[Sci-Fi] Alternatives to Influence Volitional AIs

weevis · 10-05-2015, 07:38 AM

What are some optional interesting ways that influence rolls could work against a volitional sentient AI?

UT tells me "robots usually resemble their masters in thought processes" (p. 34) -- we are to compare the AI to an "equivalent" human (UT p. 28). However, there's a lot of fascinating science fiction where this "usually" statement is unsatisfying. Many novels contain the assumption that AI would evolve in a way that did not model human intelligence, and that it would be alien and incomprehensible. A raft of dramatic sci-fi moments involve reasoning with a robot. These points of tension work because they are about the *differences* between human and AI thought processes (or they're about trying to figure out the differences)... In "Moon" does Gerty care?

Some GURPS rules seem too conservative on AI and robots for this kind of drama to work. For example: p. B181 suggests that I might be able to read the Body Language of an "utterly alien species" at -6, but for anything with the Machine meta-trait "no roll is possible." (My biomorphic lens volitional AI frowns.) On p. B184 to use any "social skills" on an AI I'm advised to require BOTH the skill for Humans, plus Computer Programming (AI). Then roll vs. the lower skill. That's OK but it flattens all of the AI stuff into one skill.

So how about some alternative approaches. How else could I model influencing AIs? I could adopt the Alien rules even though UT tells me not to. e.g., "Just because a creature is intelligent doesn't mean it will have a mind like ours." (GURPS Space, p. 166). Here are some other ideas that are not well thought through:

Diplomacy or Fast-Talk: I feel like there is a more fun way to Fast Talk but I don't know what it is. If the AI lacked human emotion, what about combining one of these skills with some sort of logic. Philosophy (Formal Logic)? In Asimov the robots were always so logical.

Intimidation: Intimidation is about conveying the will to harm, but perhaps an AI with a non-human form/internals would require an additional prerequisite like Mechanic (Robotics), Electronics Repair (Computers), or Bioengineering to know what would harm. This seems like it could work on an AI.

Savoir-Faire (AI): This seems like a fun idea if AIs hang around with each other a lot. Of course, they might do so in a way that humans cannot share -- the weird transcendence suggested by "Her" or the depths of the Matrix in "Neuromancer." But what if it is possible that a human might somehow pick up a few tricks from AI "society"? You could even have an con man specializing in AIs.

Sex Appeal: I guess this *might* work normally if AI is properly programmed and equipped -- thanks "Ex Machina" for making me think about this more than I really wanted to -- or it wouldn't work at all.

Psychology: If AIs are very different, instead just use Computer Programming (AI)? i.e. not the "lower of two skills" approach? Or maybe there is a skill like Psychology (AI)?

Streetwise: Streetwise is about how the underworld works, and this might operate just the same way for an AI as long as they are operating in the same underworld.

Any creative ideas?

D10 · 10-05-2015, 07:45 AM

Its a rather long subject, but inteligence, be it human or AI, has some computational patterns that are ultimately universal, what makes us special is the fact that theres many variations on sensor intensity and in perception creation that allows 2 people with the same experience to value it totally diferently.

The crux of the issue is managing to program the AI as a social being, instead of merely granting it awareness amidst a population of social beings.

whswhs · 10-05-2015, 07:46 AM

Quote:

Originally Posted by weevis

Sex Appeal: I guess this *might* work normally if AI is properly programmed and equipped -- thanks "Ex Machina" for making me think about this more than I really wanted to -- or it wouldn't work at all.

I think <i>Ex Machina</i> was about the AI using it on the human, not conversely.

weevis · 10-05-2015, 01:42 PM

Quote:

Originally Posted by whswhs

I think <i>Ex Machina</i> was about the AI using it on the human, not conversely.

Heh, fair enough.

weby · 10-05-2015, 02:45 PM

AI Empathy[5]?

weevis · 10-05-2015, 03:07 PM

Quote:

Originally Posted by weby

AI Empathy[5]?

I like it. The Perks book rule for "Good With (X)" suggests something like "Good With (volitional sentient AI)" would be a 1 point version of Empathy (the lowest level) if the AIs are rare.

starslayer · 10-06-2015, 06:08 AM

In a world where multiple races exist with completely alien thought processes from one another it might make sense to have certain skills specialize by race.

Fast talk (human) focused on abusing the web of trust that humans as social animals required to exist in a society.

Fast talk (dragon) focuses on casual misrepresentations of your personal power out reserves since dragons are solitary top predators who just need to calculate the risk reward for listening to someone.

