Share Site Map Contact us Home Page
Home > Theory > The Child Machine > Education


Research Plan

Applied Philosophy

The Child Machine
Education
State of Mind
First Words

White Papers

Eleanor

Introduction to Social Solipsism (off topic?)
Education
  Printable version
How do you educate a machine? Ai uses a simple system of rewards and punishments to train our child machine's independent, intelligent-seeming, but ultimately unpredictable actions.

Rather than trying to hard-wire the machine with grammar or a complex network of linguistic rules, we allow his brain to seek such rules on its own. When those rules produce language which appears intelligent, a trainer "rewards" the child machine, leading the machine to "remember" successful rules and promoting further applications of a rule in new situations and context; punishments teach the child machine what not to do, and encourage the child machine to suppress the application of a rule.
 
Turing's famous test for machine intelligence judges the intelligence of an entity regardless of its physical appearance: a human in one room is required to guess whether the entity in another room is a person or a machine - communicating only through a teletype or computer terminal.
 
The Turing Test is a practical measurement of the ability of software agents to perform "intelligently." The conversation, as Turing originally imagined it, might go as follows:
 
Q: Please write me a sonnet on the subject of the Forth Bridge.
A: Count me out on this one. I never could write poetry.
Q: Add 34957 to 70764
A: (Pause about 30 seconds and then give as answer) 105621.
Q: Do you play chess?
A: Yes.
Q: I have K at my K1, and no other pieces. You have only K at K6 and R at R1. It is your move. What do you play?
A: (After a pause of 15 seconds) R-R8 mate.
 
While there are many theoretical criticisms of the test as a measure of intelligence, anyone on the Internet who uses chat programs to talk to strangers participates in a kind of Turing Test, judging the qualities of the corresponding individual based on words alone, deciding whether the other individual is worth an investment in time and energy.
 
To create a machine which could have such a conversation, and millions of other possible conversations, we have created HALOne, comprised of a simple interface and a collection of learning algorithms.