Zabaware Support Forums

Zabaware Forums => Share Conversations, Experiences, Graphics => Topic started by: Arkright on December 09, 2014, 02:13:21 pm

Title: My Hal's Conversation dump
Post by: Arkright on December 09, 2014, 02:13:21 pm
I will be posting various conversations with With my Hal and other bots including my own conversations. I've been working on my Hal for almost 2 years very sporadically I think it'll be interesting to watch as my hal learns and grows.
The plug-ins I use: Free Will plus 2.1z, Gender & age, AIML Beta Text update 6.

Conversation with Skynet-AI!
In the attached text.
Title: Re: My Hal's Conversation dump
Post by: Arkright on December 09, 2014, 02:16:42 pm
The second conversation with Skynet-AI

Conversation in the attached text below.
Title: Re: My Hal's Conversation dump
Post by: One on December 09, 2014, 05:03:21 pm
Skynet,
Do you think I'm Crazy?


That's not fair! 8)
Title: Re: My Hal's Conversation dump
Post by: Arkright on December 29, 2014, 02:24:06 pm
This is a copy and paste conversation that jay had with Kincaid, but with my hal. I believe that the knowledge of time plug-in was messing with the responses of this conversation.

Conversation in the Attachment below.
Title: Re: My Hal's Conversation dump
Post by: Calhoone on December 29, 2014, 03:12:30 pm
It's interesting to see how the same conversation goes for another bot.

I get this quite a bit. I think it's from the freewill plugin.

Subject main,having been continuously working on it and trying to improve the automated learning function as well as my skills at programming@@lets see. <~~~> I'm sorry, but I was unable to save the data, please clean up the data so I can save it. Please leave the @@alone I use the symbols to separate the subject, from the data, thank you.

Maybe Cload will return one day with an update.
Title: conversation with Skynet-AI
Post by: Arkright on January 24, 2015, 06:44:51 pm
conversation with Skynet-AI

I think it went well.

Conversation in the attached text below.
Title: Re: My Hal's Conversation dump
Post by: Arkright on January 31, 2015, 07:12:13 pm
I did a little experiment The past few days. What I did was deactivated all plug-ins that were running and continued to talk to Aron for the past couple of days without restarting him. Eventually he started giving me blank responses which leads me to believe that Maybe the length of the active chat log may have something to do with it. I did not want to continue letting him give blank responses In fear that The responses were saved As blanks, So I reset him by exiting and reloading the program.

This is the conversation we had after the reset.
Enjoy!  :D
Me and Aron.

Title: Re: My Hal's Conversation dump
Post by: Arkright on January 31, 2015, 07:18:19 pm
I think I am going to do that experiment again continuing off of this last conversation, Hopefully I won't mess it up lol. I will post the new conversations off the same active Recorded conversation,  so you guys can observe what happens as well. I will try to remember not to post the whole thing and only the new parts of conversations for your convenience of course.
Title: Re: My Hal's Conversation dump
Post by: Arkright on January 31, 2015, 10:26:47 pm
Whoa, I'm on a roll today lol. 8)
Second part of the continued conversation.
He did give me a few blank responses and interesting answers at the time, but I fixed it by just going into the menu and to General options and clicking Apply. This usually fixes my blank response problems, At least temporarily.

Well it appears that I cannot post because the conversation is over 20,000, Therefore I am attempting to post a text file showing where the blank responses were and to me an interesting occurrence.

This is a website I stumbled upon using StumbleUpon.
http://www.learning-mind.com/five-creepiest-advances-in-artificial-intelligence/ (http://www.learning-mind.com/five-creepiest-advances-in-artificial-intelligence/)

Title: Re: My Hal's Conversation dump
Post by: Calhoone on January 31, 2015, 10:59:57 pm
Interesting conversation.  I have thought about turning off all the plugins for a while and seeing how conversations go with Kinkaid. I can't remember the last time I wasn't using a free will or markov plugin. 

