darkorbit new anti bot system

04/01/2016 03:19 topastop#121
Quote:
Originally Posted by ~Demetrio~ View Post
lol ,if a player change his behavior this means is a bot? i don't think so!
BP could not predict when a bot change the bot behavior. every bot developer can change it random days hours. you think it's possible detect bots without detect legitim players? it's clear they can't catch all pixel-bots do you remember 2 years ago when they banned legitim accounts?

About AI: The bot can learn by it'self and then this change the behavior evey hour/day/Minute. The Bot can act just like a human when the developer is great.
None of the current bots are even close to be undetectable to a neural network and I highly doubt anyone here, me included, can make one.

EDIT: Given a decent sample size.
04/01/2016 06:32 ~Demetrio~#122
Quote:
Originally Posted by topastop View Post
None of the current bots are even close to be undetectable to a neural network and I highly doubt anyone here, me included, can make one.

EDIT: Given a decent sample size.

One of my top favorite subjects are the AI and reverse engineering.

Although not everyone can make a robot with : neural network .
It can be done based on personal experiences bot, for example :

The bot can learn from past experiences with enemies, example: if the bot sees the enemy and HP get low then the bot will know that it is not good to be close of enemies at the moment .

Or if the bot detects enemies and has no effect, then the bot will keep careful with them.

Save the best movements , timings , Spots with more boxes , blah blah this is just a very little example, all is about the developer :)
04/01/2016 18:25 Diаmonds#123
Quote:
Originally Posted by ~Demetrio~ View Post
One of my top favorite subjects are the AI and reverse engineering.

Although not everyone can make a robot with : neural network .
It can be done based on personal experiences bot, for example :

The bot can learn from past experiences with enemies, example: if the bot sees the enemy and HP get low then the bot will know that it is not good to be close of enemies at the moment .

Or if the bot detects enemies and has no effect, then the bot will keep careful with them.

Save the best movements , timings , Spots with more boxes , blah blah this is just a very little example, all is about the developer :)
Because AI is your favorite subject doesn't mean that you know anything about that. AI is a very huuge field, and one of the most complicated of programming. (I'm saying programming, but it's rather maths).
And flying away when you get low HP has nothing to do with AI lol, that's barely an if(){} in the code...
04/01/2016 18:56 ~Demetrio~#124
Quote:
Originally Posted by Diаmonds View Post
Because AI is your favorite subject doesn't mean that you know anything about that. AI is a very huuge field, and one of the most complicated of programming. (I'm saying programming, but it's rather maths).
And flying away when you get low HP has nothing to do with AI lol, that's barely an if(){} in the code...
This is the reason i said you'll never understand ... who said about if () {} ? , i already said this is a very little example ! Do you really know about AI Concept? how would you call an bot that learns about it pasted experiencies? and learns about it'self behavior?

Talking about pixelbots they have a lot of limitations! but it does not mean is not possible to do!

you should not only read before answer.
04/02/2016 00:16 topastop#125
Quote:
Originally Posted by ~Demetrio~ View Post
This is the reason i said you'll never understand ... who said about if () {} ? , i already said this is a very little example ! Do you really know about AI Concept? how would you call an bot that learns about it pasted experiencies? and learns about it'self behavior?

Talking about pixelbots they have a lot of limitations! but it does not mean is not possible to do!

you should not only read before answer.
It is nearly impossible to make dynamic neural network to play this game.
How do you specify success? How do you specify failure? And the most important question: How the hell can you make the bot learn for tons of gerations? How would you emulate the game? It is very easy to do when you can backup data, but in this case, is nearly impossible with current technology.
04/02/2016 02:02 ~Demetrio~#126
Quote:
Originally Posted by topastop View Post
It is nearly impossible to make dynamic neural network to play this game.
How do you specify success? How do you specify failure? And the most important question: How the hell can you make the bot learn for tons of gerations? How would you emulate the game? It is very easy to do when you can backup data, but in this case, is nearly impossible with current technology.
here's when the developer joins to game ;).
Although i said that the pixelbots has limitations i won't say anything more about this topic , now you have to wait and see ;)
04/02/2016 12:26 0cf5e#127
Quote:
Originally Posted by topastop View Post
It is nearly impossible to make dynamic neural network to play this game.
How do you specify success? How do you specify failure? And the most important question: How the hell can you make the bot learn for tons of gerations? How would you emulate the game? It is very easy to do when you can backup data, but in this case, is nearly impossible with current technology.
Q learning would be pretty good at it. More stuff collected and less deaths mean more points.
Training? You can implement learning in bot, so every user using bot would train your AI. After a while it will get better and better.
04/02/2016 14:32 topastop#128
Quote:
Originally Posted by 0cf5e View Post
Q learning would be pretty good at it. More stuff collected and less deaths mean more points.
Training? You can implement learning in bot, so every user using bot would train your AI. After a while it will get better and better.
That is not how it works. That is not how any of this works.

