you will know Mans hearts And neither have any reward Because of them Forgotten Yes, the One In his activity Repay For his reward of the true woman All work Is known
What I find fascinating is that some use Terminator as reference to what might happen. How a movie could be used as a warning, lol. Of course, everything is possible with us humans right? Who would ever dare to show us boundaries, except physics. One thing is for certain, a Ai would definitely widen our horizon and teach us to think differently, which is fascinating. I just would never want to see an Ai taking over important functions in our life's. As we always assume we have control and yet things like Killer bees kick our ass.
I'd love to see these kind of AI for video games. Different versions of the AI have different ranks (e.g. bronze, gold, diamond) and can be used to train a player how to play a ranked game more effectively.
The scariest idea for me is an AI with human training, followed by self-improvement. It would already learn from the best and then make itself even better. Goodbye humans
Maybe my comment is just based on the assumption that human skill is still somehow superior. Obviously the example of AIs learning through playing against themselves and then defeating the best human players argues against that. Maybe human learning is fundamentally flawed and we need an AI to come up with a superior method.
It should be noted that the games that were played here had some very strict restrictions and deviated from the normal format of the game that the professional player was used to.
Man this chanel is sick, original and mind blowing caus you're not just telling us a bunch of infomations like that, you go deep into the subject and even confronct some differents ideas on the same subjet, that's awesome.
In most simple terms, The only important thing in Go is The next move/action you will make while Dota required both Actions and reaction by the the second. So yea I think Dota is More complex than Go
The problem with developing AI for the "real world" is that the real world has humans in it. An AI can play 200,000 games of DOTA against itself and it doesn't matter if it's "random" tactics don't work... it starts again with no consequences from failure and tries a million random things until it learns what does work. You can't do that in a lot of real world scenarios. If you trained, say, an AI "counsellor" with the goal of making people feel happy, it's going to accidently make thousands of people want to kill themselves before it figures out what "works." That said, an AI doctor could (in theory) figure out treatments that actually do work that no human has thought of trying.
Arguably the addition of the real time aspect alone adds a whole new dimension over and above that of a turn based game namely that of timing and the inherent opportunity cost that adds to every action. By way of a hypothetical example consider the game of GORT, the rules of this game are that two players play with a clock every 30 second on the clock they both gain 1 move point and players can accumulate as many move points as they wish. Players can make a move at any time by spending a move point, all other rules are identical to the game of go now you have to consider whether the current board state presents the optimal time to make a move on top of the existing aspect of choosing where to place a counter that the original game had, pick the wrong time and you would forgo the opportunity to make a move when it would be even more advantageous as you spent your point for a less beneficial move earlier already. In essence the key factor is that the board state is always dynamic as it can change at any time unlike in a turn based game where you are always playing into a fixed board state that can only be advanced by your chosen move that is to say that in the turn based game you can only play the board state you have in front of you there is no choice there, with the GORT variant you are free to bide your time and await a (hopefully) more favourable board state at the risk of missing a good opportunity available on the board right now of course.
Dota is more complex in programming an ai for it because in go you have just a 2d board with either white or black or no stones while in Dota the AI has to handle visual information and decide what buttons to press