[spydaz]

Recently during my Researching I found

https://responsiblemachines.wordpress.com/the-platform/ (5 mins read)
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3453527/ (15-20 mins read)

Interesting reads!

I have just written a Windows service to Hold My AI - State machine. (Not Easy)

I currently considering the idea of the Artificial intelligence State machine.
Although Variables can be preserved in a VB.NET Applications by utilising "settings" - the thoughts of our human emotions and how to implement the constant changes we go through in a single day / hour ; Various internal thought and external stimulai can spark an emotion. Our reactions to Chat/Visual stimulai can change our emotions accordingly. How many emotions are there?

"Can machine think / Feel" - I suppose it must be possible. we can simulate some emotions in text... or Avatar.. the thinking part... is this done on the fly (code-wise) or while resting (windows service-wise) .... by utlizing the AI as a service it can be thinking while the program is not active then when the program is reactivated... Requests such as What have you been thinking can be responded too.... obviously Time of day can play a part in emotions as well (night time (sleepy) ....

 Hmm... What should go in the chat-bot state machine?

PS: it seems that finite state machines are the direction to go....

[infurl]

Not sure how to display images inline here, but there's an example as an attachment.

Google for "emotion wheel" and you'll find lots more though I think the content is always similar.

[infurl]

Finite state machines are not even powerful enough to model the behaviour of an insect. You need some kind of push-down automaton at the bare minimum. Have you read Rodney Brooks paper on subsumption architecture?

EDIT: Judging by recent articles about it, it seems that even subsumption architecture isn't considered to be enough to cover the basics anymore. Personally I favour a system which supports multiple internal abstract models covering the outside world, other actors, and the system itself. These are processed independently of each other. Also take a look at recent models of consciousness. FSAs do not cut it there either.

[spydaz]

I was looking at plutarch wheel of emotions;
Sometimes my brain locks things out - For some reason he was not an interesting read! - I done some Courses on coursea / Udemy about emotions and still i can see that the current philosophies are actually missing something..... things are never really complicated .... Simplicity is key.. (for the AI its hard to think about "learning emotions" - First step is to implement the basic emotions - the center and work outwards ... wheel of emotions; is a great universal starting point.

I just downloaded the Rodney Brooks paper on subsumption architecture(will read later)
I may have also discounted it before.... when i was creating the Intelligent agent architecture... which i use as a wrapper to my AI / Sensors/Actuators etc...i also created Devices as sensors (input/Output devices) IE: the avatar - Sometimes only an Output device (actuators)...... Speech recognition (input device sensor)... This model also uses the Publisher/Subscriber design Pattern.... So subscribers can Receive updates from the Agent....
The model is actually in the Artificial Intelligence: A Modern Approach (book) Mainly used in robotics but useful in chat-bot architecture...

[spydaz]

With the windows service Running threads etc... I have the emotions returning to neutral after set time.... So no emotion Holds for too long....
Abstract thought.... Perhaps i should keep a record of the last emotion used for each user.... so when the user starts their interface - it will be neutral but pending the last emotion .... that way it can remember if he likes that user!

Also would be interesting how to Catalog some Emotional behaviour types
 

[WriterOfMinds]

Thanks for the wheel, infurl.  I was looking at some emotion stuff last week or so, and none of the diagrams I ran across were that comprehensive.

[ivan.moony]

Intelligence, by my liberal definition, is ability to reconstruct entities in the world around us. We see one system, opened or closed, we investigate it to see how it works, and we build a mental map by which we can predict system states. Systems have their input and output, and may be composed be build more complex systems. Once we have a mental map of a system (when we can predict how output depends on input), we may build another system like that one, in the real world.

Input and output notion of a system reminds me of functions, where we have input variables and output expression. Even with constants, we have a constant name as an input and constant value as an output. Functions, when approached the right way, may be expressive enough to represent any system. In science, lambda calculus deals with functions, and it is a base for many functional languages. Given an analogy between systems and functions, maybe lambda calculus is what you are looking for, as a medium for storing data..

[Freddy]

For simple emotions or moods I used sentiment analysis of incoming messages. If there were words with negative associations then Jess got sad, and for positive words she cheered up. Very simple like I said. There is a running sum or score of how things have tallied up and when asked how she feels she bases her answer on that score.

[spydaz]

For simple emotions or moods I used sentiment analysis of incoming messages. If there were words with negative associations then Jess got sad, and for positive words she cheered up. Very simple like I said. There is a running sum or score of how things have tallied up and when asked how she feels she bases her answer on that score.

I have been juggling with the 2 tier model the Script which detects the text values of the emotion (the trigger) and the Holding of the emotion;

After watching all the Lisa feildman and company about EMOTIONS.... there was only 4-5 main theories (they haven't got anywhere), yet an interesting comment was that regardless of the emotion the Measurement of the bodily seems to be the same, the label for the emotion is subjective. we place labels and facial expressions on emotions. The facial expression serves as a way to show emotion to others(learned) yet the body reacts. they made claims that the face can change your emotion (reverse lookup)  ie if you smile it makes you feel happy.... therefore they cant prove which one is affected first. but obviously if i come behind you and scare you your body reacts first...
With AI.... the emotion is received  from the environment from sensors(text, Face recognition, Reactive skin) then the body needs to Hold the emotion. then the body needs to Actuate the emotion (Avatar Expressions, Word choice, Behaviours).....

the body actually hold emotions for a short time probably for some random time (Bio-Rhythms) you cant Stay angry for example.... Hence the windows service i'm creating Holds the Emotion (for some time before returning to Neutral(some thread)).... detecting the emotion with the Words(currently) , If i can figure out how to do a cheep dirty face recognition in VB to recognise expressions then i would also use this to trigger "Symathetic responses" for the AGI/ASI

obviously ASI would be way above the need for emotions, whereas AGI needs to React like a human

"HUMAN MODEL: based on a human as a model, thinking and acting like a human as well as conversing as a human with a natural language interface".(Dyer 2016)

[Freddy]

Interesting and yes I really should do that levelling to neutral too - thanks for that

If i can figure out how to do a cheep dirty face recognition in VB

Can OpenCV be used with VB ? I never tried it...

[spydaz]

Interesting and yes I really should do that levelling to neutral too - thanks for that

If i can figure out how to do a cheep dirty face recognition in VB

Can OpenCV be used with VB ? I never tried it...

It does i was using it before with a library in vb6 but its broke now....