Aiko Learning Program.

I just recently learned of this project. In the video I watched and the article I was shown it was mentioned that Aiko has a learning program. I was curious how that worked. If you tell Aiko a fact will she be able to repeat it only in response to a specific sentence or can she respond to a sentence phrased any number of ways but still requiring the same answer?

I suppose to further this question: If for example you have her set to clean a room in your house, would she be able to recognize the objects in the room and place them where they belong? If the larger things such as desks and tables where not in the correct place would her program adapt? Would she be able to place a new object that does not normally belong in that room or is not normally in the room in an appropriate place?

Aiko can learn 3-D object in a room, however each 3d takes upto 18meg....
Aiko only have 2Gid memory.... it will fill up pretty fast.... that is the down fall of the system due to home budget...

I hope that help to answer your questions...

Their were two parts to Kachada questions. First how well can the Aiko understand context? Can it understand related phrases when the primary object is not identified with the same keyword? Do all commands need to be the exactly correct phrase to be understood? Is Aiko conversational AI based on the AIML language or some other natural language processing system?
BTW how do you search for Japanese chatbots? What are the correct keywords in Japanese? Are there numerous Chii chan Japanese chat bots under development?
For the visual processing, is Aiko able to adapt to known objects seen from an unusual perspective? Say a chair not just in a different position in the room, but say it was lying on the floor on it's side. Would it still be recognized as a chair?

Aiko looks for Keyword when talking to her.

Aiko what is this object
what is this object.

Aiko will understand both.
English and Japanese use the same in the brains software.

Aiko will doesn't recognize large object like chair, there is no reason for Aiko to learn them now, because I can't not walk.

It sounds very complicated, I know that she can identify a number of objects,
Is this your right to tell her Or use the software to identify the objects automatic?

Hi

You have to teach the software new objects. It is a lot easier to teach smaller objects than larger object like desk, chair...etc

I have an interesting question to ask then. Can she recognize right from left. Such as Aiko, is the ball in my right hand or my left hand?" or "Aiko, which hand is holding the pencil?"

i learned about using this recently, maybe this will help. you can use two internet wireless cards to connect two computers together for data transfer. a large computer or external storage media device sits somewhere else, and she transfers data to/from it. as she can't walk, maybe a wireless 1tb disc backup could be hidden below her chair? i believe some university is using some massive room sized computer for a robot that will develop and learn like a child. i believe the Smithsonian magazine mentioned that. would this variation on a theme help her intelligence grow?

Discoman wrote:i learned about using this recently, maybe this will help. you can use two internet wireless cards to connect two computers together for data transfer. a large computer or external storage media device sits somewhere else, and she transfers data to/from it. as she can't walk, maybe a wireless 1tb disc backup could be hidden below her chair? i believe some university is using some massive room sized computer for a robot that will develop and learn like a child. i believe the Smithsonian magazine mentioned that. would this variation on a theme help her intelligence grow?

That's a pretty good idea. My Nikons can talk to hard drives through a camera mounted wifi device, so I can't see why something similar couldn't be applied to Aiko. A redundant system for back up operations would be a pretty easy yet effective fail-safe.

i got the idea from a boom on building heavy duty outdoor ROV robots, and the author mentioned using standard computer wi-fi connectors. maybe a connector on the robot, and the main brain computer has a router and internet connection?

I beleive part of Le's original design specs if not already in use include Aiko's ability to surf the net and collect incoming emails.

I had similar thoughts regarding the wireless connection, but using as much as 10tB instead of just 1...

Le had replied, but it looks like he brought up a restored database as a lot of replies and psots are now missing stating he has wifi connectivity ability.

Hi

Don't know why my post got delete it. But anyway, Aiko has Wifi wirless N.
But the signal is very weak for some reason. Not sure because it is inside the silicone or interferance with the other electronic inside.

try using a MIMO wireless N antenna and matching router (they will work better if you use the card and router both from same company) if that doesnt help, boost the card to full output power and use an EM field monitor with and without her skin on to see if it may be the skin insulating against radio communications.

I did see that post. Not being familiar with the molecular structure of silicone in this fashion, I'm not sure whether it can shield a signal to that extent.

hmmm... it is worth a try. or it may be that wifi cards only broadcast apprx 60 foot radius circle? try putting a moving directed antenna on the base computer, preferably a high gain antenna. try the steerable cantenna for some ideas.

Shashinka Ichiban wrote:I have an interesting question to ask then. Can she recognize right from left. Such as Aiko, is the ball in my right hand or my left hand?" or "Aiko, which hand is holding the pencil?"

I suppose a compass could be built in the software, but I'm more interested in recognizing material.
Aiko already knows which liquids appear to be the recognized liquid, but what about grain and other foods?
Would she know wheat from sawdust, tapioca pudding from caviar, sugar from flower, or white rice from fried rice?

Artilects rule! wrote:

Shashinka Ichiban wrote:I have an interesting question to ask then. Can she recognize right from left. Such as Aiko, is the ball in my right hand or my left hand?" or "Aiko, which hand is holding the pencil?"

I suppose a compass could be built in the software, but I'm more interested in recognizing material.
Aiko already knows which liquids appear to be the recognized liquid, but what about grain and other foods?
Would she know wheat from sawdust, tapioca pudding from caviar, sugar from flower, or white rice from fried rice?

those would be useful. is it planned to have her cook? maybe something like those food and wine robots (although a bit better, the food one thought a reporter was prosciutto)