Virtual Humans Forum
Virtual Humans Forum
Home | Profile | Register | Active Topics | Members | Search | FAQ
Username:
Password:
Save Password
Forgot your Password?

Virtual Humans
Partner website:
Chatbots.org
 All Forums
 Virtual Humans
 Business Applications
 Custom, Adaptable, and Evolvable Bots
 New Topic  Topic Locked
 Printer Friendly
Author Previous Topic Topic   

hologenicman
Moderator



USA
3323 Posts

Posted - Jan 19 2006 :  22:06:11  Show Profile  Visit hologenicman's Homepage
This is an offshoot of a thread posted by GrantNZ at:

http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=87

I am proposing that a bot with a core of static code scripted with sufficient functionallity could have its configurations modified similar to "natural selection" or "selective breeding" according to the principals of survival of the fittest. Varying configuration settings for the static code control program flow and present bots with varying "traits".

These criteria for survival can be automated or evaluation based so long as the bots containing the desirable "traits" continue and contribute traits to persuant bots and those with undesireable traits go away with attrition.

Commercially viable bots for public use could be evolved based on automated or manual rating systems with input provided by the general public.

Commercially viable bots for personal use could start out as a variety of traits and evolve according to the individual's needs as their own personal "Buddy-Bot".

This whole concept is housed in the premise that the bots are contained within the environment of a 3d-Game setting. You can think of it as the Sims with brains. The evolution can take place as removal of old bots and adding of new bots or it can be handled within a single bot with undesirable traits fading away and desirable traits becoming more prevelant.

I relate the game setting with multiple bots fading and presenting to the coming and going of friends over the years as our needs and interests change. We choose to spend time with those that we are comfortable with and spend less time with those that don't match our situations. These parameters change over time, and likewise, the bots could continue to evolve(change) over time.

Survival can be based on how often we choose a particular bot over the other bots. The bot with which we interact the most scores higher on the survival rating, just like the TV rating system. Survival can also be based on manual input since we could interview the bot and fill out a quick rating chart which will influence its survival rating directly.

Once bots have been "evolved" for specialty settings such as Public Relations, Teaching, Newscasting, Healthcare, Etc, they can be sold as pre-evolved for those settings and services. The true applicabillity for this is to "evolve" those professional bots within the "culture" that they will be utilized since there can be a real culture gap from location to location.

I may have missed something, but you can go to the above thread to read the prior posts for yourself.

This is not an instant product, but rather a proposition for guidelines and goals for us to aim toward with our projects. I believe that a commercially viable direction has been identified and that it will help us put our projects to work when the time comes.

John L>
IA|AI

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Jan 19 2006 22:12:38

GrantNZ
Dedicated Member



New Zealand
2677 Posts

Posted - Jan 20 2006 :  08:07:38  Show Profile
quote:
I am proposing that a bot with a core of static code scripted with sufficient functionallity could have its configurations modified similar to "natural selection" or "selective breeding" according to the principals of survival of the fittest.

Yep, it can. KAOS will.

Note that survival of the fittest is subjective. KAOS will not consider user happiness as his main trait-survival criteria - KAOS will consider his own happiness paramount. This is important to me in enabling KAOS' own personality, rather than a submissive "people-pleaser."

Not that there's anything wrong with a "make the user happy" bot!

At first, the process will be quite simple - if KAOS likes the user, sharing happiness will reinforce that behavior, but if KAOS hates the user then insulting the user will generate personal happiness and reinforce that behaviour. Once KAOS understands relationships, the system will get quirkier. Once KAOS tries to profile the User's personality, the system will get downright chaotic!

I've considered another survival criterion - conversation length. It may not be possible to do this with current NLP technology, but the idea would be to identify as many things as possible that would prolong the conversation with the particular user. For example, greeting the user especially cheerfully may add 30 minutes to the overall conversation length. Talking about chess before 10pm could tie the user up for a while more, but after 10pm may be counterproductive. Threatening to commit suicide if the user wants to leave may also extend things a little. Hal could end up more manipulative than your standard female (A group of friends I used to spend time with had a game - we tried to keep somebody with us for as long as possible after they said "goodbye." In various ways, we could quite often keep somebody chatting for 90 minutes past their original "see ya" - it became quite a specialty and lots of fun - even to the "victim," who'd suddenly find himself centre of attention at exactly the wrong time! We never resorted to threats (other than fun ones... ) or emotional manipulation (other than ego-boosting ). Simple trick for anyone interested - try complimenting someone just before they're just about to leave. Do another compliment when they try again. It can be quite amusing to watch them struggle with their desire to hear how wonderful they are, when they also want to be somewhere else... note this is not something I do in everyday life, it's just a psychological observation )

Philosophical point - is it more economically feasible to design a bot to be good at a particular task (e.g. customer services), or to breed one that way?
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Jan 20 2006 :  18:47:39  Show Profile  Visit hologenicman's Homepage
quote:
Philosophical point - is it more economically feasible to design a bot to be good at a particular task (e.g. customer services), or to breed one that way?

