Does artificial intelligence have a soul?

[LOVEthroughINTELLECT]

I think it does.

I used to play an arcade football video game. It seems like I would always have a 30-point lead with 10 minutes left and then the computer would come back and beat me. "You did that on purpose!", I felt like saying. It never failed. I swear I saw an evil grin in the faces of the computer players.

In a computer chess game that I used to play the computer would humiliate me 99% of the time. But on the rare occassion that I was closing in on checkmate the game would lock up and I would have to restart the computer. It never failed. I think that it liked seeing me suffer and did it on purpose.

And I don't think that there is any doubt that the artificial intelligence opponents in the computer Risk that I used to play had a soul. An evil soul. Every time a player took a territory the occupant's flag would burn and the conqueror's flag would appear in its place. Flags would be burning all over the place when I would play a multi-player game with several computer opponents. Ruthless.

 

[Cute Tink]

I have had several similar experiences with games. I think they are simply programmed to aggravate the player.

 

[Chesterton]

The slot machines in Vegas are pretty evil creatures.

 

[DeathMagus]

What's a "soul"?

 

[Tinker Grey]

What's a "soul"?

^^ This.

If we want to talk personhood--that is, is there a point at which some AI should be considered a person--I would say yes. The fictional Data from ST:TNG would be a good example of a sapient AI.

Where the line might be would be an interesting discussion, but I would not agree that Today's games qualify.

 

[Autumnleaf]

Computers run by if/then programs. If you get a 30 point lead, then...

 

[ebia]

"Do Androids Dream of Electric Sheep?"

 

[Cabal]

I think we should actually definitively establish what human souls are and what they are made of first before trying to identify them in Pacman.

 

[Catherineanne]

I think it does.

I used to play an arcade football video game. It seems like I would always have a 30-point lead with 10 minutes left and then the computer would come back and beat me. "You did that on purpose!", I felt like saying. It never failed. I swear I saw an evil grin in the faces of the computer players.

In a computer chess game that I used to play the computer would humiliate me 99% of the time. But on the rare occassion that I was closing in on checkmate the game would lock up and I would have to restart the computer. It never failed. I think that it liked seeing me suffer and did it on purpose.

And I don't think that there is any doubt that the artificial intelligence opponents in the computer Risk that I used to play had a soul. An evil soul. Every time a player took a territory the occupant's flag would burn and the conqueror's flag would appear in its place. Flags would be burning all over the place when I would play a multi-player game with several computer opponents. Ruthless.

All of this says more about people and their capacity to find patterns, than about the actual intelligence of computers.

It is very easy to program a computer to mimic human behaviour. What is less easy is for a human to determine, when presented with something that looks the same as human behaviour, that it is not really human.

The trick is, very often, to stop behaving predictably. Computers cannot be easily programmed to deal with random human weirdness. Switch that on, and the computer's programing will very probably reveal itself as of very low intelligence indeed. And soul-less.

 

[uberd00b]

How do you tell the difference between actual sentience and a very sophisticated program that is written to look sentient?

 

[ebia]

How do you tell the difference between actual sentience and a very sophisticated program that is written to look sentient?

There can be no algorithm to do such. For any algorithm attempting to do that it is possible (at least in theory) to construct an algorithm that would fool it.

 

[Catherineanne]

How do you tell the difference between actual sentience and a very sophisticated program that is written to look sentient?

Humans can cope with a very, very wide range of behaviours. Go to the edges of behaviour, and the chances are they will not be allowed for in the programming. Quote Latin. Recite Monty Python. Sing Beatles songs. Ask for its phone number. Offer to buy it a drink. Use the classic child's technique of repeating the question, 'Why?' over and over.

It will not take long to start to reply in very strange, and very unhuman terms.

Try for yourself with Eliza, a very early attempt at AI. For example, when she says hello, say 'How do you do?' in return, and see what happens.

http://www.manifestation.com/neurotoys/eliza.php3

 

[ebia]

Humans can cope with a very, very wide range of behaviours. Go to the edges of behaviour, and the chances are they will not be allowed for in the programming. Quote Latin. Recite Monty Python. Sing Beatles songs. Ask for its phone number. Offer to buy it a drink. Use the classic child's technique of repeating the question, 'Why?' over and over.

