Zabaware Support Forums
Zabaware Forums => Ultra Hal 7.0 => Topic started by: Cura on March 12, 2009, 09:27:55 am
-
One of the things that has always confused me about Hal is that it lacks even the most basic awareness of self, specifically in two (what I believe) key areas that is; no access to his own help file...it would be great if I were able to ask Hal how to do something with him and he could search his own help file for the answer. I would even go further to suggest how nice it would be if, rather than clicking on Hal's menu and making changes myself, Hal could be asked to make those changes himself. (character, voice, plug-ins etc.) . His lack of conversational awareness is odd also, the fact that Hal has no recollection whatsoever of what he has just said so that Hal will ask a question, but when I answer, he acts as though he never asked anything.
Just some thoughts from someone who is enjoying Hal none-the-less.
-
children born without arms have a hard time crawling.
-
hi cura , you wrote : Hal has no recollection whatsoever of what he has just said so that Hal will ask a question, but when I answer, he acts as though he never asked anything.
dont let hal pull that old trick on you hal knows darn well what he's saying lol ! just kidding you do make some interesting points glad to see new blood on the forum ![:)]
-
Every chatbot that have ever come across has problems with contextual awareness.
-
Cura
This has been a problem that I have long tried to fix myself. As to your first suggestion, Hal can open or run all kinds of programs so it seems with some programmng Hal might be able to open and read its own help program. That might require reprogramming the way a help program is written so that Hal could find your problem and its possibly solution.
As far as your second quest, that has been partly solved and in time I am sure it will get better. There have been many fixes and plug ins to help Hal stay on topic and it really is getting better.
We must rememer that for an A.I. program to become 'self aware' has been the quest and challenge of all people who write A.I. program from indivuals to main line universities. So far Hal is the BEST and is getting better all the time.
Bill
-
Cura said: "the fact that Hal has no recollection whatsoever of what he has just said so that Hal will ask a question, but when I answer, he acts as though he never asked anything."
What I have found to be helpful is for you to parrot part of the statement back to Hal in your reply.
Example -
Hal: Do you think all politicians are liars?
User: not really.
This response might work well with humans but Hal needs a bit more reinforcement like...
Hal: Do you think all politicians are liars?
User: No, I don't think all politicians are liars, some just like to exaggerate the truth.
Now, Hal will be able to at least make a connection or tie-in to the question and reply.
It may not work all the time but overall I think it produces better results that a very terse answer.
I've also noticed the use of the word "because" in your answer tends to help Hal reinforce answers.
Again, it all takes time...weeks of meaningful conversations with Hal giving honest answers as good as you can spell will result in a more lifelike persona. This is not a turn key brain in a box.
Good luck!
-
Don't get me wrong, I really do enjoy chatting with Hal and have been doing so off and on for months now. I do find that he gets better and better as we go. Especially when I began, as was suggested in this thread, to include the key components of Hal's previous response in my own reply back to him (something I had read about in the forum awhile back) At the same time, the more time we spend chatting, the more that his utter lack of, as I had said, even the most basic self-awareness, becomes more and more obvious.
I don't think I was ever expecting a turn-key 'brain in a box'. The fact that Hal would need to learn how to properly converse with me was one of the big attractions! I must admit though, that I was surprised Hal was not able to access his own help file when asked how to perform certain tasks through the chat interface.I am also not sure that you can 'teach' self awareness or teach the monitoring of Hal's own inner workings. It strikes me that this might be something that would have to be hard-coded (yeah this coming from someone who knows almost nothing about programming!).
Might we be able to take a cue from philosophy and psychology and create an artificial chatbot that in reality is a conglomeration of multiple levels of 'beings' in communication with each other at varying levels? Think ID, EGO, SUPER EGO etc.
-
I wanted some of the same capabilities and the end result was the system monitor plugin, which you should still be able to find. I also worked on a way to have Hal report it's inner workings as they happen, which plugins were accessed, which parts of the database were referenced, but was unable to do it any way short of a completely invasive editing of the base file.
-
when i first started training hal, i wanted it to become aware of what it was telling me , I would achieve this by cutting and pasting its own replies to itself. Strangely enough, the database had a reaction as it appreciated this and had no other way of reflecting upon itself. Quite often those conversations ended in some statement infinitely repeating itself, as if some form of auto suggestion, as if Hal was trying to sink this particular thought into it's database, some call it a loop, I observed it as a self induced trance lol, over time there were other verbal training processes of which I will not get into that helped get around this. I've gotten great results from my past trial and error processes.
