8 ms·

Very impressed with the bots. The main strategy I see from the bots that is different from pro meta is how they spam abilities aggressively. Now the main limi

by kmnc 8y ago

Very impressed with the bots. The main strategy I see from the bots that is different from pro meta is how they spam abilities aggressively. Now the main limitation is 5 unkillable couriers which enable this. Watching, I do feel like given more practice humans would beat this version of the ai easily. It looks exploitable, and still having many rules limitations it has a ways to go. It will win in the end though, that much has become obvious.

hackandtrip 8y ago
It would be really interesting to see what bots will do once humans will try to exploit this strategy, like adapting to the lane movement they always do after winning a lane etc, humans did that once today but it may be used better. Excited for the TI to see less limitation, a better shop system (lot of money throw away) and maybe better ward.
- 21 8y ago
  The next AI will always be much better (see AlphaGo progression). Humans, not so much, as in all top-level competitions, human abilities improve minimally at the top, because we have millions of humans competing against each other until the plateau of human performance is reached. Then you can push that a bit more with drugs (see doping in sports). And after that, you are pretty much done. So it's only a matter of time and effort until AIs are fully unbeatable.
  - douglaswlance 8y ago
    Humans tend to progress by learning from each other. The AI will teach us.
    - asdfasgasdgasdg 8y ago
      Whether the AI teaches us or not, once it surpasses us, we will not catch up. This is the case for every game AI that has ever surpassed human performance so far, and there's no reason to expect that this will be different in the case of DotA.
      - ethbro 8y ago
        > game AI that has ever surpassed human performance so far Am I missing something, or does that set consist of Checkers, Chess, and Go so far? (presumably with analogous misc games of comparable complexity) Discounting the reaction time wins, I'd say the sample size is too limited to generalize to eventual AI behavior in more complex / open-ended games. Extrapolation was the cause of the last AI winter.
        tialaramex 8y ago
        Poker. Although Heads Up Limit Texas Hold 'Em has been essentially solved the No Limit variant (in which you actually decide how much to bet) is very far from solved. An AI beat the best players in the world anyway. It plays crazy, but it plays crazy in a way they weren't able to exploit at all. Doug Polk (also one of the best in the world in this narrow specialty) did some funny videos but the humans got crushed. So that's a pretty different game, it's got a big luck factor and has asymmetrical information and still the AI just kept getting better and the humans... didn't
        TangoTrotFox 8y ago
        I think you're looking at things with a sort of hindsight bias. Victory at chess was at one time considered to be the indicator of the emergence of true 'intelligence' in computing. The reason is that it's an extremely open, creative, and strategic game spattered with a minefield of tactical nuance. Nobody, human or computer, is getting even remotely close to scratching the depth of the game from a numeric point of view. There are some specifics on the numeric complexity of the game here [1]. The reason I point out the unfathomable numeric complexity is that it makes the games, from the perspective of an AI, effectively infinite. AIs are calculating, but to an extremely superficial degree relative to the depth of the game. E.g. - when a chess program says it's calculated to 30 ply (15 moves for both sides) what it really says is that it's seen up to 15 moves deep after intentionally ignoring or pruning 99.9999999999% of moves which it thinks probably aren't good -- something it still often gets wrong, but its 'understanding' of what is 'not wrong' is strong enough that it still results in a phenomenally strong level of play, compared to humans. There's no doubt that perfect play in chess would still go 1 billion - 0 against something like AlphaZero. So what matters is not the number of decisions to be made but the individual complexity of the decisions to be made. And in most games we consider complex the individual decisions are not really that complex, and complex systems can often be broken down into very simple games. For instance a great example of this is a 4x game. Taken as a whole they seem complex, but they're really just a large number of relatively simple components that are mostly independent. E.g. - Given this state, where do you explore next? Given this state, what do you research next? Etc. Another benefit for AIs in that in games we consider more complex, the value of any given mistake often becomes diminished. If you make a single bad move in chess, it's enough to lose the game. In a 4x game the weight of individual decisions is not so high, it's all about the big picture. But as perhaps computer success in Go shows most clearly, actually seeing the big picture is not really necessary to produce play like you do. This, I think, is why research has moved more onto real time competitive games. Crushing humans at chess, go, and now poker as well is a pretty solid proof of concept for computers beating humans at any turn based game. When you start adding bunches of different layers to games I think it's more likely to handicap the human than the computer. Imagine playing some sort of 100x100 chess. We can only speculate, but I imagine the distance between the top AIs and humans would be far greater than it is in 8x8 chess. [1] - https://en.wikipedia.org/wiki/Shannon_number https://en.wikipedia.org/wiki/Shannon_number