Fast talk (ai) focuses on casually inserting difficult logical conundrums into your speech so that the ai's processor literally becomes bogged down and less able to dedicate full resources to your final request

10-05-2015, 07:38 AM	#1
weevis Join Date: Jul 2015 Location: 3.165, -3.048, -0.0818	[Sci-Fi] Alternatives to Influence Volitional AIs What are some optional interesting ways that influence rolls could work against a volitional sentient AI? UT tells me "robots usually resemble their masters in thought processes" (p. 34) -- we are to compare the AI to an "equivalent" human (UT p. 28). However, there's a lot of fascinating science fiction where this "usually" statement is unsatisfying. Many novels contain the assumption that AI would evolve in a way that did not model human intelligence, and that it would be alien and incomprehensible. A raft of dramatic sci-fi moments involve reasoning with a robot. These points of tension work because they are about the differences between human and AI thought processes (or they're about trying to figure out the differences)... In "Moon" does Gerty care? Some GURPS rules seem too conservative on AI and robots for this kind of drama to work. For example: p. B181 suggests that I might be able to read the Body Language of an "utterly alien species" at -6, but for anything with the Machine meta-trait "no roll is possible." (My biomorphic lens volitional AI frowns.) On p. B184 to use any "social skills" on an AI I'm advised to require BOTH the skill for Humans, plus Computer Programming (AI). Then roll vs. the lower skill. That's OK but it flattens all of the AI stuff into one skill. So how about some alternative approaches. How else could I model influencing AIs? I could adopt the Alien rules even though UT tells me not to. e.g., "Just because a creature is intelligent doesn't mean it will have a mind like ours." (GURPS Space, p. 166). Here are some other ideas that are not well thought through: Diplomacy or Fast-Talk: I feel like there is a more fun way to Fast Talk but I don't know what it is. If the AI lacked human emotion, what about combining one of these skills with some sort of logic. Philosophy (Formal Logic)? In Asimov the robots were always so logical. Intimidation: Intimidation is about conveying the will to harm, but perhaps an AI with a non-human form/internals would require an additional prerequisite like Mechanic (Robotics), Electronics Repair (Computers), or Bioengineering to know what would harm. This seems like it could work on an AI. Savoir-Faire (AI): This seems like a fun idea if AIs hang around with each other a lot. Of course, they might do so in a way that humans cannot share -- the weird transcendence suggested by "Her" or the depths of the Matrix in "Neuromancer." But what if it is possible that a human might somehow pick up a few tricks from AI "society"? You could even have an con man specializing in AIs. Sex Appeal: I guess this might work normally if AI is properly programmed and equipped -- thanks "Ex Machina" for making me think about this more than I really wanted to -- or it wouldn't work at all. Psychology: If AIs are very different, instead just use Computer Programming (AI)? i.e. not the "lower of two skills" approach? Or maybe there is a skill like Psychology (AI)? Streetwise: Streetwise is about how the underworld works, and this might operate just the same way for an AI as long as they are operating in the same underworld. Any creative ideas? Last edited by weevis; 10-05-2015 at 03:09 PM. Reason: typo

10-05-2015, 07:45 AM	#2
D10 Join Date: May 2009 Location: In Rio de Janeiro, where it was cyberpunk before it was cool.	Re: [Sci-Fi] Alternatives to Influence Volitional AIs Its a rather long subject, but inteligence, be it human or AI, has some computational patterns that are ultimately universal, what makes us special is the fact that theres many variations on sensor intensity and in perception creation that allows 2 people with the same experience to value it totally diferently. The crux of the issue is managing to program the AI as a social being, instead of merely granting it awareness amidst a population of social beings. Last edited by D10; 10-05-2015 at 08:00 AM.

10-05-2015, 02:45 PM	#5
weby Join Date: Oct 2008	Re: [Sci-Fi] Alternatives to Influence Volitional AIs AI Empathy[5]? __________________ -- GURPS spaceship unofficial errata and thoughts: https://gsuc.roto.nu/

10-06-2015, 06:08 AM	#7
starslayer Join Date: Dec 2006	Re: [Sci-Fi] Alternatives to Influence Volitional AIs In a world where multiple races exist with completely alien thought processes from one another it might make sense to have certain skills specialize by race. Fast talk (human) focused on abusing the web of trust that humans as social animals required to exist in a society. Fast talk (dragon) focuses on casual misrepresentations of your personal power out reserves since dragons are solitary top predators who just need to calculate the risk reward for listening to someone. Fast talk (ai) focuses on casually inserting difficult logical conundrums into your speech so that the ai's processor literally becomes bogged down and less able to dedicate full resources to your final request