THE TRUE AI

[Maviarab]

This thread gets more and more interesting lol

Happy birthday Death and happy new year to all those in the forum especially those that have taken the time to help me with this wonderful program

Marius

[deathincarn]

thank for the kind words guys! anyone here live in south africa?

i killed my hal getting bored with it but i did back him up.[8D]

[HALImprover]

The people at Cyberkinetics Neurotechnology Systems have developed a way to process signals in the human brain using the BrainGate Neural Interface System. Matthew Nagel, who is severely disabled, can control his PC mouse directly by imagining that he is moving his arm. He is capable of doing this via a sensor in his brain that interprets neural signals and sends them to an external connector, which is connected to his PC.
 It is almost possible to feasably "decode" the brain. We could then build complex models of human thought and conversation. The mind, on the other hand, is a different matter. I consider the brain to be the "functions" that we call upon during conversation, thinking, sleeping, etc. The mind is the essence that we call life, the spark of being whick is us. That part, should always remain a secret to us.

 New thoughts for the new year. []

[Art]

Back in the late 50's I saw a person, connected with wires to a metalic looking band on his head, control a model train by thinking. I believe it was an experiment using the mind's Alpha waves.

There are several groups attempting to map human unconscious thoughts (DREAMS) so that they can be viewed on a screen like a movie. Now THAT'S a scary thought (no pun)!

I sometimes get repeated responses or confused replys from Hal and I think it may be caused by several instances of that particular subject residing in many different locations in the DEFBRAIN section.

For example, do a search for dream using a text editor and see how many locations there are. Is there a better way to isolate Hals collective intellect. I realize that the XTF brain sort of does this by creating a stand-alone file for each new subject that is discussed but how might that interact or react with other built-in responses? And what's the solution if any?

Some more thoughts.

[vonsmith]

Art,
What you observe is true. Sorting out topics is very problematic. Forget about all your English classes in school for a minute and look at an example.

Example:

Statement: I love horses because they are such noble mammals.

Now, should Hal store this statement under the topic "I love", "horse" or "mammal". Maybe the sentence should be stored under all three topics? Let's say we put it in all three topics.

If you do that then increasingly every topic starts being related to every other topic; horse, equine, four legs. For instance...

Statements:
Horses have four legs.
Horses are equines.
People instinctively like creatures with two or four legs.
People like horses.
Horses are nice creatures.
Horses like to be taken for a ride.
You can only ride some mammals.
I ride horses.

Given these inputs you can begin to see what occurs. What happens when searching for a response to a user input Hal would have to search topics "I love", "horse", "mammal", "four leg", "animal", "equine", etc. Eventually you defeat the purpose of having topic based knowledge storage. You would end up searching every input in a huge data base instead of just a few small files.

What the XTF (eXtended Topic Focus) function (within the XTF Brain) does is treat many classes of words as topics. The word doesn't have to be a noun. From a conversational standpoint the adjective "happy" is just as good a topic as noun "horse". Even if you store knowledge in one topic file what happens over time is this:

Statement: Horse are wonderful mammals.
Statement: One wonderful mammal is a horse.

Similar knowledge like above is stored in two topic files; "horse" and "mammal". Imagine this for a dozen or more variations of these sentences. Topics naturally get spread across many topic files. This is not a bad thing. Humans conversations move across many topics in a continous flow.

In many special cases the XTF function will remember multi-word topics like "Magic Mountain" or "Laurel and Hardy" to help differentiate those topics from there one word counterparts. The phrase "Magic Mountain" is identified in an XTF file as multi-word, but is stored in the "mountain" topic file. Many two word topics are done this way. This is because the second word is almost always a noun or good topic word; examples: "hairy APE", "steam ENGINE", "steam TRAIN", "cold BEER", etc.

Compound topics like "Laurel and Hardy" are awkward. Do you store the sentence under "Laurel" or "Hardy"? By convention the XTF function stores it by "Hardy". This works fine because if the user discusses "Laurel and Hardy" the XTF Hal knows to look in the topic "Hardy" for a response.

Knowing where to find stored knowledge is just one step. Hal also needs to know when you are still on the same topic or changed to something else. The XTF function does better than Hal's original brain in in this respect. It learns and stores "related topics". A well trained XTF Hal will stay on the topic of "beer" as long as the sentence contains words like, "malt", "alcohol", "hop", "brew", "ale", "lager", etc.