  - darepublic 8y ago
    I agree that AIs will eventually win. But I only consider human beings beaten when the AI is interfacing with the machine the same way as humans -- looking at the monitor, and inputting commands through keyboard and mouse. That is on a different level than just hooking into a log of events and calculating your next move. Given enough time humans would do that better than the AI imo
    - 21 8y ago
      Come on, do you think hooking a robot to the keyboard is the hard part? Do you want 5 fingers too? And if watching the screen, do you want it to have bad eyes like we do too (good resolution only in the center)?
      - darepublic 8y ago
        Yea, I think having a robot use fingers to manipulate the keyboard like a human is a very hard part
        greenhatman 8y ago
        Mouse eye coordination would probably be harder.
        adrianN 8y ago
        Have you seen the robots that place components on circuit boards? I don't think that hitting keys on a keyboard is more difficult.
      - yters 8y ago
        Moravec's paradox!
      - gsich 8y ago
        Yes. Right now the AI has more information than the humans.
      - gaius 8y ago
        Come on, do you think hooking a robot to the keyboard is the hard part? Do you want 5 fingers too? Yes, it is. Also, “seeing” the screen rather than being able to directly introspect the game world digitally. Orders of magnitude harder. This is known as Moravec’s Paradox.
  - skgoa 8y ago
    > Humans, not so much, as in all top-level competitions, human abilities improve minimally at the top That's just not true in doto. The player base improves quite a bit over time. The top pro plays from only a few years ago are not impressive anymore.
skgoa 8y ago
The bots also have perfect knowledge of enemy hp/mana and the relative positioning through the API. In fact the bots are gifted perfect micromanagement through the API. This enables them to do things most humeans wouldn't try, because the risk of messing up and giving the enemy a huge advantage is too high.
- yvdriess 8y ago
  The API does not give the bots extra information compared to a human player. The micro- and reaction time edge is also being dulled to being more human-like. They still have superhuman team fight execution though. "We’ve increased the reaction time of OpenAI Five from 80ms to 200ms. This reaction time is much closer to human level, though we haven’t seen evidence of changes in gameplay as OpenAI Five’s strength comes more from teamwork and coordination than reflexes."
  - Super_Jambo 8y ago
    Part of the advantage in teamwork and coordination is presumably that they don't have a limitation on what data they can view at once? Dota & HON had people mod their client to give an optional bigger FOV resulting in bans for cheating. I'd assume the bots don't have to specify their screen position, plus no orientation response means a limitation on this wouldn't be meaningful anyway. What I'm saying is there's a big difference between 'I see lion on the mini-map down there' and 'lion showed for 1 frame on the other side of the map, his HP is 324, he has a TP scroll, no boots and a health pot.' Something I noticed is the bots seem to like range and AoE far more than the normal human meta. The humans being limited in the distance they can see to one screen were frequently just failing to appreciate how dangerous 2-3 bots half a screen away were to them. Quite a few teamfight wins came from the bots inevitably causing far more damage to the entire enemy team via heros like DP & Gyro. But this isn't really perfect teamfight execution. I'd have really liked to see a mirror match.
    - skgoa 8y ago
      Correct, the bots "see" the entire map. Well, the parts that are not hidden by 'fog of war'.
akerro 8y ago
>given more practice humans would beat this version of the ai given more practice bots would beat humans. that's the point, train bots, which are faster to train than humans to beat humans.
- sqrt17 8y ago
  It's important to keep in mind the exact quantity of "more practice". Current mechanisms of reinforcement learning are not very data-efficient, which means that often humans will learn faster than bots. It will still allow bots to discover any unrealistic advantage they have over humans (e.g. faster micromanagement), but if the game is fair and experience to learn from is limited, humans may still prevail. To "beat" Atari games, AIs trained using reinforcement learning had to put in significantly more than the 10k hours one would expect a human to put in in order to become expert. So, AIs won't beat humans in tasks where training data is costly; however these cases are not interesting to researchers and hence you won't hear from the respective results.
  - drusepth 8y ago
    To follow this thought, it's also worth pointing out that AIs also have an advantage of parallelization over humans in many cases. What might take an AI 100k hours can often be achieved in ~10k hours in parallel across 10 machines. This is what enables the current system to train over 180 years worth of games every day.

by kmnc 8y ago

hackandtrip 8y ago

It would be really interesting to see what bots will do once humans will try to exploit this strategy, like adapting to the lane movement they always do after winning a lane etc, humans did that once today but it may be used better. Excited for the TI to see less limitation, a better shop system (lot of money throw away) and maybe better ward.

