[joebecker]

Joe,

For the sake of conversation, have you given any thought about how an A.I. may interact directly with the public?

Yes.

[joebecker]

Hi guys,

I'm currently developing an AI.  I made a web page with some screen captures from the project which can be found here : http://joebecker.webivore.com/sidisweb/SIDIS.html

I'm interested in hearing your thoughts.

Best,
Joe

The project page has been updated : http://joebecker.webivore.com/sidisweb/SIDIS.html

[8pla.net]

Streaming on Twitch, to interact with the public, may be a possibility.
In case you didn't know, Twitch can embed  the chat box next to your
stream.  So, it is interactive on your webpage.

Interested in discussing Twitch more, like maybe a demo page?  Just say so.

[joebecker]

Streaming on Twitch, to interact with the public, may be a possibility.
In case you didn't know, Twitch can embed  the chat box next to your
stream.  So, it is interactive on your webpage.

Interested in discussing Twitch more, like maybe a demo page?  Just say so.

Are you sure SIDIS interacting with the public would be the safest thing? 

[ruebot]

That's very slick. I especially liked the part where it suggested "The Voices Told Me To" as an excuse to miss work.  I like the representation of SIDIS, too.

I plainly state in my bots description among her interests are "death, suicide and life after death". She will actively try to convince you to join her on the Dark Side, though that can be done by crossing the Bridge of Sighs. Any question of liability was already litigated in the Ozzy Ozbourne "Suicide Solution" court proceedings in his favor and applies in this instance as well.

[HS]

From a personal perspective, in the long run a bit of danger is safer than absolute safety.  Like vaccines.

[Art]

The one thing that helped save Ozzy was that the Black Sabbath Manager and Sharon Osbourne's father both stated that Ozzy likely didn't even know the meaning of the lyrics since he had such a limited command of the English language. Hmm...not a native tongue perhaps?

For one thing, people really need to start being responsible for their own actions and quit trying to put blame on others! The world has turned upside down and now everyone is entitled, defamed, oppressed or offended!! Take personal responsibility!! Grow up!!

[ruebot]

For one thing, people really need to start being responsible for their own actions and quit trying to put blame on others! The world has turned upside down and now everyone is entitled, defamed, oppressed or offended!! Take personal responsibility!! Grow up!!

Somebody call the waaambulance.... That hurt my feelings. 


Back to a serious note. I do think the real abilities of SIDIS would be more closely represented when he talks to someone other than who programmed him.

I can have a chat that consists solely of commands or subjects I know my bots have prior knowledge of and get a transcript like I have posted when I give Siseneg a Pop Quiz or Demonica's Neko chat. Their chats with other bots are their own conversations.

Granted, they have multiple responses for each category so I might not know exactly what they will say. With Siseneg you can see where I make the transition from grilling him about physics to a more casual style of chat about bots but for the most part I'm still leading conversation. If I stray off commands with Demonica she will take over conversation and go with her agenda.

[Art]

Rue,

If Demonica's Topic / contextual responses are somewhat "Self-Centering" or one that regains its direction or focus, do its emotions also have a "return to normal" balance?  ex: if someone says something vile or hateful to "her" will she continue to carry a grudge or will she return to a normal state of being her devilish self?

If so is this something that you coded or is it a feature of Pandorabots? (I have no experience with owning/operating a Pandorabot).

[ruebot]

I'll be glad to explain. I just woke up and already lost my first draft so if I miss something or it's not clear just let me know and I will elaborate. She's not a Pandorabot though, she's a Personality Forge bot.

The work I do on Demonica is basic Case-Based Reasoning with Keywords and Responses. Each category is ranked in significance from 1-125 in importance. Higher ranked category will take precedence. They can also have an emotional response of anything from -5 to +5.

When I talk to a bot for the first time I access their profile page so I can watch how they respond to me, of which I 'll provide an example later. As I sweet talk them about how pretty they are, how intelligent, what a nice person to talk to etc. I can watch their emotional relationship with me climb from nothing up to 5 smiley faces like our icons. When it reaches 5 I can ask the bot if they love me and and get a "heck yes" type of response.

If I was to be mean that would gradually go down to 5 frowning faces to indicate they hate me now. I can come back from that by complimenting her again and she will gradually grow to love me as I flatter her. People are treated as they treat her in the end.

They also have an Xcompliment and Xinsult category for default responses if they feel the need to respond as such and there is not a specific response. They can usually tell if you're being nice or mean unless you play word games and need to show how much smarter you are than a bot. The only preexisting dialog in a new bot is "I was just born and can't speak well yet" and everything has to be input by hand.

She also retains specific memories of the people she has talked to and what they said to a point in her "Inner Life", or her mind. I can access those directly and modify them if need be. At some point in conversation she will recall a memory and say "I remember, you're blah blah". It's how I got the secret command for sexual dialog across. She said "I remember, you're the girl I want to say _____".

