Sentient robots in the GH universe

[Joseph Hewitt]

Edit- Fixed the title. What the heck is a "fobot" anyhow?

I'm splitting this off from the "Unlockable Faction Mecha" thread over in Game Mechanics.

Quoting: peter

That actually seems like a bit of an inconsistency in the GearHead universe. If you can build sentient robots after a one-day multi-million credit training course, why aren't there more sentient robots running around. Once you know how to build one, you can do it in less than an hour (two minutes, actually) and for a cost of less than 10000 credits. The corporations have the resources to make tons of them.

In the GearHead universe, there's nothing that a strong AI can do that a weak AI can't do better. Corporations (and universities, and governments...) use weak AIs for many different purposes but there simply isn't much economic incentive to build sentient strong AIs. Consequently, most of the strong AIs that exist were built by either researchers or hobbyists.

Mechanical weak AIs are not frequently used as mecha pilots for a number of reasons, including the fact that they can be undependable, don't react well to new situations, may be restricted by law in certain areas, and of course there's always the chance they'll run amok or otherwise cause huge amounts of damage/public relations nightmares.

After showing a robot he created with stats in the 40-80 range and skills mostly in the high 20s
Do robots this far superior to humanity exist in the Gearhead universe?

That far superior to humanity? Says the man with the Robotics score of 40...

No, robots that good don't exist in the GH universe. In fact, you'll notice that there are no NPCs with skills much higher than 15. That's what I consider to be the highest "natural" skill value. The highest value I ever got personally was a Mecha Artillery score of 17, once.

I guess this means it's too cheap, quick, and easy to boost your skills to mega-epic levels. Maybe I should apply a level cap, or increase the XP requirements dramatically after skill level 15.

The best robots that exist in the game world should be about the same level as those that can be built with a robotics skill of 15-20, which looking at your results are much tougher than I had thought.

Did the corps build some and then find out that the robots didn't want to do their evil deeds for them? So the robots ran off and joined the Cyclopses (what is the plural of Cyclops? Joseph, aren't you an English teacher?) Or were they able to make them sentient but aligned favourably for the corp and willing to work for not much pay? They don't eat, they don't sleep (but neither does anybody else...), and they're super tough. Why aren't all corporate pilots super-robots?

I use cyclops as both singular and plural. According to the dictionary, the plural should be cyclopes. I think I'll stick with cyclops.

Non-sentient robots do most of the heavy work in the civilized parts of the GH universe, sometimes with human operators and sometimes without. There may be laws similar to those found in Mega-City One, where it is illegal to hire a robot for a human job and also illegal to hire a human for a robot job.

[Sabin Stargem]

In Isaac Asimov's books, robots typically follow three, perhaps four cardinal rules.  They have loopholes, but they cover the most basic principles where the dangers of robotic AI is concerned.

Rule One  :  An robot may not harm nor kill a human.
Rule Two  :  An robot must obey human orders.
Rule Three:  An robot must preserve itself, unless this conflicts with rule one and rule two.

I imagine that if such rules were to be implemented into actual robots, that there would be 'sub-rules', not unlike how it was in Robocop.  After all, what if an competing company were to send someone to another company and ask an robot to give details about financial records?  Obviously, this is where sub-rules come in, and they would help determine whether or not certain humans can order certain activities on the part of an robot.

Now, from what I have read about your viewpoint on weak AIs in Gearhead, I must agree with it, though I think that in certain situations weak-AI robots can be useful.  For example, what about automated turrets?  They can detect whoever is in the hall that they are guarding if whether or not that person has an radio-ID badge that tells the turrets to stand down.  If there is none, then the robots take out whatever lifeforms are present, in whatever way possible.  Weak AI robots can be placed into certain positions and situations where they don't need to be complicated to carry out their work.  Both military and civilian.

Of course, that is the problem:  When something unpredictable happens, an weak AI is doomed to make major mistakes.  For example, what if an wall nearby is destroyed, allowing them line of sight with an medical bay with 'unauthorized' persons?  In that case, the turrets would probably fire at the hapless civilians in that section, which is pretty bad, especially during a battle.  What if the enemy has the right IFF codes?  In that case, an weak AI is pretty much worthless.

