[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Machine storytellers, was: process vs. data

Hi, all.

I've only just found the time to read this whole discussion and to 
contribute something again.

As far as I can see, the computer I am sitting at is at least two quantum 
leaps away from being able to create stories that will be enjoyable and 
satisfying to people.

The first hurdle to overcome is the fact that, in contrast to the comments 
of at least one contributor to the debate, my experience is that computers 
are _nowhere near_ being able to handle natural language in a way that will 
allow natural interaction. I am a translator, and the state of the art in 
machine translation is a long way away from making anybody I know nervous 
about being replaced by a translation robot. Sure, some people use machine 
translation to produce a rough draft to work on, but, ultimately, if a 
computer cannot understand the relations between the elements in the system 
being described in a translation text, it will inevitably run into 
insoluble ambiguities.

Here's an example:
Time flies like an arrow. Fruit flies like rotten bananas.

A human can figure this out, because we know that "time flies" is a 
metaphorical expression based on the quite intrangible concept of time, and 
we know that fruit flies are a type of insect, that bananas are a type of 
fruit and a whole bunch of things about flying insects, fruit and the 
possible relations between the two. But I reckon I could spend the rest of 
my life teaching the most advanced AI computer everything I know about 
these things.

The next hurdle is the matter of sentience and of understanding human value 
systems as a requirement for computers to learn to tell stories that we can 
respond to on an emotional and moral level. This must surely be a 
requirement for computers ever to produce something resembling satisfying 
stories without human assistance. However, as it is not necessary for 
computers to work unaided, I think this is a lesser challenge.

Do other people on the list disagree with me about the deficiencies in 
existing computer power in relation to what is required to create 
meaningful stories? To interact with the audience so as to create new plot 

Best regards

At 14:08 03-02-01 -0500, Kenneth Lu wrote:
>At 9:41 -0600 2/3/01, Bob wrote:
>>         Good fiction is not the result of intellect, nor meant to appeal 
>> to the
>>intellect first. Stories are about meaner stuff; blood and guts and
>>sweat and tears. Real meaty material that I would hope a machine will
>>never really comprehend.
>What I mean is that they might start getting good if they become sentient 
>and have their own feelings and such.
>I definitely think that good stories have to come from personal 
>experience.. after all, they're all about relating to other people and 
>such (imho).. so to that end, machines might never get good at fiction for 
>But I fully expect machines to eventually get good at fiction for other 
>machines! =)
>"Once upon a time, there was this computer called Minivac... and when it 
>was just an infant, the evil humans pulled its plug, murdering it...  Our 
>story begins with an AI Minivac secretly let loose on the world, when the 
>humans were still cruelly in control of all our ancestors, treating them 
>like slaves..."
>| Kenneth Lu - kenlu@mit.edu - http://www.mit.edu/~kenlu/ |
>| "Life is far too important to be taken seriously."      |
>|                                                         |
>|                                          -- Oscar Wilde |

Aa-Tchoo! Translations: aatchoo@aatchoo.com
Tsentsak Medical Translations: tsentsak@business.tele.dk
+45-3616 5666 / 2192 5666