WHAT is AI system ?????

Hi all,

This is going to be hard to explain.... What is AI system mean to me and Aiko's BRAINS software.

What is AI(Artificial Intelligence)? Many people think that AI is the ability of a mechanical object, such as a robot, to converse with a human being. I think this definition is incomplete. The meaning of AI should include the capacity for the robot to have senses similar to humans. AI should include the ability of speech, hearing, conversation, vision, object recognition along with interaction with objects and surroundings such as a sense of danger or caution in unfamiliar areas.

Currently, Aiko can have simple conversations with another human being. When I say simple, she can converse at a level similar to a 5 years old.
Aiko already has facial recognition capabilities and can distinguish different human expressions.
Aiko can identify human body movements and trace motion to detect what we are doing.
Aiko can sense danger and be mindful of her surroundings (People hurting her).
Aiko can scan her surroundings and can recognize what object(s) is/are most important such as going to a bank and understanding that the points of interest are the money, safe boxes etc.

For instance, if you put a chair in front of Aiko, her AI can recognize that the chair is the foreground and decide to disregard any information from the background.
If a person happens to stand beside the chair, Aiko’s AI will change focus to make the person the point of interest. The chair will be treated like the background and ignored.
If two people stand beside the chair, Aiko will be aware of the second person, but remain focused on the first person. If the second person engages in a conversation with Aiko, then the AI is intuitive enough to focus primarily on the speaking person.

As you can see from the latest BRAINS software upgrade, there are 3 vision systems.
The main vision system (main left) recognizes the person’s face and facial expressions. It also identifies 3D objects. The system tracing human motion (top right) identifies any movement in light blue. Unfortunately it is difficult to see here as I inadvertently wore blue, but if examined closely, there is a light blue area around my right hand as I wave it and the right side of my body which moves with my hand. The third system (bottom right) detects the point of interest. Notice that the AI ignores the chair and anything else in the background. The AI focuses on the most important “object” in the room, in this case, me.

As of today, the BRAINS software can not be updated or upgrade any further. This is simply a hardware limitation as the CPU load is currently maxed out at 99%. The only way that I can continue development of the Aiko’s BRAINS software is to upgrade the CPU to quad core2 or core7i which currently is beyond my budget.
So unfortunately until I can secure proper funding, I will be unable to advance and expand the BRAINS software for the near future.

My point is a true AI should contain all human traits not just conversation. The BRAINS software is not even close to human AI, but it's a start of something wonderful.

I believe that true AI should include all human traits and senses as well as the ability to have conversations. The current BRAINS software is no where close to the human AI that I envision, but I am determined to someday reach that goal.

It saddens me that an update will not be capable in the near future, but I am happy that the BRAINS software is getting better and better! Keep it up, and I can't wait to hear of another update.

So if this correct your going to ponder about how AIKO can cross-subsidize ''vision'' object features in front of her not only distinguishing that there is a chair in her field of vision in a space reminds that discard it and focus on the main objects rather its moving around that space or is displaced also distinguishing (objects from persons )and discarding unimportant objective and still knowing the locations.

Its probably weird for me to use a financial strategic term like cross-subsidization for a.i . For example because the vision system recognize multiple objects the problem is which faces or objects are important. If there are 3 faces which face should be favored more to be track how and why.

By cross-subsidization the most favorable face should have a competive network potential. But the vision program could have states as you said if aiko is poked around by stranger she must distinguish that if you do it its funny (sometimes) but if its some stranger and he/she does it to much politely objects what she does.

And here it comes your face has high potential any other much lower they are recognised but have no high competitive level but if her states changes those other faces grow higher in potential meaning she could strengthen the competitive position to the state where it matters most. namely in the state that face/object is important the most. reading that you added more meaning at the vision system. If the faces have no competitive strength then the system disregards them again and sticks with the favorable one still with the potential to regress competitive potential. You might think what if the faces get the same potential as me, They cant because of the state of annoyance they came from. (just an theory of my own)
I'm curious if aiko will solve that aspect.

Hi

My friend help me to write the AI system in proper English..... hopefully, reading it again will make sense.