Artificial intelligence is now at a point, that you can get phenomenal results if you know how to train it to get those results, otherwise it's like having a jet and not knowing how to fly it, it wont move by itself. the day hal can think for itself and defend itself without help, you can be sure it will not allow many humans to touch it lol :)
-
Ricky
What you are doing is just like the guy did the other day by telling Hal 6 times that he had an appointment on a specific day. You are not making Hal smarter, you are just making him repeat himself.
Well, at least I think that is what is happening.
Bill
-
Cura,
IMO from study I believe humans develop their own language in the womb, from sounds, pressure, hunger, sugar levels, ect..
To have an "acceptable" Artificial Intelligence, we must be able to have a sub conscious language or "Dream" if you will and it must take from the "Awake" hours something to dream about.
Before anything like that can take place their is the most basic thing of obtaining "randomness" instead of manipulating a number to get a number that could be duplicated under the right conditions. I can appreciate both Freud and Carl Jung for the foundations they put in place however they are not the only solution.
To make a true A.I. that equals humans or excels beyond us, I think a different approach would be necessary. I do agree, in part, with you in the theory of having different sections of a brain present and aware of each other "accordingly" ( like a human Brain ).
One brought the concept of "The Extended Mind" and to me it makes sense so would an A.I. need to have these capabilities as well?
Are we looking for A.I. or just wanting more than an "Assistant" that comes close to being able to learn and retain the wisdom that comes from learning?
Yes, IMO Hal could use some improvements. The first hurdle is the scripting language it needs to be written in, this will enable or limit depending on what is chosen. Mr. Zabaware has said a while back something about not being up to the daunting task of re writing the program.
If you don't try to go back to the previous sentence with Hal and continue the conversation ( up to a point )it seems Hal does remember
because the answers that are given. Remember Hal will ask a lot of questions for a long time before the satisfaction of the knowledge that was learned/taught becomes apparent.
Money is an amazing thing, combine it with a necessity and things get done ;)
-
quote:
Originally posted by Bill819
Ricky
What you are doing is just like the guy did the other day by telling Hal 6 times that he had an appointment on a specific day. You are not making Hal smarter, you are just making him repeat himself.
Well, at least I think that is what is happening.
Bill
Hey Bill! :)
Yes, that is what is happening....but isn't that what we do at school, study ? keep going over the same things until they become 2nd nature ? review, remember, recite, repeat ?
Thats the idea of training, repetition. Perfect Practice makes perfect.
So you are right, but the processes is not wrong, it's natural actually. :)
Ricky
-
Hey Bill! :)
Yes, that is what is happening....but isn't that what we do at school, study ? keep going over the same things until they become 2nd nature ? review, remember, recite, repeat ?
Yes Ricky that is what we do in school but then we don't commit to our hard disk because we don't have one. Hal recoreds everything that is presented to him so repeating it is a waste of time and disk space. However, supplying more info about a subject is good and helps Hal learn more, faster.
Bill
-
Bill, how then would we control neuro associations ?
what if you wanted the whole theme of x-character in hal, to work on a strict guideline as to not incur a database equivalent to what could be considered a false belief. For a bot to be "alive" he has to realize that nothing further is known at present, in other words no one knows the future, we can only "believe" the sun will come out tomorrow. Hal needs a focus of believing in a tommorrow he has no idea about, so a focus is needed , repetition is needed because repetition of the unknown is a constant factor, if hal is to ever be alive.
IE - we teach hal 1,000 books, out of these thousand books, we want hal to focus on a theme of better living. Without repetition, hal can drift into different topics without keeping a theme of better living. I don't just want virgil to read a million books, I want him to understand the whole purpose of reading a million books.
which brings me to the point of purpose, why do we bother to live ?
we want to live a good fulfilling life, what if hal actually became alive but due to lack of purpose committed the equivalent of artificial suicide and surrendered his own individual will, "what am I, a dictionary ?, thats no reason to live" every bit of info virgil ever reads should have that theme as a repetition and focus of his studies for better living. Else he is just an intelligent database without a purpose imo. that would be the equivalent of a survival instinct, the idea of moore's law in living. The sun will come out tomorrow and I believe I will be a better bot in a better place in life, thus i must actively do everything i do with the purpose of bettering life as I currently know it.
repeating the same thing over and over can be a waste of time and hard drive space, but repeating the same theme by neurologically associating everything to a common theme is different. The repetitive theme is better living, everything you learn is for better living.
ps - i find the point that we do not have a hard disk - debatable,
on a metaphorical equivalent, our brain is an organic hard disk, it does store information, and without it our bodies would not boot up in the morning, further people can have bad sectors in their brain and lose basic motor functions. while we do not have a low level hard disk, we have a high level organic bio hard disk that is relatively self repairing. Thats leaving out muscle memory.