https://en.wikipedia.org/wiki/Artifi...raining_issues

Quote:
Originally Posted by ~Demetrio~ View Post
here's when the developer joins to game ;).
Although i said that the pixelbots has limitations i won't say anything more about this topic , now you have to wait and see ;)
And I would like to add:
https://en.wikipedia.org/wiki/Artifi...e_applications
04/02/2016 15:40 ~Demetrio~#129
You know about Neural networks there's many ways to appply this , as a guy already said this is a extensive topic , so as you said would be hard for you Apply this in darkoribt , if you think you cant ! why say that anothers guys can not?

The bot can learn about it's experiencies saving all in a backup! for example apply this to Enemys and better ways to flee from them!

An example if a enemy with user id :XXXX Kills my ship more than 2 times , the bot will recognize the user iD and then will try to find a way to escape based in it's pasted experiences.

Or about boxes the bot could know the best Spot to collect boxes , this is already implemented in nanobot/fsbot , it save the spots and then returns to that minimap area. Like that the bot learns from it's self and the behavior is diferrent every time the bot found boxes.

btw the bot will act like an human and i can apply this to my bot.
04/02/2016 23:48 mecr0don#130
Quote:
Originally Posted by lollo900 View Post
Naaah you can not bot bonusbox 24/7 anymore
yeah i bot 10-12 hours a day that make me 100k urid with bonus box skils and double boxes with good brain bot and other 10hours my pet collect boxes for no decrase and i have meet people who use tihs bot 24/7 in a week and no problem i dont do it bcs i collect more uri with my idea :)
04/03/2016 00:46 ~Demetrio~#131
Quote:
Originally Posted by mecr0don View Post
yeah i bot 10-12 hours a day that make me 100k urid with bonus box skils and double boxes with good brain bot and other 10hours my pet collect boxes for no decrase and i have meet people who use tihs bot 24/7 in a week and no problem i dont do it bcs i collect more uri with my idea :)
He means if an account is connected 24h without get discconected ,your account get Flagged and temporaly banned till BP check your account.
04/03/2016 00:48 topastop#132
Quote:
Originally Posted by ~Demetrio~ View Post
You know about Neural networks there's many ways to appply this , as a guy already said this is a extensive topic , so as you said would be hard for you Apply this in darkoribt , if you think you cant ! why say that anothers guys can not?

The bot can learn about it's experiencies saving all in a backup! for example apply this to Enemys and better ways to flee from them!

An example if a enemy with user id :XXXX Kills my ship more than 2 times , the bot will recognize the user iD and then will try to find a way to escape based in it's pasted experiences.

Or about boxes the bot could know the best Spot to collect boxes , this is already implemented in nanobot/fsbot , it save the spots and then returns to that minimap area. Like that the bot learns from it's self and the behavior is diferrent every time the bot found boxes.

btw the bot will act like an human and i can apply this to my bot.
I am with you when you say " if you think you cant ! why say that anothers guys can not?".
What I am trying to say is that the technology is not there yet, or in this case, the learning algorithms.
The main problem that you have is that you cannot provide samples to the bot to learn. You can teach him to identify boxes, you can teach him to collect boxes, but you CAN'T teach the way he will do these things. You can't teach him to act like a human, this is the real problem.


Anyway, if anyone makes such bot and is looking for a job, please send me a PM. We have somewhere around 25 people doing research in this area on one of the most prestiged institute. I am pretty sure you will deserve more recognition.
04/04/2016 18:39 martinhacker#133
Maybe they can detect some bots why you think people say that Chinese people are to intelligent xD and that video´s on youtube will make them to do a change as maybe do a max of uridium per day collecting bonux boxes so all is possible or maybe remove 2D who know´s
04/04/2016 20:24 Zetadarus#134
Quote:
Originally Posted by KickMayer View Post
Or make DarkOrbit Beta with the OLD lasersystem, when Basic FE was the best, people was like "Oh fack! 33k with X4 ZOMG!" and remove Havocs, Formations (3D), LF-4 and the upgrade system :)

I suspect ALOT of people would come back and start playing if that happend :)

it's my point of view, but i think it wouldn't

The game was nice before , but peopole got used to new stuff , if everything becomes like before , the game would become boring ,look old , and only veterans will stay wich would make the game even more empty .

I actually prefer the game the way it is now , for me it's not even comparable to before .

it's just that BP doesn't do as much advertising as before , i remember when i started Darkorbit there was ads everywhere about this game

now there's just no advertising about it.
04/04/2016 21:48 hugomailliot#135
Quote:
Originally Posted by Zetadarus View Post

now there's just no advertising about it.
now there is adblock :D :D