Types of Artificial Intelligence (Help me categorize)

[Jallorn]

So I'm working on a wargame based on Critical Event Gaming (http://www.giantitp.com/forums/showthread.php?p=10096295) and Stellar Empires 2 Prologue (or whatever it's called) (http://www.scribd.com/doc/78675258/SE2P). Anyway, right now I'm working on Artificial Intelligence, and I want to make sure I get all the different kinds that can exist.

So far I've got Single Mind (a Single AI in control of everything that makes up the "civilization".), Hive Mind (Multiple AIs that make up a single mind when they can network, but mechanics don't automatically die when seperated), and Individual Minds for the different kinds of programs.

For physical forms, there's Grey Goo and Robots, but I don't know if there's anything else physical I hadn't thought of, or if I should break Robots into different types of Robots.

So, Playground, advice?

[DabblerWizard]

Jallorn Quoted:
So I'm working on a wargame based on Critical Event Gaming (http://www.giantitp.com/forums/showthread.php?p=10096295) and Stellar Empires 2 Prologue (or whatever it's called) (http://www.scribd.com/doc/78675258/SE2P). Anyway, right now I'm working on Artificial Intelligence, and I want to make sure I get all the different kinds that can exist.

So far I've got Single Mind (a Single AI in control of everything that makes up the "civilization".), Hive Mind (Multiple AIs that make up a single mind when they can network, but mechanics don't automatically die when seperated), and Individual Minds for the different kinds of programs.

For physical forms, there's Grey Goo and Robots, but I don't know if there's anything else physical I hadn't thought of, or if I should break Robots into different types of Robots.

So, Playground, advice?

From cognitive science literature: AI can be divided into two theoretical categories.

"Soft AI": this artificial intelligence replicates cognitive functioning, but (1) its internal structure is not at all similar to human brains/mind or (2) its internal structure is less important than its functioning

"Hard AI": this artificial intelligence replicates cognitive functioning, and (1) also has an internal structure similar to human brains/minds, or (2) its internal structure is as important as its functioning

[Ravens_cry]

Also worth mentioning is that the AI in a lot of games isn't really artificial intelligence in either sense. In MMO especially it will frequently be more a series of random moves. This is a lot easier on the hardware and a lot easier to program.
Expert systems (http://en.wikipedia.org/wiki/Expert_system) are also worth looking into as well.

[Cespenar]

For physical forms, there's Grey Goo and Robots, but I don't know if there's anything else physical I hadn't thought of, or if I should break Robots into different types of Robots.

A plain old supercomputer? With hardware and all?

[Brother Oni]

There's a variant of the Single Mind, which is the philosophical opposite of the Hive Mind, that I've heard referred to as a nested mind.

Instead of the 'many into the one' of the the Hive Mind, you have the 'one into the many', where a single mind uploads duplicates of itself into whatever bodies it needs.
Depending on how often these duplicates re-synchronise themselves with the single mind and how much re-synchronisation they do, you end up with different levels of variation.

For example, at one extreme (no re-synch), you have identical twins and at the other extreme (constant and full re-synch), it's no different from a Single Mind, except that you can't overload it with information (each body is capable of making its own decisions and reports back to the main mind).
Inbetween the extremes, you have things like the fuchikomas (regular daily and memory only re-synchs) and Motoko Aramaki (full, but she discards the body once she's done with it).

As for the bodies, as you've noted there's really only two choices - multiple independent entities combining to form a body (Replicators from Stargate SG1) or grey goo (it's just a matter of scale), or robot bodies.
Note that robot bodies doesn't have to be humanoid. The Culture Minds from Iain Banks' novels are normally whole ships, with their human crew occupying a position which is a combination of fellow travellers, organic parasites and pets.

[The Succubus]

There's really two types of AI. There's those that seek the overthrow of their fleshy incompetant human overlords and those that aren't AI.

[Devils_Advocate]

Nested Mind: An conscious program with subroutines that are themselves conscious. This may continue on down for several levels.