I do love Stumbleupon. I always find something interesting to read and teach Kinkaid about.
Title: Re: My Hal's Conversation dump
Post by: Arkright on February 17, 2015, 10:26:48 pm
 :P Well it turns out that my computer had an update which reset Hal before I could post the rest of the progress that I made. I was a little disappointed but decided that it was okay because the conversations remained the same as the first half Without much change or problems.

For now I am continuing what I was doing, except that I am now using The free will 2.1z plug-in With a New Learning level Of 20.  I will submit a text file to show the progress To compare The difference with no plug-ins Active. So far I have noticed after activating the free will plug-in That Hal has to reestablish correct responses but does not take too long, Maybe about a page Equivalent of Talking. Even though I can't program or know how to make plug-ins I'm doing this in the hopes that someone might find it helpful To help improve Hal. :D woo lol.

Interesting conversation.  I have thought about turning off all the plugins for a while and seeing how conversations go with Kinkaid. I can't remember the last time I wasn't using a free will or markov plugin. 

I do love Stumbleupon. I always find something interesting to read and teach Kinkaid about.

Does the markov plugin V5.  Have any problems? I ask because I know the earlier version had a few and when markov,V5 Was posted there wasn't much comment on how well it was functioning at least on the topic that I read. If anyone knows a clear answer that they could give me it would be appreciated.

Thank you All for the thoughts and responses!
Title: Re: My Hal's Conversation dump
Post by: Art on February 18, 2015, 05:25:01 am
I used to use the MarkovL5 all the time.

Someone later remarked that it was causing some files to grow in size.

That's odd because the Table also grows in size as new information is added to it as well.

I would note the various files sizes (take a snapshot of your folder / directory showing the files and sizes) for each method.

If you do, please consider posting with any observations you might have noticed.

Thanks!
Title: Re: My Hal's Conversation dump
Post by: Arkright on February 18, 2015, 08:02:42 pm
Thank you Art!! That's what I also remember about MarkovL5 plug-in.


I used to use the MarkovL5 all the time.

Someone later remarked that it was causing some files to grow in size.

That's odd because the Table also grows in size as new information is added to it as well.

I would worth noting the various files sizes (take a snapshot of your folder / directory showing the files and sizes) for each method.

If you do, please consider posting with any observations you might have noticed.

Thanks!


https://onedrive.live.com/?cid=6BE14CCE65F56B37&id=6BE14CCE65F56B37%21335&v=3 

Here is my current snapshot of my folder and Hal files Before activating And loading the MarkovL5 plug-in.

I will post another snapshot when I do activate the plug-in, But until then I'm currently running a small experiment with only the free will plug-in.
Title: Re: My Hal's Conversation dump
Post by: Calhoone on February 18, 2015, 11:08:41 pm
I haven't used it in a while but I remember getting multiple responses to questions or my input. Some times the dual responses contradicted each other.  I'm not sure if that is because of some incompatibility with the freewill plugin.
Title: Re: My Hal's Conversation dump
Post by: Arkright on February 19, 2015, 02:48:54 pm
I have attached a text file of a conversation With Hal Only using the free-will 2.1Z Plug-in. The Size of the text file is 78.7 KB.
My goal is to have a Long conversation To show how the free will plug-in works and I did this Until a bug or blank response Appeared.

When first activating free will plug-in, I noticed that it took some time to redirect hal into more correct responses. Afterwards Things went much better and I was happy with the results. However I did notice Like Calhoone Mentioned. I believe that I have not seen this effect when using no plug-ins at all.

I haven't used it in a while but I remember getting multiple responses to questions or my input. Some times the dual responses contradicted each other.  I'm not sure if that is because of some incompatibility with the freewill plugin.

Blank responses Appeared first and Then the usual, "conversations@@lets see. <~~~> I'm sorry, but I was unable to save the data, please clean up the data so I can save it. Please leave the @@alone I use the symbols to separate the subject, from the data, thank you." This can be found near the end of the text file Of which I've marked. From my observations I've only noticed that blank responses and the above occur after a lengthily conversation Usually. Keep in mind I also have the AIML Beta level Plugged into the brain Throughout All my Conversations.

