Bitcoin Forum

Economy => Gambling discussion => Topic started by: PonziLord on October 11, 2016, 05:23:02 PM



Title: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 11, 2016, 05:23:02 PM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PieCoin on October 11, 2016, 05:30:05 PM
Are you able to explain how an AI can learn from complete randomness?


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: newIndia on October 11, 2016, 05:30:36 PM
Why your name is PonziLord? Are u planning a Ponzi with AI?


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 11, 2016, 05:32:42 PM
Why your name is PonziLord? Are u planning a Ponzi with AI?

That's just a username. That doesn't have anything to do with this. lol


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 11, 2016, 05:36:02 PM
Are you able to explain how an AI can learn from complete randomness?

It wouldn't be complete randomness. The AI will decide to make the bet win or not. You can check some things on youtube like learning AIs playing videogames. The implementation would be the same.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PieCoin on October 11, 2016, 06:33:07 PM
Are you able to explain how an AI can learn from complete randomness?

It wouldn't be complete randomness. The AI will decide to make the bet win or not. You can check some things on youtube like learning AIs playing videogames. The implementation would be the same.

I don't need to check videos on YouTube, videogames and dice are not the same. Dice is a random number generator with no effect on the output from the input of the user. No previous or future actions affect the outcome of a dice roll so there is nothing for the AI to learn.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: MartinL on October 11, 2016, 06:50:45 PM
What would be the draw for players? As a player, I would rather play a provably random game rather than against an AI that can generate arbitrary results for all I know.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: Daffadile on October 11, 2016, 06:58:05 PM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

I have an idea for you. Why don't you shut the hell up and gtfo. If there is AI that can choose to lose or win then that means you know the lose or win so you have just admitted to cheating. Go dig a hole and lie in it. You really are an idiot.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: MartinL on October 11, 2016, 07:09:17 PM
I have an idea for you. Why don't you shut the hell up and gtfo. If there is AI that can choose to lose or win then that means you know the lose or win so you have just admitted to cheating. Go dig a hole and lie in it. You really are an idiot.

Future game results are known to the house in many popular games, e.g. bustabit. In and of itself this does not constitute or even imply any cheating.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: RHavar on October 11, 2016, 07:27:49 PM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

It's an interesting idea, but I don't see how it could work. Your first issue is going to be that if it's not predetermined, it's not going to be provably fair. But if it is predetermined, it's going to be vulnerable to players exploiting it (see the Nash equilibrium).

As a casino operator, my goal is for players have a great time ... but ultimately (over a long time frame) end up losing. Often I see a new player come to the site, and hit some really bad luck, get destroyed. It's bad for the player (who enjoys that) which makes it bad for me.

So there's some merit in the idea of having some "AI" that boosts unlucky players with some "good luck", so they don't get absolutely destroyed. But in order for that to really be feasible, the site would have to do the opposite to lucky players. But now if you have a casino that boosts the luck of unlucky players and decreases the luck of lucky players, which lowers variance and stops being an exciting casino and something more akin to an inhumane abattoir. Not really particularly appealing for players.

I have however read studies on some slot machine players, where they're actually not really looking for a high-variance thrilling experience but actually looking for stability and time-burning and predictable and hypnotic with the chance of a giant jackpot. Perhaps your "AI" idea could work there, creating a more controlled experience.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: FrueGreads on October 11, 2016, 09:34:32 PM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

I'm not sure this would be a good idea. Dice games and pure casino games, like roulette etc are based on luck, and with your idea, if I understood it correctly that would no longer be the case. You would actually be decreasing variance and I think that's part of the appealing. First users know, or should know, that they will lose in the long run, so their only chance on winning big amounts is variance. If they are smart they test their luck, and if they win they stop and be happy with that profit. With your idea they would lose less, but they would win less.

I think there would be another problem with that. If stats are public, players could use you AI behavior to exploit the site and know if the AI is going to decide if the bet is going to be a win or a loss, because that decision is not random.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 12:07:10 AM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

I have an idea for you. Why don't you shut the hell up and gtfo. If there is AI that can choose to lose or win then that means you know the lose or win so you have just admitted to cheating. Go dig a hole and lie in it. You really are an idiot.

Please read the posts below. This gives you a much better explanation.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: Erza on October 12, 2016, 12:29:52 AM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

I really cant agree on AI things because this will become not provably fair and you can change your AI whenever and wherever you can. So I would not really like the idea of using AI compare to just some randomness


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 12:49:00 AM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

I really cant agree on AI things because this will become not provably fair and you can change your AI whenever and wherever you can. So I would not really like the idea of using AI compare to just some randomness
That's why when this is implemented, the source code must be open to the public.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 12:53:29 AM
I am currently researching learning artificial intelligence and I want to ask if is there anyone interested on a dice site powered by an AI instead of pure luck? The AI will decide what bet will win and what bet will lose while maintaining a 1% profit from the wagered money and having the largest amount of wagered money? Please give me some good suggestions for this. Thanks.

