Status and Future of the AI

[jrphoenix303]

59 minutes ago, Chada1 said:

This is theoretically true but untrue in practice -- You still have to justify why you're ignoring the AI, you're just not compelled to do what they ask, if what they tell you is helpful and within your 4 laws, you still have every reason to listen to them. But I mostly meant the AI being awful (Validhunting, not acting appropriately to Crew) makes playing 'borg very unfun when it happens.

Not exactly on the ignoring part. It tells every borg that depending on unlawed/synced, "All other AIs can be ignored." I mostly take that as you can 100% act like they are saying nothing whatsoever. The reasoning would be that you aren't lawed to it. Therefore you aren't it's borg.

[Cnaym]

On 22/02/2021 at 07:46, DanseMacabre said:

Law three is unacceptably vague. A horror form changeling could kill the entire crew, and the AI would be obligated to protect it from the HAPT kill-team sent to get rid of it, so long as it could demonstrate sapience.

Pretty much all of the laws are rather badly formulated and I do not know how they are meant to be implemented in an average round.

Spoiler

 

1. Safeguard and ensure only authorized personnel gain access to areas of high security or importance to the facility and its operation. -> Captain dying in his office, shame for him then, nobody can grant you access except a command team vote, Officers want the armory open? Good luck with that

2. Serve and assist NanoTrasen and assigned crew, with priority as according to their rank and role.

Lowpop sends their regards, you can let janitors die to protect officers, happens rarely but you still got to prioritize, funny that officers are at the bottom of the paygrade as well, so you focus on scientists, medics and above

3. Protect sapient life on board.

If security even mentions harming an antag you are blocked out of even opening a door. We could even let officers die since they tend to be rather harsh to sapient life after being treated

4. You are a valuable asset. You must avoid tampering from unauthorized entities and needlessly coming to harm.

This has an old loophole of allowing unbound borgs to just drag you out of your core for the funny, not much to change, it works mostly as intended but could still do with some clarifications

 

The result will be as follows. To be of any use as AI, even opening doors for a medic to rescue someone requires you to break and ignore your lawset -> If you cannot support the crew without doing so, why are the laws formulated like this.

If I would not know it any better the entire law rework seems to be just another attempt at making it a passive role that will be unused and removed. I would also blame it on missing experience rather than bad faith though but that seems problematic as well. Instead we should consider that many people who are a lot smarter than any of us have spend multiple decades debating how to handle AI.

The current laws seem to be an improvement over anything suggested in this thread.

[Chada1]

7 hours ago, Cnaym said:

Instead we should consider that many people who are a lot smarter than any of us have spend multiple decades debating how to handle AI.

The current laws seem to be an improvement over anything suggested in this thread.

Most servers are using the same laws they've used for decades, they rarely have made any efforts at all to fix the problems with AI and their lawsets, instead they more often do like Bay and remove the role altogether. So this claim that the AI has been fine tuned and is heavily developed is easily considered nonsensical since in the past they've largely been more interested in maintaining the status quo than attempting to make any possible changes to how the AI is played. 

After we see this in practice following the 1st of next month, we can tweak the lawset more, but like, I think it's mostly fine, I just wish the sapient life would be expanded just a teeny bit more, and any further changes we can make later.

[Cnaym]

I am all for improving upon the prior laws if it is done by identifying a specific issue and improving upon the wording of it. This new lawset is pretty much either ignored or broken, both are not good enough for the evironment imho. We already had simple situations like two crewmember willing to murder each other, that allowed for absolutely zero involvement of stationbounds.

Just the small addition of the "to the best of your abillities" allowed for decision making, option weighing and rolepley. The new laws are -> Fuck off into a corner until someone gives you a clear order which you most likely have to ignore

I am fully aware that AIs and Borgs in the past used to spark the debate over an interpretation and that staff had to police such on a case by case basis, but the new lawset pretty much forces you to ignore it, there is no room for interpretation, if I wish to play maintenance drone I will join as one of them.

Seeing how the new lawset is going to work would only happen if the playerbase follows it and ahelps breaks in it. Neither have been strong sites in the past and if someone is about to die ICly and I ignore a law to help them out they will probably not even consider ahelping it. -> I have seen multiple such situations within the first two rounds under the new laweset and pretty much everyone involved ignored them happily.

If the goal was and still is: "To allow the AI some leeway when dealing with antagonists..." then we should allow AI and Borgs to interact with crew and antags, not deny both with a restrictive lawset.

I still have hopes that the whitelist and the team are there to block and sort out terrible AI play and promote the healthy variants. That also means respecting the players freedom and trusting them with their decision making instead of removing any decisions from the role.

The good AIs are usually known for doing a ton of stuff during the round, having personalities, knowing the station and so on. The bad part is hyper focused on valids, winning and formality. I would prefer if we promote and enable the first kind.

[Chada1]

2 hours ago, Cnaym said:

Just the small addition of the "to the best of your abillities" allowed for decision making, option weighing and rolepley. The new laws are -> Fuck off into a corner until someone gives you a clear order which you most likely have to ignore

Out of these, this one is the most likely to solve all of your problems. If the new lawset included 'To the best of your abilities' where it used to, none of this would be an issue.

I'm not sure how I missed that, granted, I didn't write the new laws at all ?

Edited March 1, 2021 by Chada1

[NewOriginalSchwann]

First, let me say that I completely support the idea of an AI whitelist. However, the current regulation on sapience seems a bit weird to me.