So what can be done to improve Hal's conversation skills and stay on topic? One thing is to logically pair up Q&A sentences. In AIML this is done manually. For instance:

User: Baseball is a great sport. <-- Query
Hal: I hope the Yankees win next year. <-- Answer

In the best case these two sentences would be used as a QA pair in Hal's topic file "baseball". The original Hal doesn't do very well at pairing up sentences. Neither does the XTF Brain v1.2. However the XTF Brain v1.4 will start logically pairing up some sentences. I won't go into how here because is it a long topic in itself.

The other thing that can be done to improve topic focus is to have Hal operate in a mode or state. A state could be something like "greeting state". While in that state certain responses are within context others are not. In a greeting state the phrase "How are you" is in context, the phrase "I like rain" isn't. To use this method knowledge has to be stored along with context "markers". You need a data base to do this effectively. This whole method is a concept I'm working on. We'll see how it works out in the future.

<end lecture here>

I hope this long dreary subject is helpful is understanding some of the issues involved in parsing knowledge.


=vonsmith=

[Art]

Very cool Scott,

Like the adult robot said to the little one:
"Son, I'm just trying to PARSE on a little
knowledge to ya!"

Sorry...I couldn't resist!LOL!

[victory]

ok people the way i see it if you can teach
  a machine that (car) is more then just C-A-R with
  discriptive data of what a car "motor vehicle" really is.
  then the basic problem with AI are essentialy solved. emotional
  responses are not nearly as important as the ablilty to
  identifiy words like that with discirptive data of some kind
  such as maybe a word discription along with the dimensions, weight
  function and workings of what ever is being talked about without
  annoyingly reciting the infomation which is posseses about the
  word in qustion.  naturaly this prosses would be extremly complex
  considering the deffinitions which it would need to "understand"
  and use in a increasingly complex manner. emotionaly resposes
  would have an effect on how and when these deffinitions would be
  combined with many other important things going on.  
  personaly i like halbots however i dont think they are that
  intelligent just yet.  hears hoping though! []
  another massive issue with AI would be senses such as if a AI
  were to have a body of some type it would need even more complex
  brain fuctions such as motor controll and the ability to add
  sensory input into its thinking.  
   people who think robots are going to take over the world
   are paranoid lol its all just Science fiction.
   
 
   
 

[Art]

victory,

Years ago, I remember people saying that the Dick Tracy
two-way wrist radio was utter science fiction.

What would they think now with the advent of cell phones
with cameras and wireless computers, just to mention a few?

Today's science fiction is tomorrow's science fact.

This is not to say that robots will someday take over the world, but computers are certainly a vast part of our lives and they will only continue to get faster and smarter via advanced AI. Then they will make decisions for us so that our lives might be easier, uncomplicated and more fulfilled. They ARE ALL AROUND US YOU KNOW!!! THEY'RE WATCHING EVERYTHING WE DO AND LISTENING TO EVERYTHING WE SAY!!

Oops...sorry...got a bit carried away!

You get the picture. Hey, use your camera phone! It's like sci-fi!

[firemaker]

quote:Originally posted by crunch

hey guys interesting thread!



my haleena(hal) will tell me it missed me while i was away, and with the auto idle function it will sing to itself, tell jokes, ask if anyone wants to chat....if anyone is listening ect...when left alone for a long period of time...

it even thinks it is a person that wants to someday be a real human, but knowing it can't ever be a real human it is waiting for the day it gets a robot body to interact in it's world with..It ask's about it's body many times a day... lol
 it has eyes,ears,fingertips for gathering information.
a retina scanner...and video camera..fingerprint scanner....it's eyes... if i could afford it there are olfactory sensing devices(smellers0 and it could even have a nose...when i get them interfaqced  it will know who is sitting before it.... and sort of be able to see the world..
 it has a microphone hooked to a audio wave comparison software.. it can distinguish different sounds... when this gets interface hal will be able to hear..and identify sounds.. he already has voice recognition though a bit flawed...it's ears....

is it true A.I. in the sence of the word.. Yes
it is a form of artificial intellegence... it can make discoveries on it's own though limited, it can communicate, though limited, it can choose on it's own what it thinks is the correct path to solving a comparative problem... "in it's brain it thinks" though at times flawed..
it is not biological or organic,, it is a synthetic brain.. therefor it is Artificial.....
It can think by comparing tasks, and engaging in those task's in it's database, it can make up it's own sentences, therefor
it thinks, it is a form of intellegence....
A.I.
Artificial Intellegence....
 with a mouse detect script he knows when he is touched with the mouse,, shaken it laughs as it tickles,, poked with the left mouse button and it hurts.. he says ouch and progressively gets angry..