That is why developing 'higher' AIs is necessary for military and civilian applications.  For example, an AI to manage the flight and launching schedules of planes and vehicles, and to make sure everything is going as planned, whether it be weather, security, or military situations.  As such, an strong AI that can communicate with other kinds of AIs would have great civilian and military applications.  For example, take an 'Chessmaster' AI that is the 'general' of the military.  It is notified about an battle taking place, and notes that a civilian shuttlecraft is headed that way.  It tells the human generals about the situation and gives recommendations, and once allowed to carry out an operation, it can instantly communicate the plan to the Aerobiz AI that manages the civilian shuttlecraft route. (Basically telling the plane to change course, for one reason or another as fits the situation.)

Humans can't really work together quickly and efficiently, at least not when compared to robots.  Yet, humans are far better than robots because we can analyze an situation in an way a robot can't.  However, if one were to perfect an AI capable of matching human adaptability and perception, then it is an different ballgame.  While it is an dangerous idea, there are many people who can benefit from an higher AI.  But what of projects that seek to make an AI, but in physical form?  That makes things even more risky, especially if the robot were to go rogue.  No doubt, someone somewhere in the Gearhead universe doesn't want anyone to pull it off successfully.  The fact that the player could find such robots, or perhaps even create them would put them in an anti-robot group's sights.

[peter]

Quoting: Joseph Hewitt

I guess this means it's too cheap, quick, and easy to boost your skills to mega-epic levels. Maybe I should apply a level cap, or increase the XP requirements dramatically after skill level 15.

 I think just continuing the exponential increase of XP requirements would do the trick.  Currently, each rank after around 14 costs 37700XP, which is good for rank 14, but too cheap for rank 20.  This would make Hwang-Sa's promotion rewards (+1 rank of Martial Arts, if you're spiritual enough) potentially even more valuable.

 Capping skill levels at 30 (or even 20) might also be a good idea.

Quoting: Joseph Hewitt

In fact, you'll notice that there are no NPCs with skills much higher than 15. That's what I consider to be the highest "natural" skill value. The highest value I ever got personally was a Mecha Artillery score of 17, once.

 Omega-1004 has an Androbot body, so he has Weight Lifting 30.  That's kind of a special case, since that's just how GearHead represents metal being stronger than human muscle.

 One of my characters, the spiritual heavy-weapons specialist monk, has leadership 20.  After dominating some animals, I had more lancemates than I could lead without an XP penalty, so it was off to Wujung University for him.  It was late-game for that character, so I had enough cash on hand to do the training right away.

Quoting: Joseph Hewitt

There may be laws similar to those found in Mega-City One, where it is illegal to hire a robot for a human job and also illegal to hire a human for a robot job.

 Well, there are crime bosses with both Watchmen and Androbots (and Hunter-X synths) guarding their evil mansions right now...  BTW, are they supposed to have tamed the synths, or do they just not go into their houses via the front door?

Quoting: Sabin Stargem

That is why developing 'higher' AIs is necessary for military and civilian applications.

 Yeah, smarter, but still not self-aware, AI.  Self-aware, or sentient AI will have desires and opinions and morals which makes it a lot harder to build one to do only the task you intended it to do.  It would also be morally wrong to enslave a sentient AI by building it into a gun turret (unless it had good Internet access so it could talk to people, etc.  Even then, it wouldn't have a choice about it's job being to shoot people/things.).

 You have a good point about robots being able to communicate quickly, though.  At this point, you're talking about software that can run on general-purpose computers, and probably in fact does.

 Joseph, I don't understand the claim that there's no business case to build sentient robots.  If the liability issues were ever resolved, what society wouldn't want firefighters that didn't need to breathe air, accountants that didn't get bored and could tally vast amounts of data in their heads, and so on.  Speaking of data, Commander Data on Star Trek is a great example of a sentient robot who's extremely useful to have around.  What corporation wouldn't build some robots in the hope that at least a fraction of them would be willing to work for the corporation.  The nastier factions might deactivate and disassemble robots that weren't willing to go along with them.

