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

101–110 of 219 posts

Re: OpenAI at the Dota 2 World Championships

#101

Post by the8472 »

NightlyDev wrote: post #95
habitue wrote: post #59

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…

You can restrain the bot by emulating human limitations if the goal is to use bots to figure out the game balance.

Add latency. Limit its actions per minute including camera movements and maybe even inertial limitations of input methods.

And even if the bot is not a perfect imitation of a human it still might figure out surprising things that would take human players years of real time to figure out by chance since they can only spend so much time on playing and experimenting.

Re: OpenAI at the Dota 2 World Championships

#102

Post by PeanutCurry »

NightlyDev wrote: post #95
habitue wrote: post #59

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…

This was generally my interpretation as well. MMORPGs like World of Warcraft already struggle to balance classes with high skill ceilings for use by both the casual players and the hardcore players while keeping those classes enjoyable for both groups. In some ways you could call that the dev-meta, making a competitively balanced meta game while still offering an appealing level of fun for the various subgroups of the player-base.

Re: OpenAI at the Dota 2 World Championships

#103

Post by vecter »

jmkapz wrote: post #92
hmate9 wrote: post #88

Earlier quoted context omitted.

Of course it's a good move. Millions watched AlphaGo, many of whom have never played Go before. Millions, almost every single gamer will watch AI vs Dota that will be amazing marketing for Valve.

I'm talking mid to long term here. To rephrase, is there a sizeable amount of players who will stop playing, knowing that they cannot ever be the best? Because when AI beats one, I don't believe there is a coming back. For the concrete AlphaGo example, of course it was hugely popular but I would be interested in the evolution on the number of players, how many of those viewers started playing balanced by how many exi…

I don't think it will matter at all. Literally, zero impact.

Chess computers crush the world's best humans, yet here we are in meatspace, still vying to win the human world chess championship.

Also, 99.99999% of players realize they'll never ever be close to being the best in the world, yet they still play the game because it is inherently fun.

Re: OpenAI at the Dota 2 World Championships

#104

Post by kahlonel »

Even with a lot of restrictions, I can't imagine how much variables this bot has to take into account; and then generate the output within a sub-millisecond time period. The most interesting part was aggressive positioning of the bot, and faking the spells to scare the shit out of Dendi (the pro player competing against it). Will definitely follow the progress of this project.

Re: OpenAI at the Dota 2 World Championships

#105

Post by unrealhoang »

hmate9 wrote: post #85
ronald_raygun wrote: post #24

Earlier quoted context omitted.

I don't want to be "that guy", but I used to play a lot of LOL, and I feel like 5v5 is in a different ballpark in terms of difficulty than a 1v1. Even champion select is strategically intensive. TSM (one of the best US teams), lost to an amateur team because they got out done in team selection. https://www.youtube.com/watch?v=g7l30-4A6WQ

It is extremely difficult. 1v1 is like beating humans at chess. 5v5 is like beating humans at Go. That's the kind of jump it is.

I would even bring that to one more level of difficulty. There are so many choices can be made in-game to gain advantages over enemy: item build, gank/farm distribution, resource sharing, vision fight, choose to defense or let go/trade of tower/objectives...

If we can have AI that beat Pro at 5v5 in Dota, it would very likely be an AGI.

Re: OpenAI at the Dota 2 World Championships

#106

Post by popcorncolonel »

habitue wrote: post #59

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…

Icefrog also balances the game for fun, not just for the theoretical limit of what's equally fair.

Re: OpenAI at the Dota 2 World Championships

#107

Post by TimPC »

habitue wrote: post #59

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…

Balance for humans and robots is different. Different play skills are easy or hard and those skills matter different amounts on different heroes . Effectively instead of balancing for fair heroes, you'd be balancing for maximizing the performance of heroes with the small human-robot gaps. It would be very easy to make the game boring because in some sense you're making the easy to play heroes the best ones and I think many games aim to be biased slightly in the opposite direction.

Re: OpenAI at the Dota 2 World Championships

#108

Post by PeanutCurry »

the8472 wrote: post #101
NightlyDev wrote: post #95

Earlier quoted context omitted.

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…

You can restrain the bot by emulating human limitations if the goal is to use bots to figure out the game balance. Add latency. Limit its actions per minute including camera movements and maybe even inertial limitations of input methods. And even if the bot is not a perfect imitation of a human it still might figure out surprising things that would take human players years of real time to figure out by chance since t…

This approach seems like it would be more useful for bug discovery than game balancing. Actively collecting a slew of performance metrics, similar to how companies like Blizzard already do, would give you a better idea of how to proceed with balance changes because you can review the data as it applies to your actual playerbase and even organize it based on ranking categories to get an idea of how your decisions would affect players at different skill levels.

Re: OpenAI at the Dota 2 World Championships

#109

Post by popcorncolonel »

bhntr3 wrote: post #21

This was fun. It's our intern's last day on the ML infra team here. And he happens to be a competitive collegiate DOTA player. We were all crowded around watching the screen, shouting. Couldn't have planned a better send off. Nice job, OpenAI!

As a ML researcher and an avid dota fan, I'm jealous! And it must have been great with Dendi the legend there too.

Re: OpenAI at the Dota 2 World Championships

#110

Post by __s »

habitue wrote: post #59

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…

Aside from human vs machine style comments, the issue this would also have to avoid is what's referred to as "over tuned". ie when content is balanced to a point where variation has to be maximally efficient. See also http://nethack4.org/blog/strategy-headroom.html
Post reply on HN