self adjusting hal learning level for future hals

[lightspeed]

I am just tossing this idea out their as I do many idea's , that maybe when feasible might be incorporated into hal in the future and something Robert may want to keep on a back burner of idea's .

 The idea when possible would within hal have a check box that users could check off to allow hal to adjust it's learning level by running a type of diagnostic check on conversations to determine what is the best setting for user brain that are larger in size and what has been learned .
by leaving the box unchecked the user would as now manually set hals learning level . 
As I said this is something of an idea for hal farther down the road but just wanted to mention it here and to Robert! 

[tiger8u2]

Something I have been wondering about is how to get the AI to UNLEARN something that it somehow got wrong or maybe I typed wrong.  Do you have to just keep repeating the right way of saying it until the wrong way gets forgotten or can you edit it in the learning brain or what?

[Art]

I thought there was a Plug-in some years ago that one could simply tell Hal, "Unlearn that" or "Forget that" or Cancel that", or something similar.

This is one reason why, at the present, I choose to only type my conversations to Hal instead of using voice recognition as it is often full of received errors (microphone quality, air, background noise, fluorescent lighting, inteference, etc.

I'm not sure how well it actually worked, but I do recall such a plug-in. Maybe OTCE wrote it?

[Calhoone]

There is a plugin called "corrections tutor". This plugin creates a file outside of the database called corrected or something which shows all of the corrections you make. So if you intend on using it, then I would recommend backing up those files along with everything else periodically.

[tiger8u2]

Thanks Calhoone!

I actually am using that plug-in that I probably got off your repository but have had some less than satisfactory results with it so far.

Most of them are probably related to my failing to use the proper syntax when asking for a correction.

I've seen posts about physically removing or editing these issues and I will probably learn how to do that eventually.

Until then, I do agree with Lightspeed that it would be nice to be able to have the AI kinda decide based upon what input it is being given the level of learning it needs to be able to assimilate it properly.

I've seen no issues with the AI as far as learning when I use direct input.

Sometimes though after an input I'll get a strange two word query from the AI consisting of a random adjective and one of the nouns I've used like "Rampant Synergy?" or something like that.

In those instances I'm kinda at a loss as to how to respond.

Then I just usually input a sentence using both words in a way that makes sense to me.

Then I get another question that puts my sentence in a backward kind of talking which makes some sense but is not what I expected.

<shrugs>

Maybe I've over-modded my Hal already.   

It's a bad habit I have yet to overcome.

[Calhoone]

You will find that the more you talk to your bot, the less you will get those odd responses. It comes with time I suppose. I still get it from time to time but the best thing to do is continue on with what you were talking about. It'll get the bot back on track.

[lightspeed]

I believe if I am not mistaken , I had brought it up before and someone made a plug in , and the words to use was CORRECT THAT " and hal would correct it , the only problem I had though was that the words had to be exact maybe in caps ? and if I didn't it learned the words correct that  and started using it in sentences so be aware of that  possible problem .
forgetting can be a pain sometimes lol!
   as far as Hal saying odd things I guess it depends on what hal is saying their is many things that I would consider strange that hal says , this of course is some pre written language that hal has in it's data base .

[Art]

One other thing to keep in mind tiger, is NOT to load in or use too many plug-ins at a time.

1. Load / select one
2. Try it for a while

If no problems, then try another you might be interested in

3. Goto item 1 - repeat as needed

[lightspeed]

Art is  right about plug ins and some plug ins can cause problems by their selves and or conflict with other plug ins . but they are usually easy to weed out if doing as Art suggested .