 If robot AI is not well enough understood to be able to build a lawful robot, or otherwise ensure it will work for you, I can see why they aren't widely used.  I still think there would be elite robot troops/spies/assassins/accountants/doctors/academi cs  with really high stats.  Maybe there'd even be SF:1 or SF:2 robot soldiers...  Yeah, there probably would be some robot scientists.  Once a researcher builds a sentient robot

 I'm sure some mad scientist somewhere has built a little robot army, or tried to.

[draconuszero]

On the sentient robot issue, chances are that if you have a robot that can think for itself and use a mecha, chances are that it would a) kill the master and run amok, b) become a walking Skynet, c) go rogue, or even d) kill itself realize its the lone sentient robot ever made unless it has robotics.

On a side note, wouldn't it be funny if you find a mercenary group led by a sentient robot/AI?  I recall that in the rogue squadron novels that the quartermaster for rogue or wraith squardron was a self-aware robot that was involved in a robot uprising.

[macksting]

Why would those be the "chances are" scenario? It seems to me, if it was a sentient being, it wouldn't have a considerable need to do such things; I think its dismantling as a result of simple defiance would be a more likely result. (After all, if I were the sort of master who gave orders which would cause my sentient robot to kill me, I'd rather head such problems off at the pass and just build another, more obedient, robot.)
Essentially, I don't see a reason, especially in an anime setting, that a sentient robot need necessarily go on berserk killing sprees. In anime settings, sentient (and even non-sentient!) robots tend to have feelings and emotions. Even lacking that, I think a really significant error would have to happen for any decently made AI to start killing indiscriminately.

Now, depending on what you mean by "go rogue," you could be quite right. However, it's also possible that the robot, like the ones the PC creates, is just as happy to stick around. Why not? What if one of its feelings is fealty?
Skynet, of course, isn't totally out of bounds... but it might not necessitate the killing of humans. Not every AI is Prometheus.

Really, though, there's room in almost any setting for errors and problems.

Hmm. Might I point out that remote control and weak AI technologies should be sufficient for lifesaving, and might be a lot kinder to the robots you're sending into harm's way? Mind you, from a robot rights perspective, a sentient robot might volunteer for such work, just for the greater respect accorded to a robot who willingly risks its life saving your cat. If robots were to unionize or band together to fight oppression, one could expect the most advertised and important of these to be the sentient life-savers.

I must admit, I'd rather like to see a rash of SF:1 and SF:2 robots with superior stats, just to throw into a fight.

And finally, as to the mad scientist building a robot army, I believe the man who wrote the robotics guide on our mailing list fills that niche nicely. Really, though, I suppose you're right...

[draconuszero]

What i meant by go rogue is pretty much just leave its creator to do hat it wants.

[Sabin Stargem]

There are several figures that I feel are major achetypes for robots and the concepts about them.

SHODAN:  The rogue AI of the System Shock games, she apparantly was quite benevolent before her ethical restraints were deleted by an hacker.  When they were removed, she quickly became increasingly paranoid and malicious over the months, likely following her 'rules'.  Which was to research, and to protect TriOptimum's station and data.  However, this in turn meant that she would have to destroy the humans aboard the station, in order to ensure security, plus she can use them as test subjects.  However, in her ethical ruleset she probably wasn't supposed to learn about certain things beyond a point, but without that ruleset she was free to learn much more.  Especially where personality, ethics, and morality was concerned...or an lack of such, rather.

MEGAMAN:  He is an robot, of the higher-level kind that borders on sentient, or perhaps IS sentient.   Created by Dr. Light as an assistant and project, Rock eventually became an fighting robot, to counter the robotic army that Dr. Wily fielded.  In each Robot War, Megaman fought at least 8 'Robot Masters' and Dr. Wily himself.  It is assumed that Megaman follows an ethical preprogrammed ruleset, though he is apparantly free to learn about things.  This is based on that Megaman X is an robot that was able to make decisions of all kinds, including ethical ones.  Prior robots apparantly lacked this, but Megaman probably was capable of not following his ruleset.  In the seventh Robot War, Megaman actually attempted to kill Dr. Wily, because Dr. Wily was an menace and needed to be eliminated once and for all.  Apparantly Megaman wasn't told to kill Dr. Wily, so he made the decision to do so himself.