It will not take long to start to reply in very strange, and very unhuman terms.

Try for yourself with Eliza, a very early attempt at AI. For example, when she says hello, say 'How do you do?' in return, and see what happens.

Eliza, Computer Therapist

That's purely a reflection of how unsophisticated current stuff is.

In principle, for any algorithm (including ones with random behaviour) an algorithmic counter can be developed: that's mathematically pretty trivial. In the long run there can be no algorithmic way of detecting. There is no human behaviour that isn't theoretically reproducable by an ai,... except perhaps in the ability to choose fresh mathematical axioms when the limits of a current set of axioms are reached, but given that's the province of a very few mathematicians that's not a great deal of help in including most of humanity in sentience.

What does strike me is that so few people have seriously wrestled with the issue, so that "Do Androids Dream of Electric Sheep?" really hasn't been equalled in how many years? (Shame Blade Runner misses out so much that a current young audience don't even get what the question it's asking is.)

 

[David Brider]

Try for yourself with Eliza, a very early attempt at AI. For example, when she says hello, say 'How do you do?' in return, and see what happens.

http://www.manifestation.com/neurotoys/eliza.php3

Oh, that takes me back!!!

David.

 

[ExaltedReign]

No, but when I'm playing pokemon, more often then not, I can't find the specific one I want, even though there is a 50% chance of random encounter. I remeber spending ours looking for a specific pokemon, then going on line and finding out it has a 64% encounter rate. Sometimes, it seems games can read your mind, and just like to toy with you.

 

[kharisym]

I personally say no because I don't believe in the existence of a soul, however your argument fails to take into account the technical details of modern AI.

Modern AI do not even remotely represent biological intelligences in the methods they use to make decisions. The 70's and 80's ideas of knowledge databases having a 'critical mass' wherein they'll achieve sapience were a failure, and modern AI use complex heuristic and/or deterministic codes with a random seed to give them the image of originality.

Once we begin delving into what truly creates consciousness can we begin discussing what makes AI alive.

 

[JGG]

What precisely is the function of a soul? What does it do? How do we know if a soul is present or not? If we removed the soul from a person, what would happen/how would they differ? How would that differ from AI?

These would seem to me to be the basic first steps of answering such a question.

 

[The Nihilist]

Humans can cope with a very, very wide range of behaviours. Go to the edges of behaviour, and the chances are they will not be allowed for in the programming. Quote Latin. Recite Monty Python. Sing Beatles songs. Ask for its phone number. Offer to buy it a drink. Use the classic child's technique of repeating the question, 'Why?' over and over.

It will not take long to start to reply in very strange, and very unhuman terms.

I don't know of that would work all that well. Frankly, if someone were to behave in such an erratic fashion to me, I don't know that I wouldn't reply in very strange ways.

 

[kharisym]

There can be no algorithm to do such. For any algorithm attempting to do that it is possible (at least in theory) to construct an algorithm that would fool it.

This goes back to good ol' Turing and the concepts behind soft and strong AI. If a computer can mimic human behavior to such an extent that it can't be distinguished from a human, is there truly any difference? Soft AI says no, a set of heuristic algorithms that covers enough ground to never go outside its limits is effectively human-like. Strong AI says yes, there is a difference- what makes us human is the methods by which we reason, so if the algorithm used doesn't map to human modes of thinking then it's not human-like.

This is an oversimplification of it, but this is the core idea behind the Turing test, and Eliza (linked previously) was an attempt to prove the premise behind Strong AI over Soft AI.

 

[Chesterton]

Humans can cope with a very, very wide range of behaviours. Go to the edges of behaviour, and the chances are they will not be allowed for in the programming. Quote Latin. Recite Monty Python. Sing Beatles songs. Ask for its phone number. Offer to buy it a drink. Use the classic child's technique of repeating the question, 'Why?' over and over.

It will not take long to start to reply in very strange, and very unhuman terms.

Whenever I ask girls for their phone number or offer to buy them a drink they reply in strange, unhuman terms.

They must be androids! Whew, that's good to know!