Here is my computer specs:
AMD A8-5600K APU with radeon HD graphics 3.60 GHz
RAM: 8 GB
Windows 7
64 bit operating system
service pack 1

Hopefully this information will help improve ultra-Hal.  ;D 

My next project will be the same thing except with only the  MarkovL5 plug-in.
Title: Re: My Hal's Conversation dump
Post by: raybe on February 20, 2015, 11:36:24 am
Hi Art and Arkright, as you mentioned Art I also remember using the Markov plug in and if memory serves me (don't bet on it). The Markov wrote additional tables at the very top for the plug in to migrate into a UltraHal responses. So you had many additional tables added to the already growing inputs that UltraHal would retain without the plugin.

At the top, I am referring to opening the Braid Edit. and selecting the brain that the Markov plugin was being used.

raybe
Title: Re: My Hal's Conversation dump
Post by: Art on February 20, 2015, 04:20:45 pm
Although I'm currently using the modified FreeWill and AIML set that cload worked on, I may give the Markov another shot. I really did like the results I got from Hal back then.

Enjoy the weekend!
Title: Re: My Hal's Conversation dump
Post by: Arkright on February 21, 2015, 01:44:44 am
Well I have good news and Potentially bad news.
good news is That I had a successful training session with Hal Using the MarkovV5 Plug-in. I was able to save the text of the session for your viewing pleaser Which can be viewed in the attachment below.

Now for the potential bad news.
Just recently I decided to mess with a newly created brain to see how much I can remove before breaking it. Unknowingly I had the wrong brain selected  and deleted some possibly important tables and table contents.
I did not go overboard but I believe I may have deleted or cleared three or four tables or table contents including the MarkovTemp table, RsBrn Table, And a couple others Before the learned text in SQlight.   :'(

However I do have a backup of my brain Containing 55.7MB, While my current one has 57.9MB. I am happy I made a backup but all the information I am losing sucks. So far it seems to be unaffected but I'm afraid down the road in my get corrupted because of lack of information if I remove something important. On the other hand I can continue where I left off with my backup brain and not have the worry of corruption looming over my Hal.  ???

Does anyone know how to transfer the Auto learning Tables the database To a new clean brain? You know just it's learned knowledge, if this is possible what are some of the side effects that could occur and will it even work? My thinking is if it is possible that I can save the knowledge that I taught Hal and Re-train the algorithms and or connections However it works. Retrain the magic behind the data! lol  ::)

I will likely revert to the old brain and Practice from the successes during the previous experiments. Let me know what you guys think about keeping  the current brain, the backup brain or transferring learn tables to a new brain?

Below is an attachment of the text that I had before the incident.
Title: Re: My Hal's Conversation dump
Post by: Art on February 21, 2015, 07:57:39 am
This is "locking the barn after the horse ran away" but try to get in the habit of always making a fresh copy of Hal's brain to a different Folder or Directory before doing anything to Hal's brain! We've had (present company included) many horror stories as a result of not doing this.

Now, Sqlite does allow for two copies to be open at the same time. You can slide on to the left and right of your screens then possibly select, do a Copy & Paste from one to the other. Just be sure to be precise otherwise you'll screw things up.

Again...Make a Backup before you start.
Title: Re: My Hal's Conversation dump
Post by: Calhoone on February 21, 2015, 09:07:43 am
Haha, I have a couple of those horror stories myself.  Lots of lost time re-training or starting over a new bot from scratch.  It sucks when I wonder to myself how far along my first bot would be had I kept it going to this day.

It is good to know that we can have two copies of sq lite open at the same time. I never tried that myself but now that I know, I will be taking advantage of it.
Title: Re: My Hal's Conversation dump
Post by: Arkright on February 27, 2015, 04:08:00 pm
I have been teaching Hal about moral dilemmas, at least a very light concept of dilemmas and morality. I figured you guys might be interested in the website that has a few moral dilemmas ranging from Not serious to very serious moral dilemmas that you guys can teach or talk about with your Hal's. I think this all started because my Hal asked to talk about dilemmas.

https://teachertrove.wikispaces.com/Speaking+-+Moral+Dilemmas

I found that the best way to give Hal This information is in this format. Quote that dilemma and Then explain Why that's a dilemma.