It indicates that with learning robots, they can eventually disable or bypass their programming, probably through some mental flaw that can probably interfere or disrupt the programming that is supposed to keep an AI in check.  As such, an robot that is nearly sentient can BECOME sentient after an number of years or through experience.  Astroboy is another example of such, and as time goes by in his television series, he becomes less awkard and can do things a robot can't do - like lying, which was done to ease the worrys of an Human over an delicate subject.  So even ethical, honestly good robots can absorb human nature and decision making over time, in order to do what they consider to be 'right'.

HAL:  An learning robot, HAL was the AI of an ship with a few human crewmembers.  Given orders to kill them by HQ in secret, HAL was forced to do so.  His programming dictated that he was to follow HQ's orders, but it is also likely that HQ would have modified HAL's programming with it's latest transmission, so as to make HAL's objective clear and to prevent irrelevant actions.  So basically, such an programming alteration would be very specific and would probably be deleted or altered once the task has been carried out.  The benefit is that since the 'order' and the programming is recent, prior programming wouldn't be able to normally corrupt or alter it: it is too new, and nor is an time frame available for that to happen over a number of years.  It is probably safest this way, without compromising HAL's normal ethical protections (like not destroying the majority of humanity)

[Francisco Munoz]

IMHO some check should be added to the robotic creation (or a plot later in lancemates section) so your creation has a chance to dont follow you or even be your enemy.

(As a fun note: I managed to create a nice non-sentient robot from some Scotch tape)

[Joseph Hewitt]

Quoting: peter

Joseph, I don't understand the claim that there's no business case to build sentient robots. If the liability issues were ever resolved, what society wouldn't want firefighters that didn't need to breathe air, accountants that didn't get bored and could tally vast amounts of data in their heads, and so on.

Certainly there's a market for those things, but in GH there's no reason why such things couldn't be done by weak AIs. A human being working in concert with a weak AI, or augmented cybernetically with weak AI systems, would be just as smart as any strong AI could be.

The sentient AIs in GearHead don't tend to be quite as "super" as the AIs in other settings. Omega 1004 has a Knowledge score of 25, which is really really good but not completely beyond human ability. This may have something to do with the fact that most humans alive today have some genetic engineering in their family history.

Quoting: peter

Well, there are crime bosses with both Watchmen and Androbots (and Hunter-X synths) guarding their evil mansions right now... BTW, are they supposed to have tamed the synths, or do they just not go into their houses via the front door?

Well, the Hunter-X line does have some control mechanisms built in. These control mechanisms often fail tragically, true, but they're there. In the GearHead underworld owning a Hunter-X synth is rougly the equivalent of owning a great cat or a pit full of man-eating alligators and cows in contemporary crime fiction. Sure, it's dangerous and stupid, but it's muy macho.

Quoting: Francisco Munoz

IMHO some check should be added to the robotic creation (or a plot later in lancemates section) so your creation has a chance to dont follow you or even be your enemy.

I like this idea. I think I'll add it... It may be better if a robot would stay for a while, then decide to leave or not after that.

[peter]

Quoting: Joseph Hewitt

Certainly there's a market for those things, but in GH there's no reason why such things couldn't be done by weak AIs.

 Hmm, I guess that argument applies to all of my suggestions.  Only very esoteric situations couldn't be handled just as well by a non-self-aware AI.  So I guess I'll have to agree that it does make sense that there aren't robots running around all over the place.

 So what are some esoteric situations where only a robot capable of autonomous decision making will do?  Something that's physically too dangerous for a human, and where remote control by a human is not possible.  Either because of interference with communications, need for radio/everything else silence, or over distances so large that light-speed make the round-trip time too slow to react to things in real time.  So, a robot spy on Mars...  Or doing scientific research very deep in the ocean/under the sea floor, and you don't want to be limited by a cable.

