Most Convincing AIs in Literature?

[Freeshadow]

@6charlong
I beg to differ
what you speak of are only scripts not an AI

[fjtorres]

Originally Posted by Freeshadow

@6charlong
I beg to differ
what you speak of are only scripts not an AI

Most games do rely on scripted/tripwired responses to simulate AI, but some developers have tried to code more general AI-like behavior in their NPCs. With no meaningful success.

Bethesda Softworks has been trying for years and they implemented some of it in early playable builds of OBLIVION but had to dial it back because emergent behavior disrupted game play. Even the watered down version that shipped could get out of hand on occasion; one documented instance, not triggered by the player, began with a city guard chasing an underclass beggar/thief and a citizen/bystander getting injured in the crossfire and retaliating against the guard. Like an old western, the incident escalates and results in a city-wide riot and eventually everybody in town is dead. All over a loaf of bread.
Most times, however, what happens is the guard chases and kills the thief without triggering any bystander's self-defense/retaliation subroutine or catches her in the act and the player only sees the dead thief.

It is important to distinguish between the simulation of sentience (which is the goal of most AI projects) and true cybernetic sentience. The former only needs to be good enough to fool most humans, the latter has to be capable of truly independent motivation and self-awareness. The former is something extremely hard but probably achievable with known principles, the latter is probably not and will, alas, likely remain a literary and philosophical construct for the forseable future.

Interestingly enough, Bioware's MASS EFFECT games make the distinction clear and a major backstory element. Most of the computer personalities in use in the milieu are database driven personality simulations, called Virtual Inteligences, and identify themselves as such, VI's. True AIs exist within the milieu (the Geth, and others) but research and (worse) deployment of neural network systems is forbidden because of the way the low-level non-sentient Geth robots bootstrapped themselves into a collective sentience and ended up kicking their creators off their planets and into exile as "galactic gypsies" searching for a habitable world to call their own. Of course, just forbidding AI research doesn't mean it stops...

(Btw, the first game is highly recomended, even for non gamers; it is a very accessible entry point into a very detailed and well-realized universe.)

[snickp]

A book I have enjoyed recently is the www wake series

http://www.amazon.com/WWW-Trilogy-Ro.../dp/0441016790

Which speaks about the development of an intelligence in the background of the internet, rather than something deliberately developed through human intervention.

What I found particularly interesting was the mechanism through which the human characters come to realised the increasing intelligence that it out there, coupled with the resources that the "intelligence" taps into to find out more about the world itself.

Might be of interest.

[Nancy Fulda]

Originally Posted by fjtorres

It is important to distinguish between the simulation of sentience (which is the goal of most AI projects) and true cybernetic sentience.

Agreed.

In the research literature, the former is further subdivided into "strong" and "weak" AI. Strong AI refers to a machine capable of performing any intellectual task a human can perform, and has not yet been achieved. Weak AI refers to machines that make rational decisions given their circumstances, and exists in every chess program, search engine, and automated telephone system on the market.

[Freeshadow]

we have kind of derailed your thread didn't we?

[Nancy Fulda]

Originally Posted by Freeshadow

we have kind of derailed your thread didn't we?

Hey, I'm not going to complain. It's a fascinating discussion.

[Steven Lyle Jordan]

Originally Posted by Solitaire1

Since "Chobits" has been mentioned, I'll mention a few other anime that features at least one AI or possible AI: A.D. Police, All-Purpose Cultural Cat-Girl Nuku-Nuku, Armitage III, Astro Boy, Big O, Bubblegum Crisis, Dragonball Z, Excel Saga, Ghost In The Shell, Macross Plus, Mahoromatic, Saber Marionette J, and Steel Angel Karumi.

Huh! Silly bag of mostly water! How can you mention Astro Boy and dare to leave out 8th Man?

And as long as comic characters are on the board, let's mention the house/armor AI "Jarvis" from the Iron Man movies.

It occasionally surprises me that I've mostly written stories with "weak" AI in them, though my Verdant series includes a "weak" AI that may or may not be approaching sentience, thanks to the events of Verdant Skies (I'm purposely keeping it vague, though I'll clear up the sub-story if I ever write a third book).

To me, the AI sentience matter--can it be considered "alive", does it have a "soul"?--is a 20th century trope that I believe we're already moving past with Iron Man's Jarvis and other similar creations. Its parallel trope, the humanoid robot, is a late-19th/20th century trope, which is also left behind by the 21st century. The matter of soul will forever be an intellectual exercise only, as long as we cannot prove the soul exists. The only real issue is, Can humans deal with AIs as they would with other people? And if so, does that qualify AIs as Persons?

[fjtorres]

Originally Posted by Steven Lyle Jordan

The only real issue is, Can humans deal with AIs as they would with other people? And if so, does that qualify AIs as Persons?

One would hope humans *don't* treat AIs (or aliens) as they treat other humans.