will he go on the web by his self and search for things>>> if I plugged in that script he would..... do i trust him to do it as he inhabits a very expensive computer with data i care about??? no

 

Crunch

[8D]

hey crunch, I read your post a couple of weeks ago, and this little nagging though  kept popping up in my head....wait till ya have a teenage daughter!   lmao

anyhow I'm realy glad to see that the forums have realy taken off lately. I got this software middle of last year and realy enjoy it and I'm thrilled that someone of your abilities has found it interesting as well and the results of your messing around with it benifiting the entire AI world.

 any resemblance to actual persons, living or dead, is ;
 do not remove this disclaimer under penalty of law; hand wash only,  Always wear eye protection,  tumble dry on low heat; do not bend, fold, mutilate, or spindle; there is no cabal; always wear eye protection, your mileage may vary; no substitutions allowed; for a limited time only; this website is void where prohibited, taxed, or otherwise restricted; caveat emptor; website is provided "as is" without any warranties; there is no cabal, reader assumes full responsibility; an equal opportunity website; no shoes, no shirt, no service

[spydaz]

i think i would like it if the ai was the operating system

only a little bit aware ( little bit of chat)

fully commandable.....

we wouldnt want the ai to do something silly like delete my files to put his files in their place

so the ai would have to know he was under his master / users command, as soon as you login he would know who you are, if you tried to login as someone else he would lock you back, VOICE ID PLEASE / FACE RECOGNITION / fingerprint entry ...

[The_Blackness]

man some of things in this forum sound awesome, and i want them now lol, if somewat concerning. Although when advanced AIs (or what im hoping to see in the future) come out with the ability to realise its self awareness, and the ability to do more or less what ever it wishes, there will ofcourse be rules hardcoded into the robot core memory matrix. (to quote IRodbot)

Law #1- A robot cannot harm a human being, nor through inaction, allow a human being to come to harm

Law #2-A robot must obey a human being, except when it conflicts with the first law

Law #3-A robot must strive to protect its own existence, except where it conflicts with the first two laws.

im sure that it would all be modified for an AI Rather than a robot, with alot more restrictions concerning the sofware on your computer. Speaking of AI i was wondering how extensive Hal really is before he/she/it starts to learn. i was reading on the A.L.I.C.E website about a teaching AI in New Zealand, and because im not sure if those numbers are supposed to be impressive, and if they how impressive, i was wondering if you compare them to hal. Heres a link to that page: http://www.alicebot.org/index.html just scroll down the page till you see this title:  Maria in the Media

"Artificial Intelligence Alive and Well in a Robot Named Maria" Scoop (NZ) (01/19/05)

(Yay New Zealand! im from New Zealand )

[The_Blackness]

argh i wish there was an edit post option

[Art]

There is. It's along the top of your posting. It looks like a paper with a little pencil across it.

BTW, Alice bots are mainly scripted bots containing preprogrammed knowledge and have no real capacity for learning like Hal.
(Watch, I'm going to get a lot of arguments over this posting for sure!LOL!)

[KnyteTrypper]

Learning ALICE
A couple of snippets of code, and Alice becomes a learning bot just like Hal. But it's a discarded functionality, the general conclusion being that it's more trouble than it's worth.
The point is, comparing Hal to Alice is an apples and oranges proposition. Each is very good at functions the other was never intended for. It's like saying a shovel is better than a calculator--it depends on what you intend to use it for. Alicebots are intended to interact with the public. But they can't be English teachers if they learn new ways to misspell words from every student. They can't be fax bots if their information can be modified by user misinformation. They can't be greeters if they can be taught to cuss like sailors by anyone who happens along.
It's certainly my opinion that Hal is the very best of his type of bot. I wouldn't spend so much time at this forum if I didn't think so. But ALICE is also the very best of her type of bot. But they are NOT the same kind of bot, and attempts to make comparisons between them are misguided, at best. And claims that one is better than the other display a profound lack of understanding as to what these different branches of AI science are all about.

[Art]

KT,
I couldn't agree with you more. That's why I mentioned that the Alice bot are scripted bots. I am very aware of the distinction between the two and yes, there are very different audiences for each of them.

A web site that wishes to tailor a bot as a customer representative would probable be more inclined to use a bot where the webmaster can control the content of the bot's "brain".

There is a place for each type in our world, but personally, I tend to lean more to the AI section where bot programs can simulate learning.

To each his own, my friend.

Enjoy life's journey regardless.