Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenAI at the Dota 2 World Championships

openai.com

Post reply on HN
Print view View original

181–190 of 219 posts

Re: OpenAI at the Dota 2 World Championships

#181

Post by backpropaganda »

ahh wrote: post #54

The developer interviewed claimed that there was no domain-specific knowledge and implied (though didn't explicitly state) there wasn't any training against non-OpenAI bots or players. (I'd love to know the reward function they used for whatever Q-learning variation they ran with.) If this is accurate, one of the things I'm most impressed with is that the bot figured out creep-blocking. (I can't find a good GIF, but…

I think they either had a reward for creep blocking or they used the "learning from human preferences" paper to coach it for that behaviour. I would be very terrified if it learnt creep blocking on its own with no help.

Re: OpenAI at the Dota 2 World Championships

#182

Post by quadcore »

dirtyaura wrote: post #155
darrenkopp wrote: post #60

Really was quite interesting to watch ( https://www.twitch.tv/videos/166172514?t=7h3m10s ). Honestly, the bot played _extremely_ well, but I think the biggest advantage was how much faster it's reaction time was and it's movements were likely much more precise than a human is with a mouse. I'm pretty interested in seeing their 5v5 results as well. It seems like that will have similar results as the bots can coordinat…

It would be a fantastic side project to teach bots to learn to speak "Dota". Learning to communicate efficiently is likely much easier as the vocabulary and intentions behind them are constrained.

Side question: if we consider a team of 5 bots, would they need to communicate?

Re: OpenAI at the Dota 2 World Championships

#183

Post by sbarre »

quadcore wrote: post #182
dirtyaura wrote: post #155

Earlier quoted context omitted.

It would be a fantastic side project to teach bots to learn to speak "Dota". Learning to communicate efficiently is likely much easier as the vocabulary and intentions behind them are constrained.

Side question: if we consider a team of 5 bots, would they need to communicate?

If you wanted them to play "within the rules", they would need some way to communicate via the game.. Text chat would make the most sense since adding in TTS / STT would seem unnecessary.. They could even communicate in some kind of shorthand language that only the bots understand..

Re: OpenAI at the Dota 2 World Championships

#184

Post by NegatioN »

sondr3 wrote: post #123
DannyDaemonic wrote: post #75

Earlier quoted context omitted.

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…

This is luckily a problem Dota2 doesn't suffer from, we've been incredibly lucky to have Icefrog be the main force behind it for over a decade. It's an incredibly well balanced game, at this International (World Championships for Dota2) 112 of 113 heroes has been picked and/or banned. The amount of diversity you have in strategy is amazing, there is a meta obviously, but teams have very individual styles and heroes t…

It's also interesting to note that icefrogs balance strategy revolves around strengthening strengths, and making initial weaknesses more pronounced. I think this helps a lot with diversity.

Re: OpenAI at the Dota 2 World Championships

#185

Post by Retric »

keerthiko wrote: post #134
dvt wrote: post #87

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…

It pulled some interesting tactical moves that I was keeping an eye out for. - 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…

Bots hesitate via thinking time. You see this in chess matches most obviously when a high level bot spends more time on some moves because it realizes the situation is unusually important or tricky.

Re: OpenAI at the Dota 2 World Championships

#186

Post by Cyphase »

sbarre wrote: post #183
quadcore wrote: post #182

Earlier quoted context omitted.

Side question: if we consider a team of 5 bots, would they need to communicate?

If you wanted them to play "within the rules", they would need some way to communicate via the game.. Text chat would make the most sense since adding in TTS / STT would seem unnecessary.. They could even communicate in some kind of shorthand language that only the bots understand..

I think the grandparent is asking whether the bots would need to communicate at all, as opposed to just "knowing" what the other bots are "thinking".

Re: OpenAI at the Dota 2 World Championships

#187

Post by Cyphase »

quadcore wrote: post #182
dirtyaura wrote: post #155

Earlier quoted context omitted.

It would be a fantastic side project to teach bots to learn to speak "Dota". Learning to communicate efficiently is likely much easier as the vocabulary and intentions behind them are constrained.

Side question: if we consider a team of 5 bots, would they need to communicate?

Just off the top of my head, they would only need to communicate if computation time is enough of an issue and they wanted to share conclusions without all of them recomputing those conclusions, or if there was some randomness they wanted to share before it became apparent.

Re: OpenAI at the Dota 2 World Championships

#189

Post by rapind »

quadcore wrote: post #182
dirtyaura wrote: post #155

Earlier quoted context omitted.

It would be a fantastic side project to teach bots to learn to speak "Dota". Learning to communicate efficiently is likely much easier as the vocabulary and intentions behind them are constrained.

Side question: if we consider a team of 5 bots, would they need to communicate?

I think so. The slight difference in latency would push their decisions slightly out of sync.

Like if the opponent changed where one unit was moving within a lag delta and a different decision was made, but also seeing a teammate bot execute a decision first would alter the inputs for the others.

Re: OpenAI at the Dota 2 World Championships

#190

Post by simonebrunozzi »

csomar wrote: post #188

Is this the start of the end of online gaming? If bots outperform humans, what to prevent someone from a running bots for online games (even Poker and Chess) and being sure that everyone loses?

Or, quite the opposite.

I played DotA a few dozen hours, mostly against the computer, to learn. Then I played only a few matches against people, but I generally didn't like them.

Why? Because there were expectations that I would know certain things, or that I would perform at a certain level, and the other players were, on average, a bunch of young kids with tons of time to play DotA and their own weird jargon and acronyms for me.

I would strongly prefer to play against bots only, and be able to adjust the difficulty, to make the game always challenging and interesting as I progress in skill level. It could also make the "competitor" in me happy by knowing what my (objective) ranking would be.

Post reply on HN