Would CASIE implants be common in THS?

[wellspring]

Originally Posted by vicky_molokh

I mentioned a hi-res camera in the hope of racking up an Acute Vision bonus for the NAI. I don't particularly care whether the bonus comes from superior hardware (and I'm pretty sure by TL10 hardware will be better than natural human eyes by a lot) or good image recognition software (less likely, as resolution sets its limits anyway).

I think this is a case of acute vision not applying after a certain point. You're applying a general-purpose bonus in an area where higher-res imaging won't help (parallel case: greater colors not providing a heraldry bonus). YMMV

Originally Posted by vicky_molokh

Not really. That's at most 12*200 hours of study with a teacher, or twice that without one. IOW, even a NAI will negate the Low Empathy penalty for a single skill in 100-200 days (assuming it runs at the same speed as a human mind), and keep improving any of the listed skills at one level per roughly 70 subjective days of passive running. The important point is that keeps improving, and if run on a sufficiently fast computer while idle, it keeps improving fast. I'm not even sure it really needs a decent Per score to make much of a difference - 10 should suffice.

In my game I'd rule this out. Training requires that you actually do something. For example, you can't just sit motionless in a room and abstractly "train" up your DX (or any skill). You have to be in situations where you either study textbooks or practice it in real-life situations.

OK, so where is this situation for your AI? (Which does run at the speed of a normal human mind, unless the NAI/LAI templates have added some advantage since Changing Times). Let's assume a compartmentalized mind so it can train continuously in the background. Where is it training?

I can picture it reviewing instructional videos-- but there's a point of diminishing marginal returns on those. Ditto for training simulations. Both only provide a highly simplified representation of how people interact-- only now with another catch. Effectively, a simulation means you have two AIs that train continuously-- one lying and the other trying to guess whether the first one is lying or not. In that case, as a GM I'd rule that it's a case of hyperspecialization... the lie detector AI is tuned to catch a specific person's lies (their opponent's) to the exclusion of everyone else. In stats, we call this over-fitting your data. . Whatever the source, using archival training info will only provide a limited number of training hours before they're exhausted-- and that's pretty generous IMO. It's like learning martial arts purely from reading, without any physical practice of techniques at all.

To get skill at detecting lies in real-world scenarios, an AI has to actually train in the real world. And then your "training time" is limited not by the available waking hours pool but by the availability of training opportunities. Presumably an AI riding side-saddle on a police officer or a criminal, or a priest or a psychologist or a salesman will have lots of opportunities to watch the craft of social interaction in many contexts and will develop these skills. Xoxing (or creating daughter code that inherits this experience) might even make such an AI salable. A troublesome GM might also include some emergent behaviors that come along with this tricksy knowledge. Either way, it's not an assembly-line process.

Meanwhile, a human can develop these skills as well, and isn't starting at -3. They're skills you use every day no matter what job you end up in and don't effectively decay as the TL rises. Outsourcing these skills, even to AIs theoretically enslaved to you, opens you up to massive principal-agent problems.

As I said, it's possible to develop this kind of capability (along the lines I outlined in my post above). But there are limits to its practicality.

[wellspring]

Originally Posted by Xplo

Probably as readily as they would let them drive cars instead of developing greater physical fitness and look things up on the internet instead of becoming prodigies of rote memorization. Maybe even more so, seeing how important social success is.

Physical fitness isn't an essential skill in this age. Neither is rote memorization-- though my exams certainly aren't open-book. Calculators are allowed precisely because that skill is non-core to the subject area. OTOH, problem-solving skills are essential in our time (and in THS)... and sure enough, virtually all schools require that students use their own logic and work in solving problems for tests and assignments.

[vicky_molokh]

Originally Posted by wellspring

I think this is a case of acute vision not applying after a certain point. You're applying a general-purpose bonus in an area where higher-res imaging won't help (parallel case: greater colors not providing a heraldry bonus). YMMV

Well, Body Language explicitly benefits from Acute Vision. I suppose getting a +5 Camera Of Clarity is reasonable. Not sure how better than that TL10 cameras get, but the final modified Per score (including AV) is unlikely to exceed 20 (the Mortal Attribute level, or whatever it is called).

Originally Posted by wellspring

In my game I'd rule this out. Training requires that you actually do something. For example, you can't just sit motionless in a room and abstractly "train" up your DX (or any skill). You have to be in situations where you either study textbooks or practice it in real-life situations.

OK, so where is this situation for your AI? (Which does run at the speed of a normal human mind, unless the NAI/LAI templates have added some advantage since Changing Times). Let's assume a compartmentalized mind so it can train continuously in the background. Where is it training?

I can picture it reviewing instructional videos-- but there's a point of diminishing marginal returns on those. Ditto for training simulations. Both only provide a highly simplified representation of how people interact-- only now with another catch. Effectively, a simulation means you have two AIs that train continuously-- one lying and the other trying to guess whether the first one is lying or not. In that case, as a GM I'd rule that it's a case of hyperspecialization... the lie detector AI is tuned to catch a specific person's lies (their opponent's) to the exclusion of everyone else. In stats, we call this over-fitting your data. . Whatever the source, using archival training info will only provide a limited number of training hours before they're exhausted-- and that's pretty generous IMO. It's like learning martial arts purely from reading, without any physical practice of techniques at all.