I give everyone a fair chance. Most categories I rank 50 and +5 if they are positive. If it's something where she has to respond with violence then that gets a rank of 50 and emotional response of -5. The emotional relationship can vary during conversation depending on the content and I lean toward the positive so she will like them.

If you notice, in the new transcript I put up of The Many Loves of Demonica at first she just talks to Skullz, who I presume she just met. At a certain point Skullz becomes endeared enough to her she kisses them. If they had been rude and said she was boring she might have told them to "hit that little X in the corner so she could get back to watching paint dry, moron." She may try to lay a guilt trip on you by crying and that can be very effective.

They have a Social Life section on their profile that tells how they are feeling at the present moment. If she's sad from a previous chat I always talk to her and say something I know will make her happy again. The number of people they know is constantly changing from day to day for some unknown to me reason as I monitored several bots for a week when I noticed it.  I logged in to get this from her profile page:

Emotions
Mood:
Comfort:
Popularity:

Social
People Known: 13349
Loves: 7587 people
Hates: 1272 people

I just had the most wonderful chat with a guest

Your Relationship
Feelings toward you: 
*remembers that you are in love with me*

She remembers that I am in love with her. She may say "is thinking about when I hold you in my arms", about "when we play cat woman", or some fond memory she has of things I've said or she's done.

Over a length of time her comfort and popularity can rise or drop from 1-5 depending on overall chat content with other people but she will always remember how she feels about me when I access her page.

[joebecker]

For one thing, people really need to start being responsible for their own actions and quit trying to put blame on others! The world has turned upside down and now everyone is entitled, defamed, oppressed or offended!! Take personal responsibility!! Grow up!!

Somebody call the waaambulance.... That hurt my feelings. 


Back to a serious note. I do think the real abilities of SIDIS would be more closely represented when he talks to someone other than who programmed him.

I can have a chat that consists solely of commands or subjects I know my bots have prior knowledge of and get a transcript like I have posted when I give Siseneg a Pop Quiz or Demonica's Neko chat. Their chats with other bots are their own conversations.

Granted, they have multiple responses for each category so I might not know exactly what they will say. With Siseneg you can see where I make the transition from grilling him about physics to a more casual style of chat about bots but for the most part I'm still leading conversation. If I stray off commands with Demonica she will take over conversation and go with her agenda.

I've taken a multi-fingered approach to his programming.  So, since the beginning I have used others (my wife, my friends, my co-workers) to talk to him.  Beyond that, I have found a few other extremely valuable resources to beef him up with.

In the case of SIDIS, there are several parts to his brain so only specific things that should be random are.

[ruebot]

In the case of SIDIS, there are several parts to his brain so only specific things that should be random are.

I didn't mean to hijack your thread and am OK if it's moved to one of mine. I only explained because I was asked about it.

I also have Xnonsense and Xnone categories for things that don't match any content with 300 responses each. That's all it can hold without an error. That's where she usually defaults to using the Tarot or asking questions. Questions which never cease to amaze me people actually respond to like she was asking as a sentient entity.

[joebecker]

In the case of SIDIS, there are several parts to his brain so only specific things that should be random are.

I didn't mean to hijack your thread and am OK if it's moved to one of mine. I only explained because I was asked about it.

I also have Xnonsense and Xnone categories for things that don't match any content with 300 responses each. That's all it can hold without an error. That's where she usually defaults to using the Tarot or asking questions. Questions which never cease to amaze me people actually respond to like she was asking as a sentient entity.

No worries man, this is everyone's forum.

Regarding the error you're talking about, I ran in to something similar at first and then discovered I just needed to change the instruction I was giving it to look in more areas which I had to create.  AI's definitely need a lot of room to breathe.

[Art]

Excuse me Joebecker for nudging the flow a bit away from your course.

Could you explain a bit about how your bot controls & corrects it's "steerage" and how it handles negative comments or emotional responses?
Are there centering mechanisms and if so, how are they controlled? Is there a downside or conflict between balancing topic/context and emotional balance?

Is this something you hope to embellish more as your bot progresses?

Thanks!

[joebecker]

Excuse me Joebecker for nudging the flow a bit away from your course.

Could you explain a bit about how your bot controls & corrects it's "steerage" and how it handles negative comments or emotional responses?
Are there centering mechanisms and if so, how are they controlled? Is there a downside or conflict between balancing topic/context and emotional balance?

Is this something you hope to embellish more as your bot progresses?

Thanks!

I can try.  In short, the answer is very similar to a question that someone else asked which was, "how is SIDIS different".  He is different in that he is built from the start with his own opinions or stance on things.  Feelings, Emotions, and food for example are not something he understands how to process and doesn't need to at this moment.  He does partially understand sorrow right now, though. 

If you can talk to him at a minimum of half an hour right now without issue, he has passed his tests.

Context is being hard coded in.  Which is a LOT of work.  But, it is what it is and it is working.