Jump to content
Maxi

OSBot 3.0 AI Transition

Recommended Posts

On 9/3/2021 at 8:48 AM, Patrick said:

only whats necessary will be public. All the serious code will be hosted on a server and is closed source.

Can you elaborate more how us ordinary plebs who can make some broken code scripts for personal use will have to script with new client? :) Still write code using methods like getAmount, getPlayers.closest() etc.?

Link to comment
Share on other sites

  • Developer
12 hours ago, Kramnik said:

Can you elaborate more how us ordinary plebs who can make some broken code scripts for personal use will have to script with new client? :) Still write code using methods like getAmount, getPlayers.closest() etc.?

Initially the hive botting will only support live (SDN) scripts. Adding support for it will be like writing a more detailed script manifest.

I don't have an exact idea on what the behaviour AI will look like in the API when we get to that point. It might just cut down on code you need to write for a script (eg. interact("Tree", "Chop-down") will know you need an axe and get it if needed).

Link to comment
Share on other sites

Maxi, this AI is very impressive work. Anything is better than nothing, I say. Do you think the AI might also be used in other areas, such as "ban cause detection"? You could probably teach it to somehow analyze which accounts get banned and which don't, and what variables they did not share in their "personification" profile.

Also, I'm curious. Would we have the option to "train" this AI by playing on our own? The "human profile" could be founded in the bot owner's personal play style, and then the AI could "fork" from there. I imagine this could have a lower detection rate as opposed to having a randomly generated personification profile that could just as easily be generated by two bots on opposite sides of the planet at the same time, if you know what I'm trying to say. Who knows, maybe forking a human's behavior would increase the detection rates for all I know.

Anyways, this is impressive and I can't wait to try it.

Link to comment
Share on other sites

11 hours ago, Patrick said:

Initially the hive botting will only support live (SDN) scripts. Adding support for it will be like writing a more detailed script manifest.

I don't have an exact idea on what the behaviour AI will look like in the API when we get to that point. It might just cut down on code you need to write for a script (eg. interact("Tree", "Chop-down") will know you need an axe and get it if needed).

Will scripters starting from rank one will have access to API? 

Link to comment
Share on other sites

On 10/18/2020 at 2:39 PM, Maxi said:

Dear community,

OSBot 3.0 AI Transition
The last two months we have been thinking, conjuring ideas, to improve the OSBot experience for everyone. Scripters, recreational botters and larger scale operators all have their own goals and requirements. Improvements in user experience is defined differently for each party. We have tried to address all angles and came up with a series of improvements that we have started working on. Resources (CPU, memory etc) requirements becoming lower, MirrorMode stability, RuneLite support, server side WebWalking, new UI and UX for the client and website ; these are all features that are well within the possibilities, innovative but not groundbreaking. We will call the completion of these features Phase One.


We want to go a step deeper, eventually, introducing complexer forms of AI to further reduce ban rates, user experience and efficiency. We will call the completion of these features Phase Two. We have big plans and will release more insights over the months to come.


Once everything is in place, Phase One and Phase Two are completed, we can mark the birth of OSBot 3.0: Autonomous artificial intelligent botting that creates a whole new level of experience.


RuneLite support
You have been asking this for a while and it is here. Expect it to be released this week as we are finalizing tests. We hope you will enjoy the compatibility and that future updates of RuneLite will not focus on breaking this compatibility. We advise against using RuneLite as we have found evidence that certain features in RuneLite could be used in bot detection.


MirrorMode stability
Many of the issues encountered with MirrorMode stem from certain incorrect assumptions made on availability of data at will. MirrorMode is a feature that literally copies the state of the game, brings it in to a seperate container where data is transferred to and eventually can be used to analyze and produce actions which are then passed back out of the container in to the original gameclient, leaving no detectable traces of the bot client. This means that if you, as a scripter, use fixed timers and expect action a to result in outcome b, you will run in to problems that break your script. Much of this requires understanding of how exactly MirrorMode works and for obvious reasons not everything can be disclosed. After all, we are the only client out there that has true Mirror technology.


So how do we know for certain where certain issues stem from? What we will do is create a series of scripts and additional features to lower banrates. Knowing the ins and outs of MirrorMode, if there are any bugs that are to blame MirrorMode, we will find them and sort them out. The result should be that any issues that technically stemmed from the MirrorMode implementation will be sorted and all scripts that were affected by this will have their issues solved too. That said, after reviewing many cases, we believe that plenty of issue arise from script implementations rather than MirrorMode issues. Going forward, we will address any issues that stem from MirrorMode itself and after that, we can know for certain what the source of ongoing issues may be.