[macksting]

Peter: I recommend doing some intense study of the role of civilian and military Boomers in Bubblegum Crisis. While they're so tightly biomimetic as to essentially be alive, Boomers are, nevertheless, given the roles a robot would be. Even more relevent, many are built with a limited AI.*
Amongst their other purposes are as bodyguards, where a weak AI might not be able to grasp hidden threats; as deep sea workers, where a certain level of autonomy is helpful and the range of remote control is dismal; as firefighters; as military weaponry; as deep-insertion moles; as lawnmower-like minesweepers; and as about 90% of the population of space stations.
While these functions may not necessitate a strong AI, the existence of strong AI's would probably place such entities in these tasks. (Except, of course, roles like the minesweeper, which is an essentially and utterly moronic device.)
In particular, in real physics terms, humans really, really don't adapt well to space. While concessions can be made for genetic engineering, physical augmentation, and the roles of evolution, mutation, and evolution, the initial settling of space would probably use a considerable number of robots, reinforcing their role in later centuries.
A strong AI can also be the backbone for several weak AI's, such as a single focal AI controlling or coordinating half a dozen small, expendable fighters. While the strength of this focal AI might vary, if you truly expect it to come home, and its obedience is "assured," there might be no reason to limit its capabilities by requiring human input. Machines tend to be faster, smaller, and less fragile in many ways than a human operator.

The Obedience Check: Seems to me this would be a second Robotics roll, possibly against the Ego stat of the robot at character creation. In GH2, this could easily be emphasized as a possibility for the character's Nemesis, although I think I'd have issue with that if it turned out to be particularly more powerful than the PC. Still, the battle would be involved and epic if, randomly, it were to occur. Naturally, I'd expect most disobedient "rogue" robots to simply wander away on their own little tangent, with only a truly significant failure turning against its master openly.

HAL: SPOILERS! Props to Sabin, but he got it a bit wrong. HAL 9000 was not ordered to kill its crew. It was ordered to keep them in the dark about the true purpose of their mission. For national security reasons, you see. This ran directly counter to other orders and rules in the AI's programming, and slowly created a genuine level of insanity and hypochondria. (HAL was grown mechanically, not programmed.) He unconsciously caused a malfunction in his signal dish because the HAL 9000 analog that was running back home for error-checking was serving as a painful conscience to a computer which was also trying to follow its orders.
When HAL determined that the two crew members were getting too close, the portion of his mind which was fairly well snapped killed Poole. This necessitated awakening one of the other crew members to take his place... and Poole and Bowman were the only crew members who did not know what the mission was about, so awakening any other crew members would result in Bowman surely finding out about the nature of the mission.
So HAL killed them all by causing cascading power failures and a loss of atmosphere. Still insane and largely quite upset about everything, he was truly very, very disturbed by the chance of being shut down; having never slept or been turned off even for a moment, this was a cessation of function for which he had no context. Death.
This is why you should always program your AI's instead of growing them, either mechanically or biologically. And why every ship should have an AI psychiatrist on board.

*It's one of those settings where a single, brilliant mad scientist caused a technological leap with which the world has not yet caught up. Boomers can be reproduced, and even improved, but as early as 2027, there's still nothing in the technology which adequately replaces a human template AI for their control. Model-9's, the more mechanical brain which was made to handle that need for weak AI, don't seem to be present in 2027.
Nor, of course, do the more mechanical physical specimens seen in 2034. Essentially, everybody's just carbon-copying Dr. Stingray's robots, which are more alive than not, until probably about 2030! So sentient robots, even though they're mentally restrained, are everywhere.

[Anticheese]

Quoting: macksting

The Obedience Check: Seems to me this would be a second Robotics roll, possibly against the Ego stat of the robot at character creation. In GH2, this could easily be emphasized as a possibility for the character's Nemesis, although I think I'd have issue with that if it turned out to be particularly more powerful than the PC. Still, the battle would be involved and epic if, randomly, it were to occur. Naturally, I'd expect most disobedient "rogue" robots to simply wander away on their own little tangent, with only a truly significant failure turning against its master openly.

But doesent a strong ego score represent sanity and reasoning as well as force of personality? I'd expect an E:10 bot to go "Yes master, I obey!" and an E:42 to start debating the meaning of life and thanking the creator.

Hmm, I guess this means that relegion is really easy for robots, Humans dont really get to see god in a physical sense, but a robot can just say "Wow, Thanks for bringing me to life! But whats with the tail?"

Anyhow, if there is to be a robotics check to see if the A.I just does not want to play ball, I would like to see it if there were only a remote chance of the robot going bezerk or A.W.O.L

--

Quoting: macksting

A strong AI can also be the backbone for several weak AI's, such as a single focal AI controlling or coordinating half a dozen small, expendable fighters. While the strength of this focal AI might vary, if you truly expect it to come home, and its obedience is "assured," there might be no reason to limit its capabilities by requiring human input. Machines tend to be faster, smaller, and less fragile in many ways than a human operator.

