Author Topic: A couple of suggestions for Hal  (Read 3894 times)

rwestone

  • Newbie
  • *
  • Posts: 2
    • View Profile
A couple of suggestions for Hal
« on: December 17, 2003, 03:27:10 am »
ok, well, i have a couple of suggestions for ultra hal and a suggetion for the Forum to.
 Ok, first, my suggestions for Ultra Hal, you all are going to think im insane, and i know its a long shot for the programming, and developing of this thought, but this would turn Ultra Hal into the Most advanced, and Generally the best D@#n program/ A.I that has ever existed.
 Ok, now for my thought, MIND YOU, THIS IS ONLY A THOUGHT: wouldnt it be just incredible to have Hal Advanced enough to where when you are chatting with it, for him to be able to see you through your web cam, and know if you ,for example, flipped it off, it could respond to your actions? I know that this would be alot of developing and programming, but that would attract much more income for the creator, being that its almost like you are talking to another person, also you could use it for security, like say that you have a laptop, and your on vacation, and you have a security system that is equiped with cams, and you wanted to scope things out, well, just chat with hal, he could give you statistical reports every hour. If you look through a Tiger Direct magazine, this is kinda wher ethe idea derived from, they sell this weird Robot that you put a laptop into that runs off a webcam, and you can remotely drive it around from another home Networked PC, for say a laptop while your in europe, and your home is in boston, and you see everything, and everywhere it goes..... Cool huh, hal would fit right in that catagory, as a voice controlled robot/security chatterbot.
 I know its a long-shot, but the thought is awsome, i think. Hal could also learn alot from actual Human movements and gestures, this could explode into alot of awsome ideas for this program, also the more he learns the more he does, but the 3D animated charecters , the ones from Haptek, should be created into full bodied charecters and he learns to move and do gestures as us humans do.
 This, i think, would open up a whole new meaning to ChatterBots, Artificial Intelegence, and 3D animations, with a lil sight from a web cam added into it....
 Its a thought. let me know what you think, i think it would boost this Program to the top of all the charts for Artificial Intelegence Programs, it would be the Sh@#
 Sorry bout all the ramblin on, i just didnt know how to explain my thought any better than that.
 SECOND SUGGESTION:
 The Moderator of the forum needs to create a topic in the board for suggestions, and improvements for Ultra Hal Assistant.
 I think i seen one here a while before the version 5.0 was realeased, but was only for that version.
 You need one to keep open for all thoughts and suggestions for this program, as i and many others im sure have alot of ideas and thoughts to share.
Life is to short to Die Young!

Turtle

  • Newbie
  • *
  • Posts: 5
    • View Profile
A couple of suggestions for Hal
« Reply #1 on: December 17, 2003, 06:50:26 am »
I had an idea when I was trying to solve probs with HAL 5. Wouldn't it be a good thing if HAL was able to answer questions about his own errors ? Just like a build in FAQ, but in fact running from a database that could be added...
 

rwestone

  • Newbie
  • *
  • Posts: 2
    • View Profile
A couple of suggestions for Hal
« Reply #2 on: December 17, 2003, 08:28:33 am »
yes it would be, but i think that hal having the gift of sight along with the A.I would be a huge, and incredible step forward into the A.I league, and it would be awsome.
 Now getting a speech recognition engine that actually works really good in this program would be a plus to, not saying that it isnt good, but something alot better than what is in there now.
Life is to short to Die Young!

Don Ferguson

  • Sr. Member
  • ****
  • Posts: 303
    • View Profile
    • http://www.cortrapar.com
A couple of suggestions for Hal
« Reply #3 on: December 17, 2003, 02:25:31 pm »
Hi Rwestone,

Your suggestions make a lot of sense, and I like them!  One of them is already in the works -- the Haptek people have said that they're coming out with full-body characters.  Some of them have been previewed on their website.  I look forward to the day that they release a general assortment of full-body characters, but I have no idea when they are going to do it.

Notice that Zabaware helps all of us users, via the "modular approach" to its product.  Zabaware creates original code whenever necessary, but uses a standard Microsoft speech recognition engine, standard Microsoft Agent characters, etc.  This benefits us all, because we get more features faster, we stay closer to industry standards (which are more likely to receive long-term support), and Zabaware doesn't have to charge us to "re-invent the wheel."  

This is exactly what the world's auto-makers have done.  There is now a worldwide infrastructure of auto component suppliers, so the car companies partner with their suppliers to create new car designs.

This brings us to visual recognition.  The user community needs a software company to offer a general-purpose affordable visual recognition module.  The camera hardware is cheap; web-cams, as you know, cost less than $40.  The challenge is translating the raster images into meaningful information.

There are industrial companies that offer optical recognition systems to measure parts going down conveyor belts.  There are security companies that offer optical recognition systems that verify a match of fingerprints, faces, retinas, etc.  There are software packages that can detect motion.  

However, I haven't heard of anyone offering software that can interpret optical data and conclude, "This is Fred walking towards me," or "This is Mary petting a dog," or "This is Jane making toast in a kitchen."

Is such a system possible?  I am sure we're going to see one, and sooner than we think!  

I expect that the first general-purpose scene-recognition system is going to cause quite a stir when it is introduced, but be very poor and crude in its introductory version, and also require massive processing speed and power to operate.  