I believe the answer would be fully dependent upon the underlying technology employed it that particular implementation.
Art makes a good point that today's technology doesn't make evolution easy, but I'm a firm believer that if nobody "actively" works on tomorrow's technology, tomorrow may never come...

John L>
IA|AI

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Jan 20 2006 18:51:11
Go to Top of Page

scratch
Curious Member



USA
32 Posts

Posted - Jan 20 2006 :  19:14:56  Show Profile
Grant, your "Good Bye" game has me laughing! I think you missed your calling, you should be a clinical psychologist You have a creative mind: in another thread you described the background image of an amusement park or carnival changing according to the mood of the bot-I thought that was very cinematic (and why are deserted amusement parks so spooky?)

Some thoughts about evolving:
I'm no biologist, so please correct me if wrong. I believe "natural" selection is a result of a diverse gene pool (created by random genetic mutation) being acted upon by a varying environment to favor reproductive success in the best adapted groups.
It seems it would be no problem to take that random mutation approach and process millions of generations in a very short time. The problem is in the selection. If satisfactory conversation is the critera, that can't really be evaluated in milliseconds!
Without the speed, the random approach isn't as useful. I think the game environment is a good way of keeping a user interested in a population of bots long enough to exert some "artificial" selection, rating the bots, etc. As Grant mentioned, some users might talk as often to the bots who are strongly antagonistic as to the ones who are friendly!
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Jan 20 2006 :  20:25:25  Show Profile  Visit hologenicman's Homepage
quote:
(and why are deserted amusement parks so spooky?)


They present a "familiar" environment that has been altered. Our minds don't consider it a new environment, but it doesn't quite fit the pattern either...

Apparently, we have "themes" that cause fear like "big thing camming at me" can be a rhinoserous or a Cadilac, but they both inspire fear. Perhaps the "similar but slightly different" theme describes the amusement park and the twilight zone...

The most freaked out that I've ever been at a haunted house was when this female dressed in white flowing garbs slowly walked toward me without saying a thing. There was no expression on her face, she made no scary sounds, she didn't move her hands, anything. She just slowly walked toward me without pausing or wavering. I was so freaked out that I picked up my date and set her between me and the woman and took off! I might have been a big stron guy, but that really frightened me. Needless to say that I took quite a bit of harrassment from my date and her sister about how wimpy I was, but I didn't care; I was out of that haunted house and away from that freaky woman...

quote:
It seems it would be no problem to take that random mutation approach and process millions of generations in a very short time.

We could also encourage the process to follow a bit of a standardized approach. Kinda like a search pattern for combinations and that would introduce a systematic means for choosing gene combinations.

quote:
The problem is in the selection. If satisfactory conversation is the critera, that can't really be evaluated in milliseconds!
Without the speed, the random approach isn't as useful.

Yeah, we would have to reduce those millions down to a few dozen. Keep in mind that the "millions" of combinations are what is required to create from scratch. We are going to be providing the "static code" already and that removes the need for millions of trial combinations to get to that point. I thing that some readers are thinking of Genetic algorythms that create the code itself and that would be quite prohibitive at the moment. I am merely suggesting that the evolution be applied to the configuration settings that control the static code that we create.

To clarify, I am only talking about evolving the configuration settings and not the program code itself.

To give a simplistic but real example, the static coded Hal can use any of several characters and configurations. Lets pick some options:

-Male/Female
-Young/Old
-Blond/Brunette/Redhead
-Jiggle/Non-Jiggle
-Black/White/Asian

Jim can correct me on the math, but I think this equals 2x2x3x2x3 possible configurations.(72)

We have Hal systematically choose configurations based on these choices and log how long the human interacts with that configuration before changing to a different configuration. Give each configuration a name so that the human can recall the configurations that they preffered to interact with. This is truly a "superficial" trait, but male users may prefer the jiggle version and female users may prefer the non-jiggle version. Some humans will prefer older and others will prefer younger.

These choices can be made directly by the human, but I am using this as a simple example of configuration choices that can be automated based on setting survival criteria such as human interaction duration and frequency. This is the self customizing bot approach that I am trying to describe. It is not future technology based. It is just an approach to automating and customizing the configuration process so that the user will be provided with a bot that will match their needs and desires.