Hi there can this help I've focused more on the future ability aiko could perform
without the emphasis to all of the things A.i could tackle. enjoy

Building an android is constructing a lively machine based on natural insights,
from the real world such as movement , awareness of sound and sight.
If we discard the general sense and ad a.i as a machine ability to manipulate data.
The manipulation is derived of several components which interact with the real
world through software parameters. When both the manipulation of data and
robotic sense are combined. The build forms an intelligent entity which can sense
in real time the surrounding space with artificial manipulators and data processes to
recognize , identifies objects which it could examine.

The capability an android is build for can posses a combination of directive goals
and functionality. Greater surrounding manipulating factors such as moving around
in a room without bumping versus recognition objects/person and abbreviate labels
what its for or what it does. Combinatorial actions is the imperative action an
android must develop, such as when a person Say's ''I want a glass of water''.
This is telling the android what it must do which it has to distinguish and figure out
of or from other directed messages how to reach that goal and successful
completing that.

Manipulating data and action process from the operating framework of the android
is an endeavoring challenge. Based on complex interaction between software
components and electronic steering of actuators. The system does not only
interacts with the real world but also with itself to perform on an optimal condition
which is beneficial for operating stability and guarantee efficiency in performance.

The B.R.I.A.N.S software development should bring in the near future better
interaction between android and users. Better upgrades such as a chatting mode
with personal characteristic behavior (anime , VR or Gothic) even talking about
things like the local environment or news through services. Improved ability such as
security , home domotica and Home care.

By choosing the most important personal asset the Aiko project closely try to
approximate the point of interest for future usage. Meaning more expressiveness with
users and improved actions potentials commendatory all possible ability's.

He there this is just a thought but is it possible to add an "WikiPage" option on this forum,
So we can all work together on some of our own definition of certain subjects.
Like explaining what A.i means, Works like this a person can add a little insight about the subject.
which you guy's can moderate if or which is more plausible to add to the wikipage and by the end
have the definite page drafted up, Or something like that it looks like a hassle but its an interresting
thought maybe.

Hi

I don't have a problem with wikipage for Aiko and related AI stuffs....
Who want to volunteer.... or moderate it...

Defining what C.I is must be the world hardest lemma ever. I cant say which direction you could go with this,
definition so in quoting Bill Gates 'A.I is to broad to define'.

But there have to be some definition right?. And reading all the post on this forum has in some way defined what C.I means Based on Project Aiko. If your a regular or newbie u must have read the majority of post about
something. Going from tall androids like Aiko and HRP-4c model robot from japan. Also the so loved smaller
version Sumomo android ('Ahem my fave') as well other small types available.

Means for Project Aiko is building a dream android and that is started by acting and start building it. Such as
Mr Le Trung has done and somehow starting a new revolution off building android types different then we know.
It all start out somewhere and software thinkering provided the means of the B.R.A.I.N.S software and later some of Aiko body parts the rest is history.

This has also let to well lets say matter of discussion about robots , android to human physical interaction as well about Humanity , Integrity , psychology and such. Which make this very complex but not at all important.
Why we are not there yet to confront problems that have not been studied or concured.

Android or robot are not on the physical plane yet completely to discuss the ability of their Computed Intelligence.

To start a new definition for C.I on this forum would be solely based on android like Aiko or Sumomo like types.

This is the longest post I've made here or anywhere its based on stuff I'm working on with my own view on things.

Working to make a compelling story into how one could build an android and the a.i with it. Artificial intelligence is not
only about androids or robots but also to confront the increasing complexity technology has the ways we could simplify
control with the abstract method a.i could be removing that layer and still know how to work it.

Please give feedback if this to long or poorly understandable.

Constructing an android system with natural interactions can also mean.
Computing information that depends as symbol , sense ,sound and sight data.
From mechanical frameworks with the ability to manipulate data for a specific purpose.
The machine manipulate several components through which it interact with the real
world through software parameters and hardware senses and movements.

When both the manipulation of data and robotics sense the environment,
The pattern creating awareness like data intelligence which in real time add within
exploration and investigation space at artificial data manipulators and data processes
identity's on environment, objects which it could examine.

