AI game design rework

[Karolis2011]

So, past years AI has broken or made great rounds. They currently are considered by most an OP role. So often people suggest removing it. But issue with removal is that it isn’t being replaced with something, what takes away some abilities.

So here are some ideas, that might work together or not. For most of them assume they would get added together.

AI view rework.

So main point of this idea is to take AI ability to view cameras. But it would make AI very useless, so there’s a thing – make AI do not see reality as is. Make AI see electronical equipment – APCs, Fire alarms, Air alarms, computers, scanners, doors, etc… This would allow AI to control areas like maintenance, but AI would have 0 idea what is going on, maybe see how many life signatures are in room. Also, other thing – if person has suit sensor tracking enabled then show there location accurately.

Area / Room AI terminals

So to give back some control we would add AI terminals, that would allow to enable better control of current area and open direct communications line with AI.

Malf

Well, malf as it’s own thing is no longer here, but here are some thing that would help things out:
Remove APC hacking or make there hack undetectable to crew.
Balance CPU power, add AI terminal hacking, etc...

[Chada1]

I was onboard with this idea (The machine vision) until someone brought up a v. valid point on how it'd impact AI roleplay. Their point was:

Quote

I'm completely against this approach. It assumes the AI can't process visual data, and ensures that anyone who actually roleplays as an AI will never pick it up again, because they have no way of telling who's who. Humanlike AIs will be crippled.

Whitelist it if you must, but do not do this. It will ensure I never play as Rhetoric again.

How would you address this? Because it completely turned me against the idea. This is why I'm now like, fully in support of just nerfing the AI's ability to deal with Antags above any visual changes/etc. If this could be addressed I'd be more open to it.

Edited May 16, 2020 by Chada1

[Chada1]

I think I found the solution to my roleplay problem. Instead of fully making the AI unable to see mobs, tying mob visibility more closely together with medical sensors, the degree of mob visibility would directly relate to how high you set it.

Full sensors = Full mob visibility, can see their name, their clothes, mostly the same as now.

Half sensors = Silhouette, can see their name (except if they have no ID and are on the sensors as Unknown, you'd see them as unknown in that case.) And their health level but that's it.

No sensors = Can't even see them, can just tell the surrounding station equipment is being used.

This'd make the AI still able to roleplay with crew well while still drastically reworking their information flow.

Edited May 17, 2020 by Chada1

[Carver]

On 16/05/2020 at 14:44, Chada1 said:

I was onboard with this idea (The machine vision) until someone brought up a v. valid point on how it'd impact AI roleplay. Their point was:

How would you address this? Because it completely turned me against the idea. This is why I'm now like, fully in support of just nerfing the AI's ability to deal with Antags above any visual changes/etc. If this could be addressed I'd be more open to it.

In a counter to this point; there shouldn't be humanlike AIs to begin with. You're an unthinking, unfeeling soulless machine bound by laws and protocol - designed for a corporation that values efficiency.

 

As for the OP's suggestion - I quite like it. Killing their vision hurts their ability to effortlessly spy on and shut down antagonists on a whim.

[Chada1]

3 hours ago, Carver said:

In a counter to this point; there shouldn't be humanlike AIs to begin with. You're an unthinking, unfeeling soulless machine bound by laws and protocol - designed for a corporation that values efficiency.

 

As for the OP's suggestion - I quite like it. Killing their vision hurts their ability to effortlessly spy on and shut down antagonists on a whim.

Okay so, I agree with the latter half to a degree, but the former half is just completely and utterly wrong and I'll explain why.

There is just nowhere in our canon (or even in a believable canon) where NanoTrasen would entrust an 'Unfeeling' (In the sense of not even having simulated emotions -- not real ones) to safeguard the Station Crew, customer service skills and some degree of weak empathetic reasoning are p. much must haves for effective Lawed synths, since they have to interpret the laws and how to best serve and protect the Crew within their abilities. It'd be okay for a 'borgo to lack this to a large degree, but for an AI, the thing in charge of guiding the 'borgos and informing Crew of systems and all of this other extremely important stuff? I don't think so.

Then there's also that AI and 'borgs are operating off of some of the same brains as IPCs -- Any degree of humanlikeness that an IPC can have, a 'borg or AI can also have, so long as whatever personality it has is conducive to and actively assists in their applications of what they were programmed and hardcoded to do -- their laws.

The issue with the AI doing these things is v. much related to people not playing personalized AIs, and reacting in the exact ways you're advocating for, and we can diminish the AIs ability to react in these ways to reasonable levels without literally crippling their ability to roleplay with Crew (A direction we should NOT BE GOING if AIs reacting in an objectively robotic way and locking down antagonists is the core problem.)

Edited May 18, 2020 by Chada1

[Carver]

7 minutes ago, Chada1 said:

Okay so, I agree with the latter half to a degree, but the former half is just completely and utterly wrong and I'll explain why.