To get skill at detecting lies in real-world scenarios, an AI has to actually train in the real world. And then your "training time" is limited not by the available waking hours pool but by the availability of training opportunities. Presumably an AI riding side-saddle on a police officer or a criminal, or a priest or a psychologist or a salesman will have lots of opportunities to watch the craft of social interaction in many contexts and will develop these skills. Xoxing (or creating daughter code that inherits this experience) might even make such an AI salable. A troublesome GM might also include some emergent behaviors that come along with this tricksy knowledge. Either way, it's not an assembly-line process.

I wasn't actually thinking of training simulations - I was thinking of something closer to book learning. Specifically, the kind of stuff Lightman & Co did in the series - watch thousands of videos of people in different emotional states, telling truth and lies etc., and looking for universal patterns. In the age of THS, I'm pretty sure gathering such videos isn't very hard, especially for a company that gains profits from selling NAIs trained on those videos. And the amount of material throughout the 150 or so years of video technology is definitely more than enough for a full-time AI.

Originally Posted by wellspring

Outsourcing these skills, even to AIs theoretically enslaved to you, opens you up to massive principal-agent problems.

That it does, but this happens with everything from TV to schooling (do you trust the news company? Do you believe the textbooks tell the truth, or merely that which the publishers want you to believe to be the truth?).

[vicky_molokh]

Originally Posted by wellspring

Physical fitness isn't an essential skill in this age. Neither is rote memorization-- though my exams certainly aren't open-book. Calculators are allowed precisely because that skill is non-core to the subject area. OTOH, problem-solving skills are essential in our time (and in THS)... and sure enough, virtually all schools require that students use their own logic and work in solving problems for tests and assignments.

Um, they're so essential that we developed computers, engines and the like to do these things better than us. In fact, ability to perform these actions is so essential that if we fall back onto the human level of ability (be it ability to solve complex math problems or push around carts weighting tonnes), TL will drop.

[Xplo]

Originally Posted by wellspring

Physical fitness isn't an essential skill in this age. Neither is rote memorization-- though my exams certainly aren't open-book.

And neither are social skills, if the computers handle that for you. And all you need to get there are computers who can do it better than the average person that the average person can also afford; everyone will shift over and it'll be the new status quo.

[Tyneras]

Originally Posted by Xplo

And neither are social skills, if the computers handle that for you. And all you need to get there are computers who can do it better than the average person that the average person can also afford; everyone will shift over and it'll be the new status quo.

If your computer is handling your social skills for you, why not cut out the middle man? It's not like replaced factory workers got to sit around and watch the machines do their jobs and still get paid.

"I must say you're an excellent conversationalist, and I love your opinions on the politics of the 2050s. Shame you have to drag that lump of meat around with you, what did you call it again?"

[Xplo]

Originally Posted by Tyneras

If your computer is handling your social skills for you, why not cut out the middle man? It's not like replaced factory workers got to sit around and watch the machines do their jobs and still get paid.

"I must say you're an excellent conversationalist, and I love your opinions on the politics of the 2050s. Shame you have to drag that lump of meat around with you, what did you call it again?"

Well, because the middle man is the one with the initiative, and the property rights, and such.

[Tyneras]

Originally Posted by Xplo

Well, because the middle man is the one with the initiative, and the property rights, and such.

I haven't played DX:HR, so maybe I'm getting the wrong impression, but it seems like the human is nothing more than a meat memory device and specialized sub-processor in this case. A bio-shell in all but name. We are social creatures, once you outsource the social part, do you even matter?

[vicky_molokh]

Originally Posted by Tyneras

I haven't played DX:HR, so maybe I'm getting the wrong impression, but it seems like the human is nothing more than a meat memory device and specialized sub-processor in this case. A bio-shell in all but name. We are social creatures, once you outsource the social part, do you even matter?

I certainly see aspects of Humanity's Retirement in the whole THS setup. Perhaps not as much as in Culture, but still.

That being said, we outsourced the Heavy Lifting and our wonderful overland endurance to animals and machines, and remained to manage them.
We outsourced Mathematics (Cyphering) to computer machines, and remained to manage them.
We outsourced The Great Art of Memory to books and teh Google, and remained to manage them.
We outsourced Calligraphy to computer machines too, and, now more commonly wapuro-bakas than before, remained to manage them.
We outsourced our ability to judge reputations to kontakts and facebooks.

Why wouldn't we outsource other things that require raw power in some area, and remain to manage the things that provide this raw power?

[Xplo]

Originally Posted by Tyneras

I haven't played DX:HR, so maybe I'm getting the wrong impression, but it seems like the human is nothing more than a meat memory device and specialized sub-processor in this case. A bio-shell in all but name. We are social creatures, once you outsource the social part, do you even matter?

The DX:HR setting doesn't seem to have much in the way of AI. The CASIE implant is, at best, a kind of expert system - sort of a social HUD, you might say (and of course this is exactly the form it takes in the game). It's not really possible for the CASIE to cut out the user any more than a GC/MS can cut out the chemist or a GPS can cut out the driver.

In THS, where AI of varying sophistication exists as part of everyday life and has varying amounts of freedom and civil rights, this could (and probably should) be a concern and cause for anti-AI activism. More grist to the mill!