OSBot Gold series
We know our scripters have done a great job at creating many scripts over the years. We want to know for sure that we can provide high quality scripts going forward and improve all existing scripts by creating our own range of official OSBot scripts. By creating our own range of scripts we will run in to any and all issues that relate to the OSBot client and sort them out. We hope that, if any issues remain with previously existing scripts, we can encourage scripters to step up their game and solve ongoing bugs that are no longer possible to be linked to our client base.


We want to support existing scripters over time to find these issues, whilst we work on our side to improve that what is on our end. The outcome should benefit the users, which in turn benefits the scripters as the userbase grows who have access to a range of official OSBot and third party scripts that are close to flawless.


Scripters won't be undercut by the official series, something we want to emphasize on to prevent confusion.


MirrorMode for all
MirrorMode has been a feature exclusive to VIP subscribers. This will remain to be for some time, but we want to make MirrorMode free for all users to decrease detectability for the collective. In return, many of the new features we are creating, will become exclusive to VIP members to ensure they get the most value out of their subscription. As we work on finding any and all issues on our end, whilst producing a series of scripts of the highest standard possible, we want everyone to benefit from zero technical detectability and by doing so automatically reduce the input our client gives to machine learning algorithms designed to find and identify botting behaviour, reducing bans on behaviour.


Personification technology
Every human is different. We are building shallow AI that will create a unique personality for every account. This means that every account will play the game differently, every activity they do will have a flavour of their own. This should greatly reduce detectability and combined with the technology mentioned in the next section, improve overall experience.


Autonomous learning hive botting
Bots focus on doing one thing for a long time. It's simply not humanlike. Humans make decisions and, as new information become available, re-evaluate their decisions over time. The botting scene has not innovated much in this regard. Aside from the mentioned Personification technology, which is focussed more on how we execute specific tasks in a unique way per person, we also all make different decisions based on our emotions, available information and preferences in terms of how we tackle our challenges and reach our goals. Humans interact with each other and consume new information, resulting in different behaviour. We set goals and then construct how to reach these goals in an efficient way, with varying degrees of efficiency over time as we learn.


We will start to build a basis to create a collective consciousness and leverage changes in information to create alternative, growing and changing itineraries of how accounts as individuals and a collective reach their goals. We will leverage this technology in the OSBot Gold series as we learn, test and verify the system and eventually make these features available to other scripts. The result will be that every account will reach their goals in diffferent ways, scrambling any sense of automated copied behaviour. Welcome to the future of AI botting.


Serverside WebWalking
This is a feature that is not so easily made available to everyone. Serversided WebWalking would greatly reduce the resources used on the client side, but one must understand that if these computations are not made on the client side, they will have to be done on the serverside and computational power simply costs money. We believe however that we can optimize this, where we can make calculations for the collective of accounts that on average reduce the total required computational power. This results in lower costs as a collective. We are working on how we can make this available at a cost effective rate as a premium feature to cover the costs, whilst making it available to you at a lower cost than having to do it on your own servers.


Other/misc
There are more features you have requested, but we need to be realistic and make decision for the foreseaable future where we deploy our limited time and resources. We believe the above plans, which is a subset of all our ideas and plans, is where we improve the experience of the entire community in a steady way. Must we fall short and there are topics you are missing, we can re-evaluate and consider changes, but we believe the above plans will make many people very happy.


Final words
We hope that everyone can be excited about these upcoming features and we hope to have everything in place, step by step releasing features and scripts, between now and the end of Q1 2021.

A big shoutout to everyone who has worked relentlessly to keep OSBot going,

Forward and onwards,
The OSBot team
 

MirrorMode for all
MirrorMode has been a feature exclusive to VIP subscribers. This will remain to be for some time, but we want to make MirrorMode free for all users to decrease detectability for the collective. In return, many of the new features we are creating, will become exclusive to VIP members to ensure they get the most value out of their subscription. As we work on finding any and all issues on our end, whilst producing a series of scripts of the highest standard possible, we want everyone to benefit from zero technical detectability and by doing so automatically reduce the input our client gives to machine learning algorithms designed to find and identify botting behaviour, reducing bans on behaviour.'

 

 

 

 

Are we any closer to getting mirror mode for non vip?

Link to comment
Share on other sites

  • Developer
16 hours ago, JaredFulton said:

Maxi, this AI is very impressive work. Anything is better than nothing, I say. Do you think the AI might also be used in other areas, such as "ban cause detection"? You could probably teach it to somehow analyze which accounts get banned and which don't, and what variables they did not share in their "personification" profile.