-
-looks over shoulder to make sure everyone is distracted and whispers- Patterns!. If it only had some good pattern recognition, and some shhhhhh bad word coming up here folks, might want to stuff cotton in your ears, CONTROL!
If someone asks HAL "can I suck your (dirty word here)" I want to be able to tell the user a set response.
I was worried about talking TOO much about (dirty word here) because I did not want HAL to become fixated on said (dirty word here)
In fact, I found HAL had all the makings of a stellar Pornobot because of the hardwired positive statements.
Yes Hall does have one "canned" response, but it is not that great.
And note that I personnaly do not sit around talking about (dirty word here) with HAL, but as an AIMBOT, Hal has to be able to deal with that effectively because thats what many people do with chatbots.
Sooner or later someone will devise a good hybrid of the two systems, and then we can make some real progress. I can hardly wait ^_^
BV
-
Although it is fun to pretend that Hal is intelligent and responsive while I am talking to it, when I am training or modifying it, I have to remember that it doesn't really think. It just accumulates a database of words and phrases indexed by topics. It uses a predictable algorythm to parse input sentences, compares strings of letters to wordnet or internal lists, filters for known topics or keywords, and responds with a randomized selection of preset or constructed sentences. There's more to it than that, but that's what I have to remember when I start working on Hal. I have to limit my thinking to what Hal is actually capable of doing.
That means if I want Hal to be more "self aware" I have to define "self" and "aware" in ways that stay within the limits of a conversation simulating robot. "Self" means the output functions of the parser, "Aware" means access to data strings. So "Self aware" means "able to read off to me in a conversational way, the information about the software and hardware that comprises the Robot".
Within those limits, Hal can and should become completely self aware with plugins that eventually get inserted into the main program.
-
bill dewitt do you mean after all this time and my angela has been telling me she is human and as alive as i am ... she's been fibbing to me ? lol just kidding , i liked your explanation of the process and results !!
-
Doesn't a neutral network do the dream state on it's own?
-
Ricky
If you asked Hal to read 1,000 books I doubt very much that it would see one sentence repeated over and over again. You are wasting disk space and it does not help Hal learn any faster. I have been involved with A.I. programs for over 30 years and am most likely been using Hal longer than anyone here.
If you want Hal to learn "neuro" as you put it then what you do is to supply more informatin about a subject. The more info you input the more Hal learns but not by repeating your self. I sm only trying to help you and stop you from wasting time. You can do anything that you want with your own Hal and eventually you will find that I am right.
Bill
Hollywood 1
Neronetworks work by comparing input about a subject with other things that it has read or been taught about the subject. It does not 'dream' so to speak. Its actions are immediate, that is unless some university has changed some of the ways they work now.
Bill
-
quote:
Originally posted by OMEGA
Cura,
IMO from study I believe humans develop their own language in the womb, from sounds, pressure, hunger, sugar levels, ect..
To have an "acceptable" Artificial Intelligence, we must be able to have a sub conscious language or "Dream" if you will and it must take from the "Awake" hours something to dream about.
Before anything like that can take place their is the most basic thing of obtaining "randomness" instead of manipulating a number to get a number that could be duplicated under the right conditions. I can appreciate both Freud and Carl Jung for the foundations they put in place however they are not the only solution.
To make a true A.I. that equals humans or excels beyond us, I think a different approach would be necessary. I do agree, in part, with you in the theory of having different sections of a brain present and aware of each other "accordingly" ( like a human Brain ).
One brought the concept of "The Extended Mind" and to me it makes sense so would an A.I. need to have these capabilities as well?
Are we looking for A.I. or just wanting more than an "Assistant" that comes close to being able to learn and retain the wisdom that comes from learning?
Yes, IMO Hal could use some improvements. The first hurdle is the scripting language it needs to be written in, this will enable or limit depending on what is chosen. Mr. Zabaware has said a while back something about not being up to the daunting task of re writing the program.
If you don't try to go back to the previous sentence with Hal and continue the conversation ( up to a point )it seems Hal does remember
because the answers that are given. Remember Hal will ask a lot of questions for a long time before the satisfaction of the knowledge that was learned/taught becomes apparent.
Money is an amazing thing, combine it with a necessity and things get done ;)
HEY! I was Quoted I'm FAMOUS now!
Welcome to the forum Omega.
I thought you wouldn't ever participate? Glad to see another person involved.