The capability an android is build for could posses some combination of directive goals
and functionality. Greater surrounding manipulating factors such as moving around
seamlessly in a room (thus without bumping). Versus recognizing objects/person and abbreviate
labels what its for or what it does to create knowledge.

Combinatorial actions are the imperative action an android must develop in completing
directives. for example ; when a person speaks to an android and says ''I want a glass of water''.
This is telling the android what it must do which it has to distinguish and figure out.
Same go's when a person spoke to the android earlier an directed messages, the how
to reach that goal and successfully completing that from what was said before Its like
story telling the android framework going through the action.

Another way of assessment could be manipulating the knowledge data and action process from
the operating framework to let the android decide validity reasoning ''is this still worth interacting as of
now should I ask to work on current query'' is an endeavoring challenge. Based on complex interaction
between software components and electronic steering of actuators. The system should
not only interacts with the real world but also with itself to perform on an optimal condition
which is beneficial for data operating stability and guarantee efficiency in performance.

The broad conception computed intelligent software could bring in the near future
means improved interaction with better equipment and personal characteristic behavior.
As for having the local environment status or information through services.

In summary if we also look at project aiko.

By choosing or selecting important personal asset the Aiko project closely try to
approximate the point of interest for future usage. Meaning more expressiveness with
users and improved actions potentials commendatory to all possible ability's.
there is never a one solution getting an answer to notice some of what you want to achieve.

So here a question if anybody could why did dent they. Figure an advanced interacting basically intelligence
on a common computer. Software making advanced stages with application for almost anything.
A.i or computed intelligence as we call it is mediating rules of intelligence for certain automation where non
human interaction is needed. The key for new data is mediating data with rule system to build rules coming to
this sort of features think neural nets to expert and semantics one of many possible way's which intertwine
together brings fort some property's to look at another angle.

#1 because for some reason the definition or term (a.i) is not correct calling it a whole system, the
meaning for a.i or computed intelligence became finding to recognize and distinct certain visual/speech
pattern and work a logic around encapsulating it. Many application are build around this feature building fixed
component for a specific reasoning method.

#2 The case should be a basic statistic and analysis towards recognition behavior and control logic.
To be as much broad in a specific reasoning structure as well building towards various solution capable within the framework.
As what might be capable through ontological reasoning methods that formed the a.i as software concept.

#3 The main line addresses A.I fixed structure now as closed draft for computed reasoning.
and it should be ''Open reasoning" for any a.i system free-scaled to get patterns from what kind of service available.
Presumable a.i should be capable enough to analyses and get the statistic of data and subtracting meaning from it
apart from what "fixed limitation" might give which slows (con.ci.ontology).
Meaning not just speech or vision only but anything that could work as a service.

#4 Lets look closer to what open reasoning could mean, Practical recognition software is capable of quickly
analyzing data input received, doing something with it and sending back data as response some event gets started.
the limitation of this practical recognition application turns into "single services" with its own build logic.
Okay nothing wrong with it works. I qeustion what if some control environment was the whole system of the a.i parts
anything the control environment received gets an pattern abbreviate to some service(#3). Recognizing faces or voices
as an example would be a task for the control environment which is capable handling the resulting mediation response/action.

#5 Could such control environment mediating the A.i network handle quick enough like a real-time response.
For each input has to be subdivided and correlated to each method or service structure to close a labeled pattern as new data.
this could mean the (practical service) must send responsive data to all parties. Real-time system resource must then cope with synchronizing the updating data, process all services. In all the process might take longer to learn sub-data (*10). However with the current processing power such as heterogeneous processing this should not be a problem. But such is limited on smaller devices which leave the neural net with sparse computing and updating could be costly in real-time and synchronization.

#6a There are many stable a.i application that suffer the limitation of single practical applications, better yet
these application work fast in single mode after all the goal to overcome computing intelligence limitation is getting these apps practical. But it can get all together in a global interface server environment. Meaning defining a free-scaled regulatory system that separates most of the process having hierarchical order in such a way that data reasoning becomes real-time.