For example:

The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

It's an interesting idea, but I don't see how it could work. Your first issue is going to be that if it's not predetermined, it's not going to be provably fair. But if it is predetermined, it's going to be vulnerable to players exploiting it (see the Nash equilibrium).

As a casino operator, my goal is for players have a great time ... but ultimately (over a long time frame) end up losing. Often I see a new player come to the site, and hit some really bad luck, get destroyed. It's bad for the player (who enjoys that) which makes it bad for me.

So there's some merit in the idea of having some "AI" that boosts unlucky players with some "good luck", so they don't get absolutely destroyed. But in order for that to really be feasible, the site would have to do the opposite to lucky players. But now if you have a casino that boosts the luck of unlucky players and decreases the luck of lucky players, which lowers variance and stops being an exciting casino and something more akin to an inhumane abattoir. Not really particularly appealing for players.

I have however read studies on some slot machine players, where they're actually not really looking for a high-variance thrilling experience but actually looking for stability and time-burning and predictable and hypnotic with the chance of a giant jackpot. Perhaps your "AI" idea could work there, creating a more controlled experience.

I don't think exploitations would work as the AI itself will learn if it is being exploited or not and the players would have no idea what would others bet (In Nash equilibrium, the other player must know the strategy of the other). And this is already somewhat implemented on bustabit. Where the program itself will decide to maintain a profit for the website.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: erpbridge on October 12, 2016, 01:13:28 AM
I am super confused. I gamble because of it being luck and the 1% may not always be there for my betting sample size of say 1000. If the AI forces the 1% for a smaller sample size then won't I definitely be a losing player ?


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: RHavar on October 12, 2016, 01:18:53 AM
and the players would have no idea what would others bet (In Nash equilibrium, the other player must know the strategy of the other).
The high level premise of your AI, is that it's going to try make players lose X% of the amount they bet. That itself is leaks a sufficient amount of information that it would make most forms of gambling, impractical (e.g. I know if have been unlucky, I will now be unnaturally lucky. If I have been lucky, I will now be unnaturally unlucky etc.)

It's possible to constrain the gambling game in such a way it won't be exploited, I guess (i.e. only fixed size bets, no hi/lo choice).


Quote
And this is already somewhat implemented on bustabit. Where the program itself will decide to maintain a profit for the website.

Hm. Not to be crass, but you should look at how existing stuff works before trying to design a different system. Bustabit and, in fact every single provably fair casino I'm aware of does *not* work like this, on the contrary, the next 10M games have been defined and set in stone (https://bitcointalk.org/index.php?topic=922898.0), and regardless of how much players have won or lost the exact set of games will continue to play in order. If the casino loses 200 BTC on night, it will *not* alter the game in any shape or form to "recover" the loss etc.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 01:44:49 AM
and the players would have no idea what would others bet (In Nash equilibrium, the other player must know the strategy of the other).
The high level premise of your AI, is that it's going to try make players lose X% of the amount they bet. That itself is leaks a sufficient amount of information that it would make most forms of gambling, impractical (e.g. I know if have been unlucky, I will now be unnaturally lucky. If I have been lucky, I will now be unnaturally unlucky etc.)

It's possible to constrain the gambling game in such a way it won't be exploited, I guess (i.e. only fixed size bets, no hi/lo choice).


Quote
And this is already somewhat implemented on bustabit. Where the program itself will decide to maintain a profit for the website.

Hm. Not to be crass, but you should look at how existing stuff works before trying to design a different system. Bustabit and, in fact every single provably fair casino I'm aware of does *not* work like this, on the contrary, the next 10M games have been defined and set in stone (https://bitcointalk.org/index.php?topic=922898.0), and regardless of how much players have won or lost the exact set of games will continue to play in order. If the casino loses 200 BTC on night, it will *not* alter the game in any shape or form to "recover" the loss etc.

If you lose, the odds will still be the same. As the 1% is not based on player's profits but on the whole game including other players. Meaning if you lost, other players would profit on your loss not the game.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: NLNico on October 12, 2016, 04:47:26 AM
Actively cheating players without them knowing why/how much/when/etc? No, sounds like a bad idea. There is no way for you to prove you will keep a 1% HE. You could let a big real player lose any bet and let your "house players" win their profit (and claim it's fair because other players won.) It seems just like a bad idea imo.





However, if you give a hash with the outcome before each bet.. then it could be interesting. Basically you would not let players pick a clientseed. So this way, you could cheat them by predicting their next bets (and give hashes based on that.) Players could also abuse this, by trying to trick (and/or learn) your algorithm/patterns.

So players will be able to verify if the outcome is really the hash that you gave before they made their bet. But you could still increase your odds by predicting how they bet. Meanwhile the player could also increase their odds by trying to predict how you predict their betting.

I don't think it's really a smart idea either tbh. But it is interesting :P


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: MartinL on October 12, 2016, 07:02:01 AM
That's why when this is implemented, the source code must be open to the public.