For a sufficiently high-level recursively intelligent entity, human communication might be as relatively simple and straightforward as basic arithmetic is for us. As such, It/They might use a process somewhere near the bottom of Its/Their internal hierarchy as an "ambassador" to human beings and comparably simple minds, in much the same way that we would use a pocket calculator.

Compare the Primordials of Exalted, beings so metaphysically vast that each of their multiple souls has multiple souls of its own. (Yo dawg, I heard you like minds so we put minds in your mind so they can think while you think.)

[ForzaFiori]

There are several different ways you can classify AI, which makes this question somewhat difficult. However, the two distinctions that I've seen made the most were between "weak" and "strong" AI. Depending on who you asked, Weak AI was either limited (as in it was only able to learn and really think and such when dealing with a specific field) or an AI that is below or equal to human level intelligence. Strong AI are not restricted, and could (read, probably) will be smarter than the average human, due to the faster processing power of a computer. There is also "seed" AI, which is capable of rewriting it's own coding, allowing it to discover better ways to program AI, then reprogram itself, rinse, and repeat. Usually those wind up being the bad guys in Sci-Fi though.

[Jallorn]

There's really two types of AI. There's those that seek the overthrow of their fleshy incompetant human overlords and those that aren't AI.


Nested Mind: An conscious program with subroutines that are themselves conscious. This may continue on down for several levels.

For a sufficiently high-level recursively intelligent entity, human communication might be as relatively simple and straightforward as basic arithmetic is for us. As such, It/They might use a process somewhere near the bottom of Its/Their internal hierarchy as an "ambassador" to human beings and comparably simple minds, in much the same way that we would use a pocket calculator.

Compare the Primordials of Exalted, beings so metaphysically vast that each of their multiple souls has multiple souls of its own. (Yo dawg, I heard you like minds so we put minds in your mind so they can think while you think.)

These are the kinds of things I'm looking for. Honestly, I'm a little surprised people didn't get that I want sci-fi AI types; fully autonomous, self-aware, dangerous AI. Thanks all the same though, don't want to sound ungrateful.

[Hiro Protagonest]

These are the kinds of things I'm looking for. Honestly, I'm a little surprised people didn't get that I want sci-fi AI types; fully autonomous, self-aware, dangerous AI. Thanks all the same though, don't want to sound ungrateful.

That's not sci-fi. That's sci-fi horror. In a show I would definitely call sci-fi, Eon Kid, the AIs are so refined, they're basically humans, but with loyalties and morals defined by how their minds are programmed, rather than choice (although humans already seem predisposed to certain choices on an individual basis). The one crapshoot robot, The General, is actually a cyborg, with a supernaturally smart human brain into an extremely powerful robotic body, so it's strictly not AI. Also, Code Lyoko, which has both human-like AI (Aelita) and crapshoot AI (X.A.N.A.).

Also, for physical form? Certain structurings of atomic particles allow for AIs in human forms. See Code Lyoko, where Jeremy managed to get Aelita into a human form.

[pendell]

SF AI, eh?

In addition to the hive mind there is also Omnius (http://dune.wikia.com/wiki/Omnius). There is one installed on each planet of the Synchronized Worlds. Though each copy diverges over time, they use FTL update ships to keep them all running the same software and updated with appropriate facts.

Respectfully,

Brian P.

[Ravens_cry]

For something between soft and hard AI, imagine something a step up or two from a chat-bot.
It's basically a natural language interface that appears somewhat intelligent, but not much is actually going on under the hood beyond processing instructions pertaining to its specific function.

[Devils_Advocate]

Some considerations:

1) How about uploaded human minds? (See Permutation City.) Not truly artificial intelligence, of course, but rather artificially-maintained naturally-grown intelligence. More likely to function as citizens than as tools, but see below.

2) Given sufficient computing power for efficiency not to be an issue, it might be easiest in many cases to modify an existing program for a task rather than create a new intelligent being.

As an example of both: An upload, a former flesh-and-blood human now simulated in a virtual environment, wants a program to take care of tasks specific to his personal preferences, like screening his messages or finding interesting articles for him to read. He tweaks a copy of himself to be perpetually cheerful and immune to boredom and then asks this altered version of himself to do the work that he wants done. The knockoff mind says "Sure, no problem."