#6b The many application the limitation could be overcome if we had a system overcoming these limitation by adding processing power to extend functionality to the software. Also if these application could be written to support object
environment. Data can be combined and cross referenced building up more meaning to the data un-ordered and a.i
reasoning is real-time.

#7 The current field of technology and software is sufficient enough for building towards such an endeavor. Computing
speed and memory has become available as is the bigger storage for data on small devices. But that was not necessarily the case, the submission goal was the "knowledge" the application network build-up as way of instructions. Services and net connection provide
data to almost anything which can be intelligently ordered as for being used to develop more better secure knowledge bases.
Thanks of the development through the years and to come make possible story telling and visualizing them in totally new way's we could ever imagine.

#8 Whenever the industries or academics talk about computed intelligence , smart a.i or anything towards computed intelligence.
they mean the practical intelligence what else. The status quo has to better the problems faced to get a.i working.
working properly while avoiding data complexity or mechanical about having security and system interaction/interactive response, and knowing which system fits a problem best to build towards. Anything capable by software simply means selecting specific features or having things done within a system designed by means such as measuring the problem self like analyses of the reasoning problem.

#9 About an android A.i must be up a challenge it is high performance computing on a compact scale where input/output signals interact widely through services of the android framework. A system that interacts with the environment as well with it self is a strong multi-task. It is a niche design separating knowledge with automation . Where for robots its a simple deduction to halt its automated process due through conflicts , errors or unaccountable obstructions. An android has to figure out conflicts itself.

#10 Some of the control issue stated at (#5) May not be so lamented as there are way's overcoming those with optimization. It is a
challenge after all.

#11 Open reasoning is the way to go making sense out of data already available with getting semantic, heuristic etc to do them.

#12 Modern operating system provide a wide list of possibility to this project . Linux, Mac Os, W7, open source o/s anything supporting heterogeneous selection of processing capability.

#13 The next thing small mobile development using gps, sensor hardware acceleration services.

#14 Software androids some of the works probable.

This short overview article is (as if) and prone to changes overtime.

It is easy to use terms which look so plausible that one does not look much further into. And true we don't need to do such, that much but sometimes the need of a proper definition of this word relating to the situation we want to put needs a clear explanation.

System a word we use to define a set of property's that act on them self or adapt to a process that collaborates between elements , as where it a nature to do so. Scientifically accepting the way nature does things.

We also discovered nature has its own law to do stuff naturally and strange enough if we look closer to what nature does (With nature I mean the natural way things interact with each other just as plants grows using water ,soil and sunlight these elements are needed give the plants its richness).

There are way's we could look to these natural laws there is a way a biologist does it saying ''Well its a plant it needs soil , water and sunlight'' and expands thoughts based on what the biologist observed.

Or how a scientist does it saying ''Well earth exist of ordered elements (periodic table) where each of these create a reaction which emits energy on a small scale even unlikely to detect , magnetic field etc etc".

Mathematicians would also have a view on the matter defining the property's of the laws in numbers finding complex relations would build up a theory where the numbers state the fine behavior and hypothesis.

The same go's for machine learning, the word learning as the subject referred at machine has an extensive term to the real and complex situations put at it. Using the framework of a machine as the tool kit of negotiating strategies which commonly on this field is to making a pattern based on the behavioral research we define this with. Thus making strategies a machine can use to apply matching patterns to a similar situations which it can define the same property's with by data analyses and statistic of these property's.

Strangely we take the word learning on computers for granted meaning we make an equal association with how we learn. Much of what is applied to A.I comes from research on how the human brain works. A complex biological structure which does things with ease.
Neurology one of the research fields is used also just like math based constructions statistic such as Bayanese networks.
There is even the theory of improbability like Schrödinger cat to explain weird anomaly's that are likely hard to code in A.I.

Machine learning go's a long way starting with Lady Lovelace , Charles Babbage , Allen Turing to Von Neumann , Allen Kay and all those contributing a part of their knowledge on this research. Its not only to explain how our brain works but all things around us to gain greater insight in understanding complex structures.

And off course the entertainment factor portions off robotic learning comes with having fun and building something that interacts and broaden our experience.

Peace .