Also, I'm curious. Would we have the option to "train" this AI by playing on our own? The "human profile" could be founded in the bot owner's personal play style, and then the AI could "fork" from there. I imagine this could have a lower detection rate as opposed to having a randomly generated personification profile that could just as easily be generated by two bots on opposite sides of the planet at the same time, if you know what I'm trying to say. Who knows, maybe forking a human's behavior would increase the detection rates for all I know.

Anyways, this is impressive and I can't wait to try it.

Yes we're also looking to use the AI for tracking bans pretty much how you mention it.

Training the system based on human input is also something we're looking to implement 😛 

Link to comment
Share on other sites

10 hours ago, Patrick said:

Yes we're also looking to use the AI for tracking bans pretty much how you mention it.

You should make a "botting@home" distributed computing project for OSBot members to run in the background. Let it run bot experiments on their machines and log all the causes and effects. I'd run it if it means the quality of our software gets better. 😅

10 hours ago, Patrick said:

Training the system based on human input is also something we're looking to implement 😛 

Come to think of it, my "botting@home" idea could also include a feature that tells people to play on the account normally for a few minutes for the humanistic analyzer to study. From the behavior data, you would probably gather even more data than Jagex has in terms of how a human plays. You'd be able to see attention spans, reaction speed, decision trees, input habit variations... everything. And in my head, it doesn't seem like a complicated thing to gather, which is important. Gathering should be easy. The data gathered is timeless. All that matters in the early stages is that you have it, and you have as rich of detail as possible. You don't have to have good (or any at all) analysis capabilities right away. That data will be just as good a decade down the line as it was the minute it was recorded.

  • Like 1
Link to comment
Share on other sites

3 hours ago, JaredFulton said:

You should make a "botting@home" distributed computing project for OSBot members to run in the background. Let it run bot experiments on their machines and log all the causes and effects. I'd run it if it means the quality of our software gets better. 😅

Come to think of it, my "botting@home" idea could also include a feature that tells people to play on the account normally for a few minutes for the humanistic analyzer to study. From the behavior data, you would probably gather even more data than Jagex has in terms of how a human plays. You'd be able to see attention spans, reaction speed, decision trees, input habit variations... everything. And in my head, it doesn't seem like a complicated thing to gather, which is important. Gathering should be easy. The data gathered is timeless. All that matters in the early stages is that you have it, and you have as rich of detail as possible. You don't have to have good (or any at all) analysis capabilities right away. That data will be just as good a decade down the line as it was the minute it was recorded.

yeah botting@home or osbot@home or something like that would be awesome! Would definetly run that on my pc!

Link to comment
Share on other sites

6 hours ago, maximiliano1 said:

Hi, in which way you're going to use AI?It's quite broad term and doesn't tell the story. There's Deep Learning (pixel based or state based). Reinforcement Learning ...

I interpreted it as a fatigue system similar to that of Script Factory. I think the AI is "shallow" like they say in that it is only applied to certain things like wait times. I don't see how any AI could actually change script behavior without a rewrite of the scripts themselves, so that's best explained by them.

Link to comment
Share on other sites

  • Developer
14 hours ago, JaredFulton said:

I interpreted it as a fatigue system similar to that of Script Factory. I think the AI is "shallow" like they say in that it is only applied to certain things like wait times. I don't see how any AI could actually change script behavior without a rewrite of the scripts themselves, so that's best explained by them.

The first release will indeed not change the actual script behavior, it's more like a 'smart bot manager' which will learn over time at which times certain activities can be done best.

The personification AI will introduce API changes that can change script behavior (Think like slower over time to simulate getting tired, skilling script might afk more during working hours etc, but also the way how people interact with objects (camera/mouse movement, walking etc))

First one will be trained via botting, second one by human playing.

Link to comment
Share on other sites

6 hours ago, Patrick said:

The personification AI will introduce API changes that can change script behavior (Think like slower over time to simulate getting tired, skilling script might afk more during working hours etc, but also the way how people interact with objects (camera/mouse movement, walking etc))

First one will be trained via botting, second one by human playing.

I've noticed that Script Factory has a similar feature already out right now. Can you elaborate on its sidedness (server or client), and how long it will take to train? Or, will it be telemetry based, combining play data from thousands of humans and then using that to create an algorithm that can be seeded, or will it be stored locally and used only to seed local accounts? That way, a new customer would only need to connect and get a "personification profile" from the server, but it runs the risk of being similar to anyone else who connects. I'm not an expert so I don't know what the sweet spot is between individuality and convenience.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.



  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...