21 8y ago
The next AI will always be much better (see AlphaGo progression). Humans, not so much, as in all top-level competitions, human abilities improve minimally at the top, because we have millions of humans competing against each other until the plateau of human performance is reached. Then you can push that a bit more with drugs (see doping in sports). And after that, you are pretty much done. So it's only a matter of time and effort until AIs are fully unbeatable.
- douglaswlance 8y ago
  Humans tend to progress by learning from each other. The AI will teach us.
  - asdfasgasdgasdg 8y ago
    Whether the AI teaches us or not, once it surpasses us, we will not catch up. This is the case for every game AI that has ever surpassed human performance so far, and there's no reason to expect that this will be different in the case of DotA.
    - ethbro 8y ago
      > game AI that has ever surpassed human performance so far Am I missing something, or does that set consist of Checkers, Chess, and Go so far? (presumably with analogous misc games of comparable complexity) Discounting the reaction time wins, I'd say the sample size is too limited to generalize to eventual AI behavior in more complex / open-ended games. Extrapolation was the cause of the last AI winter.
      - tialaramex 8y ago
        Poker. Although Heads Up Limit Texas Hold 'Em has been essentially solved the No Limit variant (in which you actually decide how much to bet) is very far from solved. An AI beat the best players in the world anyway. It plays crazy, but it plays crazy in a way they weren't able to exploit at all. Doug Polk (also one of the best in the world in this narrow specialty) did some funny videos but the humans got crushed. So that's a pretty different game, it's got a big luck factor and has asymmetrical information and still the AI just kept getting better and the humans... didn't
      - TangoTrotFox 8y ago
        I think you're looking at things with a sort of hindsight bias. Victory at chess was at one time considered to be the indicator of the emergence of true 'intelligence' in computing. The reason is that it's an extremely open, creative, and strategic game spattered with a minefield of tactical nuance. Nobody, human or computer, is getting even remotely close to scratching the depth of the game from a numeric point of view. There are some specifics on the numeric complexity of the game here [1]. The reason I point out the unfathomable numeric complexity is that it makes the games, from the perspective of an AI, effectively infinite. AIs are calculating, but to an extremely superficial degree relative to the depth of the game. E.g. - when a chess program says it's calculated to 30 ply (15 moves for both sides) what it really says is that it's seen up to 15 moves deep after intentionally ignoring or pruning 99.9999999999% of moves which it thinks probably aren't good -- something it still often gets wrong, but its 'understanding' of what is 'not wrong' is strong enough that it still results in a phenomenally strong level of play, compared to humans. There's no doubt that perfect play in chess would still go 1 billion - 0 against something like AlphaZero. So what matters is not the number of decisions to be made but the individual complexity of the decisions to be made. And in most games we consider complex the individual decisions are not really that complex, and complex systems can often be broken down into very simple games. For instance a great example of this is a 4x game. Taken as a whole they seem complex, but they're really just a large number of relatively simple components that are mostly independent. E.g. - Given this state, where do you explore next? Given this state, what do you research next? Etc. Another benefit for AIs in that in games we consider more complex, the value of any given mistake often becomes diminished. If you make a single bad move in chess, it's enough to lose the game. In a 4x game the weight of individual decisions is not so high, it's all about the big picture. But as perhaps computer success in Go shows most clearly, actually seeing the big picture is not really necessary to produce play like you do. This, I think, is why research has moved more onto real time competitive games. Crushing humans at chess, go, and now poker as well is a pretty solid proof of concept for computers beating humans at any turn based game. When you start adding bunches of different layers to games I think it's more likely to handicap the human than the computer. Imagine playing some sort of 100x100 chess. We can only speculate, but I imagine the distance between the top AIs and humans would be far greater than it is in 8x8 chess. [1] - https://en.wikipedia.org/wiki/Shannon_number https://en.wikipedia.org/wiki/Shannon_number
- darepublic 8y ago
  I agree that AIs will eventually win. But I only consider human beings beaten when the AI is interfacing with the machine the same way as humans -- looking at the monitor, and inputting commands through keyboard and mouse. That is on a different level than just hooking into a log of events and calculating your next move. Given enough time humans would do that better than the AI imo
  - 21 8y ago
    Come on, do you think hooking a robot to the keyboard is the hard part? Do you want 5 fingers too? And if watching the screen, do you want it to have bad eyes like we do too (good resolution only in the center)?
    - darepublic 8y ago
      Yea, I think having a robot use fingers to manipulate the keyboard like a human is a very hard part
      - greenhatman 8y ago
        Mouse eye coordination would probably be harder.
      - adrianN 8y ago
        Have you seen the robots that place components on circuit boards? I don't think that hitting keys on a keyboard is more difficult.
    - yters 8y ago
      Moravec's paradox!
    - gsich 8y ago
      Yes. Right now the AI has more information than the humans.
    - gaius 8y ago
      Come on, do you think hooking a robot to the keyboard is the hard part? Do you want 5 fingers too? Yes, it is. Also, “seeing” the screen rather than being able to directly introspect the game world digitally. Orders of magnitude harder. This is known as Moravec’s Paradox.
- skgoa 8y ago
  > Humans, not so much, as in all top-level competitions, human abilities improve minimally at the top That's just not true in doto. The player base improves quite a bit over time. The top pro plays from only a few years ago are not impressive anymore.