Dealing with AIs as if they were people is too low a bar; dealing with them as equals will be the real breakthrough.

I'm thinking of Melinda Snodgrass' MEASURE OF A MAN; how even humans with a couple centuries' exposure to "others" needed a Shakespearean-grade speech to accept Commander Data as a person. And that in a "hopeful" fictional universe. In real life...?

[DavidKitson]

Originally Posted by fjtorres

One would hope humans *don't* treat AIs (or aliens) as they treat other humans.

Dealing with AIs as if they were people is too low a bar; dealing with them as equals will be the real breakthrough.

I'm thinking of Melinda Snodgrass' MEASURE OF A MAN; how even humans with a couple centuries' exposure to "others" needed a Shakespearean-grade speech to accept Commander Data as a person. And that in a "hopeful" fictional universe. In real life...?

I don't think the problem is treating AI's as human - because like with other humans, that's really a decision left to the individual.

Likewise, many humans will also treat a completely inhuman AI as human because of our tendency to anthropomorphise things.

I think the greater question is how do you determine if an AI is worthy of being treated as a human, with the same rights and expectations.

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.

Regards
David

[John K]

Genesis by Bernard Beckett takes an interesting look at these questions.

Blindsight by Peter Watts, to be honest, torqued my mind a bit. He comes at the question from the opposite point of view: What IS the human idea of the relationship between intelligence and consciousness and is it a rational use of resources? Could there be other intelligent life in the universe that functions in a way that we would call AI but would actually make more sense? Is human consciousness an anomaly that that is doomed to fail the evolutionary test of time?

Blindsight is available here at MR for free.

[Freeshadow]

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

[EatingPie]

Originally Posted by Steven Lyle Jordan

To me, the AI sentience matter--can it be considered "alive", does it have a "soul"?--is a 20th century trope that I believe we're already moving past with Iron Man's Jarvis and other similar creations. Its parallel trope, the humanoid robot, is a late-19th/20th century trope, which is also left behind by the 21st century. The matter of soul will forever be an intellectual exercise only, as long as we cannot prove the soul exists. The only real issue is, Can humans deal with AIs as they would with other people? And if so, does that qualify AIs as Persons?

Hah, we can't even deal other people as other people!

Even though I'm kidding, it's true, something we see demonstrated on a daily basis -- either by ourselves or others. And this is explained by a central tenet of Christianity (sin), which would then lend credence to the notion of a soul.

So I would not be so quick to dismiss it as a trope. Nor would I relegate it merely to the 20th century (beside the fact that the 21st Century opened with one of the most horrible atrocities in the US history). The soul is an ancient concept.

-Pie

[Solitaire1]

Originally Posted by Steven Lyle Jordan

Huh! Silly bag of mostly water! How can you mention Astro Boy and dare to leave out 8th Man?

And as long as comic characters are on the board, let's mention the house/armor AI "Jarvis" from the Iron Man movies.

It occasionally surprises me that I've mostly written stories with "weak" AI in them, though my Verdant series includes a "weak" AI that may or may not be approaching sentience, thanks to the events of Verdant Skies (I'm purposely keeping it vague, though I'll clear up the sub-story if I ever write a third book).

To me, the AI sentience matter--can it be considered "alive", does it have a "soul"?--is a 20th century trope that I believe we're already moving past with Iron Man's Jarvis and other similar creations. Its parallel trope, the humanoid robot, is a late-19th/20th century trope, which is also left behind by the 21st century. The matter of soul will forever be an intellectual exercise only, as long as we cannot prove the soul exists. The only real issue is, Can humans deal with AIs as they would with other people? And if so, does that qualify AIs as Persons?

I apologize for not mentioning "8th Man." I simply overlooked the series when I was writing the anime list, based on the DVDs I had on hand. But it does bring forth a question: Since 8th Man was created from the downloaded memories of a murdered detective, does he fall under the category of an A.I. or a just a human intelligence transplanted into a robot body?

[Nancy Fulda]

Originally Posted by Freeshadow

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.

This is one way. I think a system based on genetic algorithms might work as well. But I don't think we should discount the importance of hand-coding.

If you look at a human, we come pre-packaged with a set of reflexes, nonverbal communication strategies, and urgent needs that function as a foundational framework. Even language is built in. A baby isn't born knowing its native tongue, but the ability to parse sound bytes into individual words and the drive to communicate are inborn. Any learning we do on top of all that really is just the tip of the iceberg. An important tip, but a tip just the same.

Grammar is not consistent across languages, but the idea that there must be a grammar and that it must be discovered and utilized is present in every infant. A baby's -- in all senses extraordinary -- ability to learn whatever language it is presented with is possible only because a lot of the work was already done.

Regarding Strong AI, the problem is finding the right combination of hard-wired scaffolding and learning algorithms.

[fjtorres]

Originally Posted by Nancy Fulda

Regarding Strong AI, the problem is finding the right combination of hard-wired scaffolding and learning algorithms.

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.