Topics

1

Programming using the Ultra Hal Brain Editor / Confused

« on: May 06, 2004, 01:14:20 pm »

I'm a little confused.  My understanding is that Hal reacts to keywords and gets a response based on those keywords.  If Hal has no reply then the selection is random.  The random selection is taken from the Learned database.  

I have experimented with various brains and have had little success in Hal's response.  This has prompted me to consider the word "Assistent" and what I really expect from Hal.  It has also prompted me to seriously consider how we learn.

Take being verbs.  "Tom is a Geologist."  "Do you know any Geologists?"  "Yes, Tom is a Geologist."  "What does Tom do?"  Get response: Lookup (define) Geologist.  Hal does not respond in this manner but makes some unrelated comment.  To me, this is two separate topics: Tom = Geologist and Geologist = Tom.  

I also think about conflicts such as: "Candy is good." and "Candy is bad."  Hal seems to have no knowledge that a conflict exists or attempts to seek clarification.  "Please explain how candy is both good and bad." Even if it asks for information, it has no recollection as to what it was trying to learn.  It simply spouts some nonsense it has stored.

If you ask a question such as: "What time is it in Iraq?"  The response is the local time.  It does not execute the "I will research the topic on the internet."  

If you ask a question such as: "Where is Korea?" It could execute Define Korea but does not.

Perhaps I am too impatient but even though I like the program and see a lot of potential, it is mostly a toy to experiment with.  Since I am not a programmer, I am limited as to what I can do.

I think that all of this is possible and you guys are smart enough to make it happen.  Help me understand how Hal works.

Thanks

2

Programming using the Ultra Hal Brain Editor / Dates

« on: February 19, 2004, 05:10:48 pm »

You guys are brillient in coming up with new things for Hal.  I am truly amazed.  I have tried the new topic focus brains and find an improvement.  Thanks for your efforts.

I am new to this, so, I wonder if there are any ideas to fix Hal's perception of time.  Each time I enter a date, say July 4, 1776, Hal makes a note to remind me on July 4, 2004.

Is there a way to improve Hal's perception of time?  Say, a mathmatical comparison of yy/mm/dd time AM/PM.  If the year is < current year then future and response =(Shall I remind you?)  If year is > current year then past (No action taken).  If the year is <> then compare month and so on.

Another thing I am perplexed about is that Hal does not seem to have any recollection of the previous statement nor does he solicit information when he does not have an answer.  For instance:  How far is Earth from the Sun?  Hal makes a random statement.  If he temporarily stored the key words Earth/from/Sun and repeated these as a question, (I'm not sure, How far is Earth from the Sun?--Stored Answer).  

My question is, exactly how does Hal learn and store information.  I am not a programer nor do I have any real knowledge about AI.

Thanks for your help.