Dear Yusuf,

I would like to build an autonomous agents which can interact with me without any needs to press or control other device. It would be better if the agent be able to detect and understand my moods and feelings, and act based on those analysis. So, there is a little physic interaction between agent and human, or even none physic interaction, but still the agent could act accordingly with response to human stimulus.

I would prefer using a linguistic interface which resembles the way humans speak which is not always using "precise" terms but "vague" yet meaningful generalizations. This may be brought about using a fuzzy logic based system. Based on some of my research, I feel that the closer we get to human interaction, the more robust, adaptable and effective we get.

Dear Yusuf,

I would like to build an autonomous agents which can interact with me without any needs to press or control other device. It would be better if the agent be able to detect and understand my moods and feelings, and act based on those analysis. So, there is a little physic interaction between agent and human, or even none physic interaction, but still the agent could act accordingly with response to human stimulus.

In my opinion agents' goals should be parts of the language for the interaction human-agent. I mean that human may have the opportunity to delegate some needs in form of goals the agent try to achieve. I am supporting this aspect in my recent papers: if interested, here is an introductory description of the approach:

https://www.researchgate.net/publication/236969537_GoalSPEC_a_Goal_Specification_Language_supporting_Adativity_and_Evolution?ev=prf_pub

Conference Paper GoalSPEC: a Goal Specification Language supporting Adativity...

Make sure any human user will be made aware this is a robot, not a human agent: for ethical reasons as well as for maintaining user understanding.

All great points above, eg voice rec and face-rec (expressions, and gesture-rec) of the human user would be great.

And all just my 2c, below...

I think - apart from it speaking English (just because: that's easier for me),

(1) it would be good to make its thought processes `transparent' so we can `see' its `thought processes'. Have an extra "brain scan" window - that shows (or - summarizes) the chain of logic (or fuzzy logic) it used to arrive at a response.

It freaks me out, not to know the (dumb) mistakes a computer agent is making.

(I want to see not just the Wizard but also: The Man Behind The Curtain...)

Also (2) give it emotions (fake/simulate this, if you have to) see: `The Expression of the Emotions in Animals', by Darwin.

Give it an avatar with `display-emotions & gestures & facial expressions' so that I get the feeling it's not just rational but emotional too. (Less estrangement, but risks more uncanny-valley.)

(3) Maybe also (on top of this) use colours like (I dunno) a mood ring - or a mood-t-shirt ?? to show its mood/aura as well. For its current emotions/attitude, I mean. (I don't mean new-age hippy stuff. Then again maybe I do. I haven't thought that through. Maybe life would be much easier, if we could all see each others auras. Red = angry, stay away, green = I'm feeling friendly / etc etc)

(4) Maybe don't make it human, make it a cute cartoon dog-avatar or something.

(5) What if, it also communicated in geometry - or say, symbols, or maths as well as in language AT THE SAME TIME. Sometimes, some concepts are better communicated in symbols - or maths - or pictograms or images than just language alone, as language is such an abstraction - and is so ambiguous (I'm a creative writer :) eg Think of how much a shot in a film (or a photo) conveys as opposed to a page in a novel

I've played with some agents like HAL and Tahni and Pandorabots and whatnot, and they were cool - but: so buggy. They take so long to train and are still pretty dumb.

I suppose - like Tahni you could make it customizable visually, and now and then it could even hit on you. Add some sexual attraction to the mix.

Also it should be *very* clear, when: it is lying.

eg Grows a long Pinocchio-nose or something, when it's telling you a whopper.

And should also be clear when it is `unsure'/feeling out of its depth/just guessing, Give it some body-language too.

(6) Give it a sense of humour.

eg: http://aimlhumour.blogspot.com.au/

But - let me pick (from a list) the type of humour, for that day. (eg witty, or sarcastic, or Woody Allen, or Oscar Wilde, or Louis CK, or Groucho, etc etc)

...All these features should be pretty quick& easy to implement, right?

LOL ;)

Cheers

JT

I thank you all of you for your useful contributions. How do you suppose agents can show:

affection

understanding of utterances

make understandable utterances

use linguistic prosody like hmm, nodding