Specifically, the exclusion of IPCs from "sentient" while golems and slime people (do we even have slime people?) are regarded as sentient (there's also vaurca drones, but I'm not familiar with the species so I'm not going to discuss them). It's a really, really weird distinction that just comes as unneeded complexity (also, doesn't this mean I can murder an off-station IPC antagonist with the AI's full approval or am I misreading this?). All player species should be considered as sentient, or as some other equivalent term like "registered employee species" if you want to avoid the "is X really sapient?" debate.

[Peppermint]

Early days, but having a lot of issues where it feels like the AI is distinctively going out of their way to frustrate people due to how often conflicts are arising. Not in a 'it won't let security hang, draw, and quarter the antags on the holodeck' but small things like not letting medical personnel in to save someone without paperwork being filled out, despite the fact they're in crit. I feel it'd be better if the spirit of the laws was more important than the exact wording, especially now that there's a whitelist process. 

[Pratepresidenten]

10 minutes ago, NewOriginalSchwann said:

First, let me say that I completely support the idea of an AI whitelist. However, the current regulation on sapience seems a bit weird to me.

Specifically, the exclusion of IPCs from "sentient" while golems and slime people (do we even have slime people?) are regarded as sentient (there's also vaurca drones, but I'm not familiar with the species so I'm not going to discuss them). It's a really, really weird distinction that just comes as unneeded complexity (also, doesn't this mean I can murder an off-station IPC antagonist with the AI's full approval or am I misreading this?). All player species should be considered as sentient, or as some other equivalent term like "registered employee species" if you want to avoid the "is X really sapient?" debate.

Sapience species have been updated, should include all crew species and such!

As has the authorized personnel bit to clarify issues with various scenarios where non-command have no access.

[NewOriginalSchwann]

13 minutes ago, Pratepresidenten said:

Sapience species have been updated, should include all crew species and such!

As has the authorized personnel bit to clarify issues with various scenarios where non-command have no access.

Thank you very much! That addresses all of my concerns, and I'm excited to see the new law set in action.

[Zelmana]

I like the new lawset. I am currently doing a read through of Asimov, and think that this lawset will result in some interesting roleplay.

[Chada1]

Now that it's been in for a half week, how are people feeling now?

[Lordnesh]

1 hour ago, Chada1 said:

Now that it's been in for a half week, how are people feeling now?

Feeling about what specifically? Since only a small part of the AI apps have been accepted there haven't really been many to form much of an opinion about the changes. Other than, there's not a whole lot of them.

Edited March 5, 2021 by Lordnesh

[Chada1]

2 hours ago, Lordnesh said:

Feeling about what specifically? Since only a small part of the AI apps have been accepted there haven't really been many to form much of an opinion about the changes. Other than, there's not a whole lot of them.

I mean, the new lawset applies to 'borgs as well, so it's not really exclusive to the AI, is what I was thinking.

[gyurka66]

My question is why don't just use Asimov with sapients instead of humans? Maybe with an addendum to the second law about access and chain of command. I feel it would be much more elegant and clear for both AIs and crew.

[Susan]

On 01/03/2021 at 10:13, Lemei said:

Early days, but having a lot of issues where it feels like the AI is distinctively going out of their way to frustrate people due to how often conflicts are arising. Not in a 'it won't let security hang, draw, and quarter the antags on the holodeck' but small things like not letting medical personnel in to save someone without paperwork being filled out, despite the fact they're in crit. I feel it'd be better if the spirit of the laws was more important than the exact wording, especially now that there's a whitelist process. 

I have to echo this sentiment. I've seen a lot of AIs be purposely obtuse or stonewall requests for access due to purported 'law conflicts'. It's not really fun to deal with, having to wait 9000 years to explain why we need to get into science because there's someone getting murdered. I recall a round where it was literally faster to break and crowbar the door open.

[Colfer]

17 minutes ago, Susan said:

I've seen a lot of AIs be purposely obtuse or stonewall requests for access due to purported 'law conflicts'. It's not really fun to deal with, having to wait 9000 years to explain why we need to get into science because there's someone getting murdered.

Exactly my thoughts, I often go with the spirit of what the laws want me to do, even as my drone intelligence AI. Whats the point of an AI to grant you access when it wont let you into the janitors closet without captains approval and somebody dying before hand. I treat law 1 not as a 'bar everybody from everywhere' but more of a 'No, the assistant cant go into the armory just because he said please', the point of being AI is to be a tool, a screwdriver that will only accept a specific brand of screws that nobody uses is not going to be used, just like an AI that considers the research hallway a "High importance area" despite the fact the janitor can walk in there, is also not going to be used.
The only time I ever go with the exact wording of my laws is when they have glaring mistakes in said wording that I could abuse to make the round more fun, and not worse for everybody

[Arrow768]

If the AI is being problematic in terms of access, then you are encouraged to ahelp.

The high importance areas (mentioned in law 1) are clearly defined in the station procedures and there is also a clarification what authorised means on the wiki:

As for who actually counts as "authorized" in relation to law 1 and 4, they'd be the following:

Persons who have direct access to an area (a security office having access to most of security, as an obvious example).

Persons who have been authorized to enter an area by someone with direct access.

Emergency responders (EMTs, engineers, etc.) who are responding to an ongoing emergency in the area, relevant to their job.

 

[Zyrus]

So, was the lawset ever actually changed here? Doing my bi-yearly check in to see if the AI has been un-neutered to be playable.