Most Convincing AIs in Literature?

[Nancy Fulda]

Originally Posted by fjtorres

HP has recently done work pointing towards microchips that can reconfigure themselves on the fly using memristor crossbar latches. It's going to require a whole new programing model but the tech could provide a suitable hardware base to host a Strong AI... someday.

Oh, that is very cool.

[JasonB]

I know of at least two AIs in "classic" literature that have not been mentioned yet. Both are examples of "emergent" intelligence - where the AIs aren't just turned on from being built and think right away, but gradually become aware of themselves:

Sven from The Turing Option by Marvin Minsky and Harry Harrison.

(That one is perhaps my favourite example of this particular sub-genre.)

H.A.R.L.I.E from When H.A.R.L.I.E. Was One by David Gerrold.

BTW: This is one of my favourite subjects. I have a Philosophy degree, and questions of mind, perception, and identity were my main focus.

Later: Oops! theducks did mention When H.A.R.L.I.E. Was One.

[Steven Lyle Jordan]

Originally Posted by DavidKitson

And of course, what process is necessary to produce such an AI that can be trusted with the sort of tasks that we should only entrust to another human.

Considering the importance of tasks we currently entrust to those we know we can't trust (Congress comes easily and quickly to mind), I'd say the answer to your question is "Not much of one."

[JasonB]

The trilogy-to-be (I assume) that includes The Unincorporated Man and The Unincorporated War, by Dani and Eytan Kollin, has a series of narratives that includes one by an AI among a race of AIs. (Only briefly mentioned in the first book, the second has much more focus on this.) While there isn't much philosophical discussion that goes on in the series about "convincing" (or real) AI, they are treated as another form of intelligence that can relate to humans at their level - and there's a bit of talk about their different perceptions and interactions. Still, the authors at least imply that whatever form an intelligence takes, that any culture of one intelligence will end up having the same kind of problems as a culture of any other kind of intelligence. Specifics aside, they seem to say that form doesn't matter.

[JasonB]

Okay, I just have to mention two more...

Robert J. Sawyer's The Terminal Experiment, about a scientist who creates 3 computer models of himself, each missing a key psychological awareness, in an effort to find out what death and possible afterlife is like. One of the 3 AIs ends up committing murder.

The series of murder mysteries, by Donna Andrews, where the sleuth, Turing Hopper (LOL), is actually a computer program. The first in the series was You've Got Murder. I have to admit that I'd never heard of this until I started Googling around trying to get the names of the books I'd been thinking of for this thread. It looks interesting, in a tongue-in-cheek way, and I'm probably going to have to read it now!

[Nancy Fulda]

Whew! Okay, after a very late night writing session, the article is complete and off to the editor.

I wish I could have fit in all the fabulous AIs mentioned here. I did manage to include Mycroft, Data, Terminator, Asimov's Bicentennial Man, Robert J. Sawyer's www:wake books, Tik-Tok from the land of Oz, and a few others. Not bad for 2000 words, but really, the topic is deserving of a far more extensive treatment.

Ah well. Next decade, perhaps.

[Nancy Fulda]

A couple more references that I found intriguing:

Eric James Stone's Unforgettable -- features an evil AI based on Quantum Computing.

Murray Leinster's A Logic Named Joe. This one didn't fit anywhere in the article because there's no real description of how he works, but good heavens. This is a pre-1950's story. Computing is in its infancy, and Mr. Leinster has predicted the internet with disturbing accuracy. Wow.

[fjtorres]

Originally Posted by Nancy Fulda

Murray Leinster's A Logic Named Joe. This one didn't fit anywhere in the article because there's no real description of how he works, but good heavens. This is a pre-1950's story. Computing is in its infancy, and Mr. Leinster has predicted the internet with disturbing accuracy. Wow.

