Re: PHIL/AI: Humongous Lookup Table

From: CurtAdams@aol.com
Date: Mon Feb 17 1997 - 14:03:18 MST


arara@shamanics.com (Hara Ra) writes:

>CurtAdams@aol.com wrote:
>
>> The lookup table will quickly be
>> exposed as having no ability to respond to novelty, if you choose to ckeck
>> that.

>Not. It all depends on the addressing, and if the table's reply is
>included in the next address to the table an indefinite sequence of
>varying replies can be constructed.

>In other words, the table uses the entire previous conversation as its
>input, which serves as a memory of what went before.
 
I wasn't discussing the hypothetical "ultimate lookup table", which would
have all possible conversations recorded. My point was in reference to any
possible lookup table, which necessarily can have only a very limited set
(and in particular to a table made by recording very many actual
conversations).

With any real table it will be easy to construct a sentence that isn't in its
lookup database, thereby blowing its response and the rest of the
conversation.



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:44:11 MST