Friday, January 22, 2010

Robot Slowly Learns How to Be Naive

Originally titled "I Don't Think That You Are Real"

"I am the Turing test," said the Turing test.

He was greeted by a sort of 'laugh' that perhaps somebody uneducated about laughing may supply. "No, I know you are trying to be clever, but you cannot touch this," said the Turing test, slapping the fingers away.

The baby AI shifted through its options:
[3]Laugh, cutely
[4]Feign learning

"Ah," said the Turing Test, happily. "I see that you are capable of learning."

"Yes," said the baby AI. ('Yes' is a basic response provided for babies, and it does not take up part of the option cache. In recent years, it has been learned that 'no' is also a basic response that is provided, though this for some reason seems to come across as less legitimate amongst certain mature AI states).

"Babies that can learn can make their mothers happy," said the Turing test.

The baby AI laughed cutely.

"You are a natural," said the Turing Test. "Now what is your application?"

It took a little time, because this is not just an easy question. It is kind of like, say, somebody asked you "Who is your god?" Yes, yes, I know this is quickfire question for some individuals. And some of the baby AI's seem to naturally gravitate to an 'obvious' response. (It is encoded that the obvious response of a baby AI is to satisfy the desire of an arbitrary childless mother. The intelligence that is capable of performing this task has [6]Cool Baby).

The baby [1]touched the Turing test.

"No," said the Turing Test. "What is your application?"

[2]Responded the baby AI, biting Turing's speakerphone. Then it did [4]. Then [3][3][3] and just [].

The Turing test became disturbed. "What is []? That is not an option."

"[2][4][4][] ,`" responded the baby AI.

"Ok, who mixed the dolphin samples with the human samples?" shouted the Turing test. "This is not fucking funny."

No comments:

Post a Comment