Re: PHIL/AI: Humongous Lookup Table
Tue, 18 Feb 1997 20:29:51 -0500 (EST) (Michael Lorrey) writes: wrote:
>> With any real table it will be easy to construct a sentence that isn't in
>> lookup database, thereby blowing its response and the rest of the
>> conversation.

>The solution is to operate with more than one type of lookup table, sort
>of a super thesaurus/dictionary. One table for definitions, one for
>synonyms, one for antonyms, slang, plus grammatical logic, as well as a
>mathematics/english translator (which can for example extract data and
>operators from a paragraph structure for actual calculation).

>Given this, one such AI could even deal with new information by
>analysing it and analogizing it to existing knowledge.

Well, duh. If you write an AI that can understand English, it can understand
English. But the sort of thing you describe is very difficult (none of it
has been done; you need much more than just lists of synonyms to use them;
you also need and understanding of how the words differ in connotation). The
intelligence here would be coming from places other than the lookup table of
conversations. Indeed, it really wouldn't need a conversation lookup table
at all, anymore than we do.