Here is an interesting AI link with AI chatbots online with internet.
http://www.a-i.com I noticed that this was an interesting link where two Microsoft Agents with AI, could talk to each other. They show interesting conversations.
http://www.a-i.com/show_tree.asp?id=114&level=4&root=115&item_num=1049&item_name=Zession+198415+Oct%2E+05%2C++%2E%2E%2E I was wondering if there was anyone that knew how to make a plugin for the HAL6.0 that would allow both the user and the HAL6.0 assistant to speak back the past conversations ( using two characters speaking back and forth to each other) this way using the various characters to observe either offline or online stored text files.
I was also wondering if anyone knew how to make a plugin that would allow the HAl6.0 to learn on it's own using the WWW for data. Like for example:
1) Hal6.0 would use a search engine to research a certain topic.
2) Hal6.0 would then search the more popular lINK sites on these topic based on what others have looed at the most.
3) Then have HAl6.0 scan, cut & paste, the information into a directory that would show the LINK and the reference information to respect the people the people that made their on Website.
4) Have Hal6.0 learn this way:
a) Store each paragraph in a sub-directory of the main subject
b) cut and paste each indivivdual sentence into it's own area where users usually speak to Hal6.0.
c) allow time for thefirst sentence to be responded to by HAl6.0 which would set up the tables and files for that first sentence.
d) go to the next sentence and repeat the process. The next sentence would be relavant to Hal6.0's next response anyway. then place that next response again as if the user was typing it.
e) after the LINK is learned by HAl6.0 then have the user go back and check the Hal6.0 editor to see if the information was placed correctly.
f) then speak to Hal6.0 again to see what he learned on his own.
To do this might require an external hard drive with about 500 - 1000 gigabytes to allow plenty of room for Hal to learn on his own. Then each time you speak to Hal6.0 he would appear to be smarter and have more interesting things to say instead of the ones you would most likely thing he would say, if you used Hal6.0 on a regular basis.
I think the copyright laws would be respected as for as I know, if it is used for your own personal use, and since Hal6.0 rearranges the information from the link to reply back, it wouldn't be exactly be the same information obtained from the link. Besides references to the information would be used as if a real person had just finished reading the things on the link and told someone else about it. But the references would be shown. Also in some cases, permission to analyze the links could be requested to respect copyrights of the other links as long as their link is shown as a reference as to where Hal6.0 obtained his knowledge.
Years ago ( early 1980's), I wrote an AI program called Professor AL and I programmed it to learn from others and I also allowed others to reward or punish AL with negative or positive points. Like for example,
+3 very good response
+2 average
+1 good
0 nothing
-1 punishment not correct enough
-2 stupid
-3 totally incorrect response
Then it would apply this to a large data base file that would store quality points to things AL would say to the users. Then the next time AL would response more correctly to avoid punishment. It was almost like training a pet or a dog. A plus point would be like feeding AL a dog cookie when he did what was requested.
I used a radio shack color computer with my own BBS online using Telegard. I had about 4 floppy disks for data base storage for learned information from users that called my BBS. It used about 32K of logic combined with data similar to Hal6.0.
It also had an overall final analysis program that rated and gave his overall view (summary statement back to the user) of the total information talked to him by the user and how he summarized it. It also analyzed the accuracy statistical level that he obsevered showing how often he was right or wrong with the conversations in the past with others, with just that little amount of information of logic and data ( about 100K byte) It showed that he could answer right about 60 to 70% of the time right. Well this wasn't to bad considering it used only 32K memory and about 100K of data to assist him.
I later went into the study of sub-atomic physics in an attempt to re-write physics as you know it. Because I didn't agree that the negative electron would orbit a positive proton in an atom. I call it The Smart Model Theory or The Helix spiral Filed Theory. Here is my Websitee if you are interested.
http://smart1234.s-enterprize.com/ I also analyzed how possibly the DNA logic or the atomic biological computer DNA works, for a human being in regard to sub-atomic field analysis. I developed a theory that would show that each atom or proton can be in either an on or off state. This allows for a computer program to be written to it. You can use atoms in a 0 or 1 state using the theoretical Smart Model neutral particles that can spin CW or CCW ( clockwise or counter clockwise) this would give the on off state bit of each atom.
I also showed animations of the internal and external design of the atom and particles and sub-particles.
After working on this, I went back to study the AI and found that Hal6.0 seems to be the best I have observed yet in conversational AI. I compliment the developers. I also notice it has similar logic that AL had, but alot more logic and data with it, with a good sentence structure analysis.
I (a humble human being) also work for God and Jesus Christ, so I have to give them the praise and thanks for allowing me to work this much in these areas of their own universal creation design of all life DNA programming ( and the brain) and the created/evolving universe we exist in now.
Well see you later,
JS