Indeed.
Leinster is one of the great ones (FIRST CONTACT alone qualifies him) and quite capable of coming up with the idea on his own.
But the timing of the story--1946--suggests he might have been inspired by Vannevar Bush's seminal 1945 paper describing the "MEMEX". From there to the network of "logics" was still a hefty leap and showed true insight, though.

Another example of knowledgable writers extrapolating with scary accuracy (not AI, though) are the pocket computers in Niven & Pournelle's MOTE IN GOD'S EYE, from 1974. They pretty much invented connected PDAs a generation early. The novel is easily one of the three best SF novels of the 20th century, btw. Ridiculously well recommended.

[EatingPie]

Originally Posted by Nancy Fulda

Whew! Okay, after a very late night writing session, the article is complete and off to the editor.

I wish I could have fit in all the fabulous AIs mentioned here. I did manage to include Mycroft, Data, Terminator, Asimov's Bicentennial Man, Robert J. Sawyer's www:wake books, Tik-Tok from the land of Oz, and a few others. Not bad for 2000 words, but really, the topic is deserving of a far more extensive treatment.

Ah well. Next decade, perhaps.

Are you going to have mercy on us and tell us what the article is titled and who it's for?!?!

-Pie

[Rob87]

As to produce real AIes I think the only way is to set up a self-learning neuronal network and let it learn and develop. (which is AFAIK the way chosen to achieve the goal)
Making one finished out-of-the-box seems impossible to me because it would require full understanding of what intelligence and personality fully and exactly is.
I.e. full understanding of the brain - with the brain, which is paradox

Artificial neural networks have been around a long time now and it seems not much progress has been made in creating human like intelligence.

Probably because they require too much hand holding - from my limited understanding it sounds like you need to tell them what to learn, how to assess their performance at it and through lots feedback they eventually improve at what they've been coded to do.

If thats the best we can do Human like A.I is miles off

[EatingPie]

Originally Posted by Rob87

Artificial neural networks have been around a long time now and it seems not much progress has been made in creating human like intelligence.

Hey, do your research buster! Mr. Data is made up of Neural Networks and Heuristic Algorithms, and he seems positively human at times!



-Pie

[Nancy Fulda]

Originally Posted by EatingPie

Are you going to have mercy on us and tell us what the article is titled and who it's for?!?!

-Pie

I don't believe in jynxing deals by revealing information too early. (I'm pretty sure the magazine will take it, since they commissioned it and specifically requested the literature tie-ins, but the head editor hasn't approved it yet.)

However, I hereby promise to tell all once the paperwork's been signed.

[Nancy Fulda]

Originally Posted by Rob87

Artificial neural networks have been around a long time now and it seems not much progress has been made in creating human like intelligence.

Probably because they require too much hand holding - from my limited understanding it sounds like you need to tell them what to learn, how to assess their performance at it and through lots feedback they eventually improve at what they've been coded to do.

If thats the best we can do Human like A.I is miles off

Back when I was writing my Master's thesis, one of the biggest problems with neural nets (and most other learning algorithms) was setting up the reward structure right. Computerized systems are very, VERY good at learning what you tell them to do.

If you give a mobile robot negative feedback whenever it bumps into a wall, guess what it will learn to do? Yup. Sit there like a rock. So the researcher has to add in a positive reinforcement for moving around, at which point the robot will learn to spin in endless circles -- it's moving, but it's not hitting any walls, see? So then you start layering on things like rewards for being 'curious' or exploring new territory, rewards for reaching certain goal locations, and so forth. The whole process is dreadfully complicated, but in a very real sense, it's not the neural network's fault. We just don't know the magic recipe for laying out the right reinforcements.

Well, okay, then there are also generalization problems, but that's a different kettle of fish.

[Freeshadow]

whenever I read something about such self learning systems, there were always minor yet striking behavioral patterns involved the nets developed which were unexplainable to no one which is the basis for what Isaid about explaining intelligence with brain.

[Freeshadow]

Nancy have you ever heard of this game? http://en.wikipedia.org/wiki/Creatur...ife_program%29