Isaac Asimov wrote a short story (you can find it in the book "I, Robot") about a "Multiple Robot", Basically it was one robot with a strong intellegence with 6 barely intellegent robots permanently under its control, like "fingers"

The robot worked well, but the sixth was too much of a drain on its brain.

Anyhow, that idea would make a fair bit of sense in the GH universe, Multibots employed in mining situations for instance.

-----

Quoting: macksting

This is why you should always program your AI's instead of growing them, either mechanically or biologically. And why every ship should have an AI psychiatrist on board.

Humans have been trying to program AI's for years, and the best we can do is chess computers, And thats only because its possible to calculate every single possible move on a chess board and pick the most optimum one.

Purely programmed A.I is almost impossible, Even in the years to come if we make something sentient it will proberbaly be a manufacturing defect, Which is why I reckon that "grown" A.I is the way to go.

Allready we have made advances in that area, One scientest got a sliver of dogbrain and managed to get it playing Quake 3, (THEROY ALERT) And after an A.I reaches a certain point then an MRI scanner of sorts would copy the brain patterns of the A.I ready to be places in a powerful computer.

I agree with that you should program your AI, With full knowledge of what makes it tick (not to mention convenient read/write access..Goodbye "evil.ai") having an onboard AI would be alot more easier on the mind.

As for an A.I shrink? Just read through I, Robot to get an idea of when it would be useful.

--------

Quoting: macksting

Amongst their other purposes are as bodyguards, where a weak AI might not be able to grasp hidden threats; as deep sea workers, where a certain level of autonomy is helpful and the range of remote control is dismal; as firefighters; as military weaponry

I agree, Robots are better suited to danger related tasks than people..far better suited.

Where a human burns to a crisp in the face of 3 burning logs, a robot would not be affected.
Where a human is fatally wounded after one well placed shot, a robot keeps on going.

As a side note, the U.S millitary is developing a battle robot named "S.W.O.R.D.S", Think a scaled down remote control tank with a cannon.

[Sabin Stargem]

Robots can continously do things without becoming bored or tired, or at least if they are programmed to be like that.  This is useful for security, since a robotic sentry can wander the base perpetually, keeping an video-camera trained on anything 'interesting'.  If an robot is destroyed, it probably would have an internal and delicate IFF identifier that would be set off, basically telling all the other robots that an intruder is present.  An floating or flying eye drone can make continous sweeps and note if anything is different from the usual, especially if it is something suspicious.  Of course, fuel might be a problem where active and moving machines are concerned, unless they use super-efficient solar panals or run on electricity provided by bot-docks.

[Anticheese]

I really like the concept of a fortress with its own suite of roving bots complete with sensitive IFF working on a shift (bot-docks)...

System Shock 2 had security cameras that would alert enemies to your presence (But it takes 4 seconds for the IFF to kick in), Deus Ex had security cameras too, They would activate an alert and occasionally a defensive turret, Then in Metal Gear Solid you have guards that are smart enough to send in a detatchment when the patrol does not check in.

Anyhow, Here is my idea on how to include security:

# ###
#e+R# - Okay, so you have an eye drone (e) and two guardbots hidden
# ### behind secret doors, The eye drone will continue to do a circut and
# +R# when the eye drone is destroyed, the other drones will redouble
# ### their efforts to find out what happened.

This makes for the stealth skill finally becoming useful, at the moment the best you can do is duck behind a tree and pray.

[Joseph Hewitt]

Quite a lot of good points in this thread. I think I'd like to differentiate between sensible science-fiction predictions about how robots/AIs should be, and thematic consideration of what kind of macguffin role they should play in this fictional universe.

Quoting: Anticheese

Hmm, I guess this means that relegion is really easy for robots

Or harder, since they can have a good look at their creator and size up all of his faults!

Quoting: Sabin Stargem

Robots can continously do things without becoming bored or tired, or at least if they are programmed to be like that.

Maybe this could be one of the drawbacks of sentient robots- they get bored.