131–140 of 219 posts
Earlier quoted context omitted.
> but I think that it was one of those cases where it was "coached" (same as the creep blocking). What makes you think that? In the interview, they explicitly disclaim this.
In the interview, they literally said they coached it to have certain behaviors.
As a longtime DotA player and someone who's following the pro scene, this is very impressive. Especially considering how it's beaten Sumail, widely regarded as one of the best 1v1 players in the world. Can't wait to see what OpenAI have in store a year from now for 5v5.
Sumail won once until they gave the bot insane creep blocking skills. https://twitter.com/Phillip_Aram/status/896162260455800832
That's insane that he figured out how to beat it so quickly. I feel there are other ways to cheese it too. Like maybe survive until 6 -> rush shadow amulet -> smoke -> activate and walk into lane during fade time -> ult when the wave/bot is on top of you. I bet the bot has never seen invisibility and wouldn't know what to do
So an interesting thing you could do as a game company with an "unbeatable" bot is to use it to balance a metagame. Let the bots all play each other and tweak character stats etc until they win a proportional amount of the time. (This presupposes that the bot learns the game the way OpenAI claims it is, without needing to learn from player replays or from playing in a very different way than human players do with tre…
Saw this live and as someone that's played Dota for many years, it's not as impressive as people make it out to be. A Shadowfiend 1v1 is primarily a test of mechanical skill and not judgment. A similar comparison is if a Counter-Strike or Quake bot would instantly snap to your head getting a headshot. Like, yeah, cool I guess, and it would beat 100% of human players 100% of the time, but it's not even remotely as imp…
- In game 1, when the AI triggers his flask of health regen (which gets dispelled if you take hero damage) it zones Dendi out knowing he's going to come in for an attempt to dispel his regen. It does this with lots of non-committal, pre-emptive shadowrazes. It doesn't do this any other time. It's aggressively predicting what Dendi may favor based on its own cirucmstances.
- In game 2, Dendi said he was going to try letting one creep ahead and see if it would give him a lane advantage. [very technical DOTA mechanics] What this does is if the opponent creep blocks better than you, their creep wave strikes your overextending creep without taking any damage, under their tower, thus significantly weakening your creep wave, making the tug of war swing very heavily in your favor. You counter this by also letting one of your creeps out ahead as soon as you realize it. The AI that was perfectly blocking all the way till the creeps met up in G1, as soon as it saw the leading creep let one of its creeps forward. A perfect response that is highly unlikely to have been coached but determined as the best choice to, what I'm guessing is to return the game to a familiar state.
- The bot cornered Dendi between some trees and his tower, and then with 0 fakeouts/hesitation shadowrazed him. It was not just a matter of knowing the target was in range, but also accounting for the probability that he could dodge it. Not 100% certain about the mechanics here, but I believe: Shadowraze has a ~.2s cast animation. It can be canceled during the first .1s (fake out), but after that it is committed to, you will complete the animation, spend the mana, and cast the spell. If the cast crosses this window and the opponent is capable of moving the short distance needed to dodge in this time, it is theoretically not something the AI can guarantee on pure mechanical skill. So cornering is an approach advanced players use to reduce the probability and room for dodging a skillshot, and I found it taking instant advantage of that opportunity very impressive.
Besides the obvious mechanical precision of a bot once it's made a decision, the other inhuman factor here is what Dendi mentions, that it has zero hesitation to take advantage of even the smallest opening it has available. A pro human who is even instantly aware of an opportunity, no matter how theoretically certain they are that they should take it, hesitates at least a little bit to think "is there a trap? Am I not considering something?"
Hesitation is an incredibly human trait.
OT: Is backdooring prevented by the game engine yet? Annoyed me that a useful tactic was "against the rules"
It's always been prevented in Dota 2. Buildings have backdoor protection, which makes them regenerate lost HP from recent attacks unless there are creeps nearby.
Earlier quoted context omitted.
> but I think that it was one of those cases where it was "coached" (same as the creep blocking). What makes you think that? In the interview, they explicitly disclaim this.
In the interview, they literally said they coached it to have certain behaviors.
Earlier quoted context omitted.
How does having a bot succeed cause appeal to be lost? First you're suggesting that the strategy will be "found" and never beaten (no room for improvement, not even in hero selection). Secondly.. Chess & Go have been around for thousands of years and they still are played en masse daily.
> How does having a bot succeed cause appeal to be lost? You might think not beating bots is OK. Not for me and many people have been involved in competitive scene. The moment machine beats human, a large part of the competition is gone. The essence is that you want to be the best. But if all you can do is to beat some other inferior opponents, what's the point? > Secondly.. Chess & Go have been around for thousands…
So an interesting thing you could do as a game company with an "unbeatable" bot is to use it to balance a metagame. Let the bots all play each other and tweak character stats etc until they win a proportional amount of the time. (This presupposes that the bot learns the game the way OpenAI claims it is, without needing to learn from player replays or from playing in a very different way than human players do with tre…
Humans are not machines. Games are mostly based on the skills of the players, so should games be balanced by assuming a human can do what a machine can? Of course not! A bot can usually make computations way faster and more presice than humans. Let's say a bot plays against another bot, both with different characters. One character might seem really overpowered compared to the other, but that might only be because th…
AI bots such as this one are usually limited to emit actions every N frames, so they can't possibly react faster even if they can solve the perception problem. There is nothing interesting for AI advancement in competing for the best reflex speed - it's all about strategy.
So an interesting thing you could do as a game company with an "unbeatable" bot is to use it to balance a metagame. Let the bots all play each other and tweak character stats etc until they win a proportional amount of the time. (This presupposes that the bot learns the game the way OpenAI claims it is, without needing to learn from player replays or from playing in a very different way than human players do with tre…
I think balance is a bigger problem than people realize. It's much harder than you'd think, and the most common way of balancing things is to simply make them more "samey". I was part of the Warcraft 3 beta. When it started all the races felt so unique. It's a bit foggy now but I think, for example, all the elves buildings - which appeared tree like - would automatically regenerate. Once they started balancing the ga…
which is the wrong way to balance a competitive game