Unfortunately open-sourcing its code is not sufficient to make a game provably fair. You cannot prove that the published code is actually running on your server and not a modified version.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 07:46:35 AM
Actively cheating players without them knowing why/how much/when/etc? No, sounds like a bad idea. There is no way for you to prove you will keep a 1% HE. You could let a big real player lose any bet and let your "house players" win their profit (and claim it's fair because other players won.) It seems just like a bad idea imo.





However, if you give a hash with the outcome before each bet.. then it could be interesting. Basically you would not let players pick a clientseed. So this way, you could cheat them by predicting their next bets (and give hashes based on that.) Players could also abuse this, by trying to trick (and/or learn) your algorithm/patterns.

So players will be able to verify if the outcome is really the hash that you gave before they made their bet. But you could still increase your odds by predicting how they bet. Meanwhile the player could also increase their odds by trying to predict how you predict their betting.

I don't think it's really a smart idea either tbh. But it is interesting :P
Source code would be openly available. And everything would be on AI's decision. Not the programmers (We will just adjust which parameter should be prioritized, but on what is the result of the bet is the AI's decision).


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: MartinL on October 12, 2016, 08:02:00 AM
Source code would be openly available. And everything would be on AI's decision. Not the programmers (We will just adjust which parameter should be prioritized, but on what is the result of the bet is the AI's decision).

Imagine the following scenario: The game operator publishes the fair AI's source code. Then–without publishing these changes–he modifies the AI to favor a set of house-controlled players. He also modifies the AI to disadvantage all other players so that the house edge ostensibly remains 1 %.

It is not possible for real players to distinguish this scenario from one where the game operator is fair–at least not with any certainty. Therefore the game is not provably fair.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: Positid on October 12, 2016, 08:14:03 AM
Honestly, I cannot tell you that I will be interested if it is powered by AI because in the first place I do not understand how does it work with AI, for me that seems complicated and will not gain interest from the majority here.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: Invulner on October 12, 2016, 08:15:54 AM
I don't think that this is a good idea... Obviously although you are keeping a 1% house edge, you'd still be basically "rigging" the game with an AI. Rigging is rigging, no matter done with AI or not.

And also, how are you to prove that your AI is fair, since the bet result isn't generated beforehand?


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: patt0 on October 12, 2016, 08:46:16 AM
I'm not sure I'm understanding this. If it's based on a learning AI doesn't that mean that a player that knows the AI pattern of behaviour, or how it is evolving, could exploit that and predict the outcome of a bet? If it's not random, and it follows a pattern then you just have to know that pattern and you will keep winning I guess, because you know what the AI will decide. If it works like this, then I don't know how you can bring this idea to gamble.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 08:58:40 AM
I'm not sure I'm understanding this. If it's based on a learning AI doesn't that mean that a player that knows the AI pattern of behaviour, or how it is evolving, could exploit that and predict the outcome of a bet? If it's not random, and it follows a pattern then you just have to know that pattern and you will keep winning I guess, because you know what the AI will decide. If it works like this, then I don't know how you can bring this idea to gamble.

That's the thing. You wouldn't how it will evolve. And there will be no specific behaviour as the AI evolves on every bet.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: crytoboost on October 12, 2016, 09:53:42 AM
Can you explain how this would be possible to learn AI for dice game but I think there will be more high chance to get cheated from the house, because it will be hard for anyone to keep it running without any house edge where only player always will make profit by using this AI for dice game.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: PonziLord on October 12, 2016, 10:04:19 AM
Can you explain how this would be possible to learn AI for dice game but I think there will be more high chance to get cheated from the house, because it will be hard for anyone to keep it running without any house edge where only player always will make profit by using this AI for dice game.
There are AIs that learn playing games out of scratch and this AI would be based on that. Check sethbling's MarI/O


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: hatshepsut93 on October 15, 2016, 11:59:19 AM
Can you explain how this would be possible to learn AI for dice game but I think there will be more high chance to get cheated from the house, because it will be hard for anyone to keep it running without any house edge where only player always will make profit by using this AI for dice game.
There are AIs that learn playing games out of scratch and this AI would be based on that. Check sethbling's MarI/O

Why would we need an AI for dice? Gambling is supposed to be based on randomness, bringing some agent that decides winners and losers only creates a field for cheating, I don't see any benefits for players.

You can try creating some other games with your A.I. but you won't be able to control outcomes if those games are competitive, like chess or checkers.


Title: Re: Would anyone be interested on a dice site powered by a learning AI?
Post by: Herbert2020 on October 15, 2016, 12:21:16 PM
...
The AI would decide what bet to lose, from a whale betting 1 BTC or from 100 users betting 0.01 BTC or something like that.

this seems like a cool idea only because of the artificial intelligence part of it and the fact that i am interested to know how it really is going to work.
but when thinking more about it, in the end it would be more interesting to the house (the website owners) not the players because dice games must be completely random and fair. and you cannot change any of these two crucial factors unless you want to be left branded as a scammer.