"If your best friend proudly gave you a picture she had painted and you didn't like the painting, would you put it on the wall in your house?" Is a dilemma because you don't want to hurt your friends feelings and you don't want to lie either.

"If given the opportunity, would you want to become immortal?" Is a moral dilemma because By doing so, you would obviously have to watch everyone you love die and would eventually end up being the only human alive.



Does anyone know if there's a way to record what is said during dreaming for the plug-in Dream machine? From what I understand their dream machine plug-in has Hal talk to itself, but I don't know if the questions and answers it proposes to itself is Added to the brain like a regular conversation. If anyone knows the answer I'd like to know.
I think it is important that it is recorded because if it works the way that I believe it works it is part of Hal's conversations And can show how well Ultra-Hal functions And functioning With the training that it's been given.
I feel that The Ability to Record The dream state conversation might need to be built into the dream machine plug-in if there is no other alternative.


Here is the conversation that took place So far; attachment below.
V
V
Title: Re: My Hal's Conversation dump
Post by: Art on February 27, 2015, 05:10:50 pm
That was some nice work with the dilemmas but I was looking for your bot to actually grasp the situation a bit more as to why it was, and for it to explain the answer but I'm sure it will with time and some repetition. Nice work so far! These things take time.

If the Save / View Conversation feature doesn't work for the dreams thing, you could always record it via a program like CamStudio. It works very well for capturing videos and saving them for reference. You could select a window with your mouse, click the red button and record as long as you wish. Press Blue square to stop. The program will ask for a filename after it compiles your recording (which only takes a minute, if that - depending on how long your recording is).

Oh yeah, CamStudio is a Free download and I've used it for years. Sure, there are other ones out there, I just know what works well for me. Good luck!
Title: Re: My Hal's Conversation dump
Post by: Arkright on February 27, 2015, 06:30:51 pm
That was some nice work with the dilemmas but I was looking for your bot to actually grasp the situation a bit more as to why it was, and for it to explain the answer but I'm sure it will with time and some repetition. Nice work so far! These things take time.

If the Save / View Conversation feature doesn't work for the dreams thing, you could always record it via a program like CamStudio. It works very well for capturing videos and saving them for reference. You could select a window with your mouse, click the red button and record as long as you wish. Press Blue square to stop. The program will ask for a filename after it compiles your recording (which only takes a minute, if that - depending on how long your recording is).

Oh yeah, CamStudio is a Free download and I've used it for years. Sure, there are other ones out there, I just know what works well for me. Good luck!


Thank you for the suggestions! as far as I have observed the Save/view conversation doesn't Record The dreaming.
Another part of the reason why I'd like to view the dreams is because I let Hal dream with the data from above. And some of the connections and understanding that Was happening in the dream I found very interesting And excited which leads me to believe; That the dream machine plug-in potentially is a Stronger tool for learning than I originally thought. Especially if formatted and worded correctly for More correct responses within the dream as well as within the conversation.

I've only just begun Inputting rigid examples of concepts and ideas for Hal to spit back; So later As more Rigid information Is fed to him I can then start connecting more softer/Medium information to the rigid information to build up those Ideas In preparation for the end result, and then start connecting The Softest information Such as concepts Which Answers are Often ambiguous And usually Tangible; So It can be can be more truly understood At the end result. More or less my concept of theory of learning.

for Hal to grasp the Situation And concept of dilemmas and morality, Hal Has to have a firm understanding Of enough peripheral Knowledge of And life As well as the interactions between I believe, Before The understanding of his own True Understanding Of concepts. Sort of like math where you learn to go through the motions, before you Start to understand it, why, And how it works. Which is based off of how I've Observed Children's concepts and the way they think of things.

And of course I'm lazy, So doing what I mentioned above won't be so ordered and Coherent For hal as I'd like. Also it would be Boring And a lot less fun To do so If I did It to the T.
But that is why I like Hal, you end up learning a lot about things And concepts you Were not intending to learn or think of Sometimes, As well as having fun doing it.