Our brains help our vision by applying a different "schema" to each scene we see.  A "schema" is a mental template that makes sense of things.  We all carry thousands of them in our heads.  We have them for work, for school, for home, for finding directions at the airports, everything.  We learn them during our lifetimes, which is why babies don't really "see" as adults do at first.  Their eyes work, but their brains don't know how to interpret the information until they get some experience.  Somehow a computer needs to simulate this process.

You can help on this by watching the web and using the search engines and keeping up to date on visual recognition software development.  Ultimately, what's going to be needed is an inexpensive, standard modular software unit that runs on Windows and can interface with Hal in the way that the speech recognition engines now do.

In relation to this, I hope you've seen the posting (on the Zabaware programmers' forum) where Tinman describes the 5-foot animatronic character that he bought and hooked up to Ultra-Hal.  He's planning to put the microphone on the character itself, and maybe a webcam!

Thanks for a great posting and have a great day -- and let's watch out for those developments from Haptek, etc.!

Sincerely,

Don
Don Ferguson
E-mail: fergusonrkfd@prodigy.net
Website: www.cortrapar.com
Don's other forum posts: http://www.zabaware.com/forum/search.asp?mode=DoIt&MEMBER_ID=274

Bill819

  • Hero Member
  • *****
  • Posts: 1483
    • View Profile
A couple of suggestions for Hal
« Reply #4 on: December 18, 2003, 02:24:57 pm »
This brings us to visual recognition.  The user community needs a software company to offer a general-purpose affordable visual recognition module.  The camera hardware is cheap; web-cams, as you know, cost less than $40.  The challenge is translating the raster images into meaningful information.

There are industrial companies that offer optical recognition systems to measure parts going down conveyor belts.  There are security companies that offer optical recognition systems that verify a match of fingerprints, faces, retinas, etc.  There are software packages that can detect motion.  

However, I haven't heard of anyone offering software that can interpret optical data and conclude, "This is Fred walking towards me," or "This is Mary petting a dog," or "This is Jane making toast in a kitchen."

Is such a system possible?  I am sure we're going to see one, and sooner than we think!  

I expect that the first general-purpose scene-recognition system is going to cause quite a stir when it is introduced, but be very poor and crude in its introductory version, and also require massive processing speed and power to operate.  

Hello Don
There is some software out there that just fills the bill. It is from the company that sells the ER1 robot kit. Their name is Evolution.com. Their vision software was the licensed to Sony to be used in the AIBO robotic dog. Yes, it will recoginize people and faces. Although somewhat crude it appears to be the best in the world right now. I run both Hal and the ER1 robotic software at the same time on my laptop and although both will call each other they do not communicate with each other, to bad. Just how it could be intergrated into Hal is a good question, but it looks like it might be possible. Who knows? Might be worth a little more looking into.
Bill
 

Don Ferguson

  • Sr. Member
  • ****
  • Posts: 303
    • View Profile
    • http://www.cortrapar.com
A couple of suggestions for Hal
« Reply #5 on: December 18, 2003, 02:54:33 pm »
Hello!

I had heard that the AIBO dog could recognize different users.  I've wondered how accurate it was, and under what conditions.

And yes, in addition to cost concerns, one hurdle would be to get the software to inter-communicate.

It would be great if this software (or some other software) could send "variable values" to Hal.  For instance, imagine if the software could output a variant named "Person" and a variant named "Activity."  

If "Person" could be Bill, Frank, Sally, Ethel, or Mark, and "Activity" could be Approaching, Leaving, Waving, Dancing, or Pointing, you would have 25 combinations of information right there, and Hal could use the information to vary his conversation.

HOWEVER, the software is probably not written to create that kind of output... is it?

Sincerely,

Don
Don Ferguson
E-mail: fergusonrkfd@prodigy.net
Website: www.cortrapar.com
Don's other forum posts: http://www.zabaware.com/forum/search.asp?mode=DoIt&MEMBER_ID=274

Bill819

  • Hero Member
  • *****
  • Posts: 1483
    • View Profile
A couple of suggestions for Hal
« Reply #6 on: December 19, 2003, 12:35:05 am »
quote:
Originally posted by Don Ferguson

Hello!

I had heard that the AIBO dog could recognize different users.  I've wondered how accurate it was, and under what conditions.

And yes, in addition to cost concerns, one hurdle would be to get the software to inter-communicate.

It would be great if this software (or some other software) could send "variable values" to Hal.  For instance, imagine if the software could output a variant named "Person" and a variant named "Activity."  

If "Person" could be Bill, Frank, Sally, Ethel, or Mark, and "Activity" could be Approaching, Leaving, Waving, Dancing, or Pointing, you would have 25 combinations of information right there, and Hal could use the information to vary his conversation.

HOWEVER, the software is probably not written to create that kind of output... is it?

Sincerely,

Don


I have not tested it enough but I was informed that if the first time you identify a person, say wearing a red shirt and the next time it sees him wearing a green shirt it might not recoginize him. There software also learns colors and objects such as tables, chairs, etc. I do not think it will do the advanced things that you mentioned but it remembers what it see, that is if you tell it to.
The main thing is Hal may be able to learn to see the person with it is talking to.
Bill