Superhuman IQ [Space/Spaceships/Ultratech]

[jason taylor]

Originally Posted by AlexanderHowl

AI gods are part and parcel of a lot of science fiction stettings (especially transhumanist science fiction settings), so I thought that it would be to assess how they would be used in campaigns.

You may still be overdoing the problem. A superintelligent being might have MORE capacity for interacting with humans, not less especially if empathy is built into said capacity.

A dog cannot comprehend building of space shuttles (well neither can I, but I can understand the purpose and some of the processes). It can understand the things at which a human and a dog are more or less equiv especially as it is in human's nature to condescend for the benefit of said dog. A dog can for instance understand, "Big chief human, mighty hunter, bring much meat" even if he cannot comprehend a rifle or shotgun. Likewise it can comprehend, "Guard territory, protect human cubs" even if it cannot comprehend what the Common Law says about burglary.

What a superbeing would have is some part of itself that is outside the comprehension of humans and it would likely be giving the impression of this to the PCs from time to time. The part where it does interact with humans, should be parts that are reasonably close to a human perspective. Just as there are parts of humanness that are enough like caninity for a dog and a human to interact.

[AlexanderHowl]

Originally Posted by David Johnston2

Then what you need to do is look at the actual mechanics. The truth is, the gap between an IQ 42 super intelligence and a human isn't really bigger than the gap between a human and an ant. Numerically speaking that's a ten fold increase. Moreover there's a qualitative difference in that ants just can't learn any skills or languages. At all.

So looking at the actual mechanics the super intelligence presumably has Psychology and social skills around 38, but also a substantial penalty because it's a nonhuman trying to understand and manipulate humans. It doesn't have the initial advantage of actually sharing the psychology it is interpreting. And the people it is dealing with aren't going to be IQ 10 either.

An ant is less than a tenth as intelligent as a human being. We should probably look at differences in IQ as being log function rather than a linear function. An IQ 10 human is eighty times as smart as an IQ 1 ant while an IQ 42 AI god is over 10,000 times as smart as an IQ 10 human.

As for the penalties, what penalties actually matter at skill 42? Even if it suffers a -10 to defaulting between its Psychology (AI) and Psychology (Human), it is still rolling against a 32 (against a 29 when Cultural Familiarity is considered). With a 29, it probably understands human psychology better than any human.

[David Johnston2]

Originally Posted by AlexanderHowl

An ant is less than a tenth as intelligent as a human being. We should probably look at differences in IQ as being log function rather than a linear function. An IQ 10 human is eighty times as smart as an IQ 1 ant while an IQ 42 AI god is over 10,000 times as smart as an IQ 10 human.

I doubt we should. There is no equivalent of basic lift for IQ. Certainly the difference in reflexes between DXs don't work like that. It is however true that a Psychology skill of 38+ is huge. So if a superintelligence (who has Psychology) devotes all of its attention to predicting what a single person will do we can expect that projection to be basically flawless as long as there are no unknowable variables are in play. But gee that would have to be a pretty important person.

[jason taylor]

Originally Posted by David Johnston2

Then what you need to do is look at the actual mechanics. The truth is, the gap between an IQ 42 super intelligence and a human isn't really bigger than the gap between a human and an ant. Numerically speaking that's a ten fold increase. Moreover there's a qualitative difference in that ants just can't learn any skills or languages. At all.

So looking at the actual mechanics the super intelligence presumably has Psychology and social skills around 38, but also a substantial penalty because it's a nonhuman trying to understand and manipulate humans. It doesn't have the initial advantage of actually sharing the psychology it is interpreting. And the people it is dealing with aren't going to be IQ 10 either.

Manipulate implies distasteful motives. It may regard humans as pets rather than chess pieces.

Benevolent Sufficiently Advanced Beings in the trek franchise aren't as common as less so. And when they come they seem to get into the Can't Argue With Elves sort (the Organians for instance are just doing gunboat diplomacy on another level to what the Federation does regularly and making airs about it is absurd). However the Squire of Gothos' parents would qualify as benevolent, and maybe the Prophets on DS9.

None of those are of course AIs. The one example of an AI superbeing is the Borg which is/are far from benevolent.

[AlexanderHowl]

AI gods do not have to be benevolent or malevolent, they can be scary enough if they are indifferent. An IQ 42 entity is 32 'levels' above humans in intelligence while humans are only 9 'levels' above ants, so they might not acknowledge humans anymore than humans acknowledge ants. They might find us an interesting field of study for a few years, and they might prevent our extinction due to a desire to maintain uniqueness, but they will likely not care if there are 10 billion humans or 100 million humans.

[DataPacRat]

Originally Posted by AlexanderHowl

AI gods do not have to be benevolent or malevolent, they can be scary enough if they are indifferent. An IQ 42 entity is 32 'levels' above humans in intelligence while humans are only 9 'levels' above ants, so they might not acknowledge humans anymore than humans acknowledge ants. They might find us an interesting field of study for a few years, and they might prevent our extinction due to a desire to maintain uniqueness, but they will likely not care if there are 10 billion humans or 100 million humans.

Whatever motivation AIs with such levels of IQ and skills have, I expect they're more likely to be Blue-and-Orange Morality than anything we'd recognize as reasonable.

[David Johnston2]

What would be hilarious in such a setting is a person with zeroed. Imagine how much they'd screw up master plans just by being a person about whom the super intelligences are constantly drawing a blank.

[smurf]

Originally Posted by AlexanderHowl

AI gods do not have to be benevolent or malevolent, they can be scary enough if they are indifferent. An IQ 42 entity is 32 'levels' above humans in intelligence while humans are only 9 'levels' above ants, so they might not acknowledge humans anymore than humans acknowledge ants. They might find us an interesting field of study for a few years, and they might prevent our extinction due to a desire to maintain uniqueness, but they will likely not care if there are 10 billion humans or 100 million humans.

GURPS IQ is not just intelligence it is all various bits of knowledge. The IQ tests were predicated on a racist default.

Unlike ST which can be determined by how strong beings are or DX has some idea have how good X is versus Y, IQ has a wider varied context than just computational power.

The human mind is capable of many computations. It has to deal with various social interactions. This cannot simply replicated by high IQ.

[DataPacRat]

One way that might be worth using to model better-than-human intelligences is less IQ, and more Advantages; specifically, something along the lines of a "Path to Victory" set ala:

* Common Sense (10), Conscious (+50%, pP45) [15]
* Intuition (15), Inspired (+100%, pP56), Cosmic (No Die Roll Required, +100%, pP101), Reliable +10 (+50%) [52]
* Oracle (Digital) (15, pP65), Inspired (+100%), Cosmic (No Die Roll Required, +100%, pP101), Reduced Time (from 1 hour to instantaneous, +260%) Reliable +10 (+50%) [91]
* Precognition (pB77, pP68) (25), Directed (+100%), Cosmic (No Die Roll Required, +100%, pP101), Reduced Fatigue Cost 2 (from 2 to 0, +40%), Reliable +10 (+50%), Reduced Time (10 minutes to instantaneous, +220%) [152]
* Psychometry (20), Directed (+50%, pP69), Immersive (+100%), Sensitive (+30%), Hypersensory (-50%), Cosmic (No Die Roll Required, +100%, pP101), Reliable +10 (+50%) [76]

If we use THS's approach that an AI who's spent CP on things other than skills costs an extra $30k per such CP, then the above is a mere $11.58M, likely a fraction of the price of the hardware.