John L>
IA|AI

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Jan 20 2006 22:23:40
Go to Top of Page

art
Moderator



USA
374 Posts

Posted - Jan 21 2006 :  01:15:26  Show Profile
Funny stuff Grant,

People also respond well to talking about a particular in which
they either have a vested interest or a good deal of knowledge.

Case in point:
When I was in school (we won't discuss how long ago that was), whenever one of our teachers was about to give us a test, one
of us would ask him something about his favorite subject.

He would start talking about it. The more he talked, the more
attentive the class became...focusing all our attention on his
every word...asking more leading questions in an effort to firmly
set our hook in his mouth not unlike a big fish.

It actually went on for several weeks until he finally got wise
to our game and at that point he no longer swallowed the hook!

It was fun while it lasted, but what you said is true. People
love to feel important, knowledgable, loved and they love to talk
about it.

Good observation, Grant!

- Art -
In the world of AI it's the thought that counts!
Go to Top of Page

laackejim
Committed Member



USA
3274 Posts

Posted - Jan 21 2006 :  06:22:08  Show Profile
John, You got the math.

I also think you are right about the evolution (genetic code) process. evolution of existing organisms is single (or rarely more) gene mutations at at time. Much more than that tends to make a non-viable code. large changes occur incrementally over time (from our perspective even the catastrophic model of evolution--big jumps, not little steps--require thousands of years to accumulate impact).

For our Vspecies (phylum?) it would probably be inappropriate to apply the biological selection system too tightly. The evolutionary pressures would be applied in an entirely different way and survival has a totally different definition. In general, organisms have a schedule, a tight leash in a way, of needs to be met or they do not survive. There are quite a few that have found a way to button up for long periods (centuries in the case of some species of brush) and perhaps longer in the case of some primitive crustacions. Our Vperson has the characteristic of being alive (potentially viable) during long periods of non-existance (shut down). This alone introduces a step-around of selective pressure and a thing to keep in mind when we conceive of the propagation of "Venotypes".

At some point we should seriously consider what survival criteria WE would prefer and build them in deeply and strongly. At some point I raised an idea in this forum, sort of a bug in my head. It goes something like this. Imagine a well designed Vperson infected by a computer virus. Consider that causing havoc or damage or whatever may be the intention of the human creator of the virus, but to be successful the virus must first be equipped with the intent and the means to -- guess what -- Survive. Insert that imperative into a well designed Vperson and (maybe) off it goes as something new.

Summary, we should seriously consider the possible ways survival criteria would be expressed in the environment we intend and reinforce the good, buffer against the bad.

Today's technology IS likely the limit at the moment. And if you are running into the problem of limits already then pretty soon you/us should begin to seriously consider building a structure that will accomodate the stuff (re: lots of old postings).

I just read a posting from Anthony Carson that suggests that in his new work the "new" has been developed to incorporate beyond us. But in spite of the fact that my amazement at what he has accomplished knows almost no bounds (I refer to him as the time lord Dr. Who --BBC sci fi series --because he saves the universe regularily and does it with a sense of humor, albeit it British Humor) I still think (re: the inspite of a mile and half back there) building a skull of sufficient capacity to harbor the brain you are building is an important task.

I agree, if the criteria is conversation, generations of evolution will take time. Relatively speaking Lots of time. But I am not sure that conversation is the only, or even the strongest, pressure.

The game format (hologen) is quite possibly the arena. Those that survive, full time or as visitors, will form a range of characteristics (picture a 3D bell curve, for each user). The distribution of characteristics will not match, but will overlap, with those of other users, inturn defining a distribution pattern of viable Vpeople. From that distribution comes the primary (most common) survival characteristics and those that lie at the extremes (but are viable). Those at the extremes are the source of the mutation of the population in response to changing environments.

Jim

Uncle Jim (e=mc2)
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Jan 21 2006 :  08:12:48  Show Profile  Visit hologenicman's Homepage
quote:
The game format (hologen) is quite possibly the arena. Those that survive, full time or as visitors, will form a range of characteristics (picture a 3D bell curve, for each user). The distribution of characteristics will not match, but will overlap, with those of other users, inturn defining a distribution pattern of viable Vpeople. From that distribution comes the primary (most common) survival characteristics and those that lie at the extremes (but are viable). Those at the extremes are the source of the mutation of the population in response to changing environments.


I like it. Very well designed approach, and it allows for keeping fringe characteristics as "Biodiversity/Technodiversity" for surfacing in times of differing environmental "pressures".

John L>
IA|AI

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Jan 21 2006 08:14:27
Go to Top of Page
  Previous Topic Topic   
 New Topic  Topic Locked
 Printer Friendly
Jump To:
Virtual Humans Forum © V.R.Consulting Go To Top Of Page
This page was generated in 0.23 seconds. Snitz Forums 2000