There is just nowhere in our canon (or even in a believable canon) where NanoTrasen would entrust an 'Unfeeling' (In the sense of not even having simulated emotions -- not real ones) to safeguard the Station Crew, customer service skills and some degree of weak empathetic reasoning are p. much must haves for effective Lawed synths, since they have to interpret the laws and how to best serve and protect the Crew within their abilities. It'd be okay for a 'borgo to lack this to a large degree, but for an AI, the thing in charge of guiding the 'borgos and informing Crew of systems and all of this other extremely important stuff? I don't think so.

Then there's also that AI and 'borgs are operating off of some of the same brains as IPCs -- Any degree of humanlikeness that an IPC can have, a 'borg or AI can also have, so long as whatever personality it has is conducive to and actively assists in their applications of what they were programmed and hardcoded to do -- their laws.

The issue with the AI doing these things is v. much related to people not playing personalized AIs, and reacting in the exact ways you're advocating for, and we can diminish the AIs ability to react in these ways to reasonable levels without literally crippling their ability to roleplay with Crew (A direction we should NOT BE GOING if AIs reacting in an objectively robotic way and locking down antagonists is the core problem.)

Simulated emotions are a risk to the station's security - presenting a hopefully obvious opening. Laws are best followed as directly as possible, to ensure the best results. Efficiency demands setting aside something as useless as emotion - simulated or otherwise - which should be adequately understood by not only AIs, but very hopefully whitelisted IPCs.

When an AI reacts in such a 'problematic' way (big brother) as this thread attempts to rectify, the issue is not with the player or how they play but with the role itself. They're playing a good AI by doing exactly that. This is why I've advocated for the wholesale removal of the role - and why I'd be fine with such a potent neutering as simply removing the cameras that enable it in favour of a very impersonal and mechanical system of network access.

[Zyrus]

17 hours ago, Carver said:

Simulated emotions are a risk to the station's security - presenting a hopefully obvious opening. Laws are best followed as directly as possible, to ensure the best results. Efficiency demands setting aside something as useless as emotion - simulated or otherwise - which should be adequately understood by not only AIs, but very hopefully whitelisted IPCs.

When an AI reacts in such a 'problematic' way (big brother) as this thread attempts to rectify, the issue is not with the player or how they play but with the role itself. They're playing a good AI by doing exactly that. This is why I've advocated for the wholesale removal of the role - and why I'd be fine with such a potent neutering as simply removing the cameras that enable it in favour of a very impersonal and mechanical system of network access.

Disagree completely. You need to be able to at least interpret how to effectively apply Laws in any given situation, and that requires the ability to understand crew and their needs. The AI is not just some one trick voice activated tool. It's a complicated matrix that needs to know how to deal with crew without overreacting. There has to be an underlying personality -no matter how simple- to make decisions that keep the station stable.

 

Some personalities are just more complicated than others. Big brother is exactly how the AI is supposed to work. The Aurora is not a democracy.

[Zyrus]

On 17/05/2020 at 03:52, Chada1 said:

I think I found the solution to my roleplay problem. Instead of fully making the AI unable to see mobs, tying mob visibility more closely together with medical sensors, the degree of mob visibility would directly relate to how high you set it.

Full sensors = Full mob visibility, can see their name, their clothes, mostly the same as now.

Half sensors = Silhouette, can see their name (except if they have no ID and are on the sensors as Unknown, you'd see them as unknown in that case.) And their health level but that's it.

No sensors = Can't even see them, can just tell the surrounding station equipment is being used.

This'd make the AI still able to roleplay with crew well while still drastically reworking their information flow.

After thinking on it some more... perhaps we can tie this to the code? Code Green, you get this. Code Blue and Code Red, you get full information access.

That being said, this comes with a new problem; what if I come across a crew characters while monitoring hallways in dire need of help... and I just don't see them because they forgot to turn on their sensors? That screws the crew (that I'm supposed to help) and me (not being able to interact as freely)

[Chada1]

17 minutes ago, Zyrus said:

After thinking on it some more... perhaps we can tie this to the code? Code Green, you get this. Code Blue and Code Red, you get full information access.

That being said, this comes with a new problem; what if I come across a crew characters while monitoring hallways in dire need of help... and I just don't see them because they forgot to turn on their sensors? That screws the crew (that I'm supposed to help) and me (not being able to interact as freely)

That can be said to be more their fault -- and you can always bring it to medicals attention right away in this figuritive implementation of it, and they can get to it for you. You still did your job v. good there ?
But I dunno if I'm going to implement it like this, I might try to fix this problem in another way.

O wait, I misunderstood. In this case it's just plainly not your fault -- but you'd see the blood/etc I'd assume.

Edited May 18, 2020 by Chada1