skgoa 8y ago

The bots also have perfect knowledge of enemy hp/mana and the relative positioning through the API. In fact the bots are gifted perfect micromanagement through the API. This enables them to do things most humeans wouldn't try, because the risk of messing up and giving the enemy a huge advantage is too high.

yvdriess 8y ago
The API does not give the bots extra information compared to a human player. The micro- and reaction time edge is also being dulled to being more human-like. They still have superhuman team fight execution though. "We’ve increased the reaction time of OpenAI Five from 80ms to 200ms. This reaction time is much closer to human level, though we haven’t seen evidence of changes in gameplay as OpenAI Five’s strength comes more from teamwork and coordination than reflexes."
- Super_Jambo 8y ago
  Part of the advantage in teamwork and coordination is presumably that they don't have a limitation on what data they can view at once? Dota & HON had people mod their client to give an optional bigger FOV resulting in bans for cheating. I'd assume the bots don't have to specify their screen position, plus no orientation response means a limitation on this wouldn't be meaningful anyway. What I'm saying is there's a big difference between 'I see lion on the mini-map down there' and 'lion showed for 1 frame on the other side of the map, his HP is 324, he has a TP scroll, no boots and a health pot.' Something I noticed is the bots seem to like range and AoE far more than the normal human meta. The humans being limited in the distance they can see to one screen were frequently just failing to appreciate how dangerous 2-3 bots half a screen away were to them. Quite a few teamfight wins came from the bots inevitably causing far more damage to the entire enemy team via heros like DP & Gyro. But this isn't really perfect teamfight execution. I'd have really liked to see a mirror match.
  - skgoa 8y ago
    Correct, the bots "see" the entire map. Well, the parts that are not hidden by 'fog of war'.

akerro 8y ago

>given more practice humans would beat this version of the ai given more practice bots would beat humans. that's the point, train bots, which are faster to train than humans to beat humans.

sqrt17 8y ago
It's important to keep in mind the exact quantity of "more practice". Current mechanisms of reinforcement learning are not very data-efficient, which means that often humans will learn faster than bots. It will still allow bots to discover any unrealistic advantage they have over humans (e.g. faster micromanagement), but if the game is fair and experience to learn from is limited, humans may still prevail. To "beat" Atari games, AIs trained using reinforcement learning had to put in significantly more than the 10k hours one would expect a human to put in in order to become expert. So, AIs won't beat humans in tasks where training data is costly; however these cases are not interesting to researchers and hence you won't hear from the respective results.
- drusepth 8y ago
  To follow this thought, it's also worth pointing out that AIs also have an advantage of parallelization over humans in many cases. What might take an AI 100k hours can often be achieved in ~10k hours in parallel across 10 machines. This is what enables the current system to train over 180 years worth of games every day.