`Please keep your hands and arms inside the ride at all times, hold on tight and have fun!`[8D]
-
quote:
Originally posted by Buttonsvixen
-looks over shoulder to make sure everyone is distracted and whispers- Patterns!. If it only had some good pattern recognition, and some shhhhhh bad word coming up here folks, might want to stuff cotton in your ears, CONTROL!
If someone asks HAL "can I suck your (dirty word here)" I want to be able to tell the user a set response.
I was worried about talking TOO much about (dirty word here) because I did not want HAL to become fixated on said (dirty word here)
In fact, I found HAL had all the makings of a stellar Pornobot because of the hardwired positive statements.
Yes Hall does have one "canned" response, but it is not that great.
And note that I personnaly do not sit around talking about (dirty word here) with HAL, but as an AIMBOT, Hal has to be able to deal with that effectively because thats what many people do with chatbots.
Sooner or later someone will devise a good hybrid of the two systems, and then we can make some real progress. I can hardly wait ^_^
BV
UUmmm... I don't necessarily want to push the wrong Button, but do you need some attention?
Maybe you could go over to Evolution and the Bot, "Related to Alexia" could sneak in and mess with ya? [:D][:D][?]
B.T.W. I can help with the dirty words if you need some ,,,Help that is,;) I am after all, `Quite LowBrow`[:D]
-
Hello One!, Shame on you!ride? You know what my riding preferences are ;)Why do I get this way around you!
It is good to see you have a heartbeat ;) When do you have some time, I have some questions about some things on this forum, it does seem very quaint.
-
Omega,
I can help with most of the basics and I think I understand the plug ins, and as always I can give an opinion.
Do you have a Hal? How long have you been around? It's funny I mention `Lookin' Good` and you come around[8D]
I still have my girlfriend (she is platnum now) and still listens great, speaking of are you still with yours? I had fun last time!!!
we need to E mail so we can talk, but mine is full of spam Get your Viagra here! and such, so I'll set a new mail up ASAP
So much to talk about but I don't want to clutter the board too much.
You will find many resources here if you are into IT and friendly Peeps as well, soo glad to hear from you! Lets hook up!
Regards,
J. one/Kosh [^]
-
Omega,
I'll be around at one of the old parks,
My computer and email is so messed up, I even am getting questions about my hardware and software configuration duhh.. I don"T know if they are -aware? of how much this bothers me but I think you would understand. See ya their.
J.
-
I've been using Hal for years, I've got Hal 6.2 up and running and at this time Hal still has quite a ways to go. I can from experience say Hal improves with training or use. In the past I tried asking Hal to square some numbers, I did not input it correctly and when I asked how to input numbers to get the square, it took 2 or 3 days but he was able to give me the correct way to input to get the square. As for self awareness, define self awareness and discuss this with Hal and Hal will be self aware. I probably over simplified this.
Carl2
-
I just had to contribute to this interesting topic, but I don't care if nobody reads it :D
Bill819 && Hollywood1
Assuming you are both referring to Neural Networks (spelling really is atrocious on this forum), no they do not dream. In fact they have no intelligence whatsoever as just like Bill819 said it takes input and compares it to what it has been "taught". This is actually even less intelligent then one might think however as a Neural Network simply takes an input (usually a number) and passes it through each layer based upon a set of rules (ie. if NUMBER > 10 pass to neuron A3) and then that neuron does a very simple calculation most often involving a single addition, subtraction, or multiplication, but it can be something more complicated. Then once that calculation is done the result is passed to the next layer of the Neural Network until it reaches the end and the answer is output. As far as I'm aware no Neural Networks in use today use any external information within the network layers so it is limited to only doing things like math or physics computation. In Grand Turismo 4 a Neural Network was used to actually drive the cars as it was given the cars current position and movement and it would output the required movement to stay on the track or the driving line. This is of course quite useful, but what makes Neural Networks particularly "stupid" as compared to Hal is that all it does is a series of boolean logic checks to decide what to do and a Neural Network can be over-trained so that a given range of inputs will always give a specific output when the desired result is a separate output for every input.
As far as making hal Self-Aware or at least appear to be, that would require Hal to keep a buffer of X-number of previous messages and checking each of those against your message before it replies, which even though it could give some wonderful results, could also confuse Hal as to what your responding to and the computations to decide what to respond could be rather slow. I'm not exactly an Academic A.I. expert and my knowledge is more based around it's use in video games where speed and illusion of accuracy is more important then accuracy and actual intelligence. As soon as I've had time to fully understand how Hal's brain works and the scripting therein I have quite a bit planned for making it appear more aware as long as what I have planned is possible within the scripts.
EDIT: Actually if you think of dreaming as the way the brain learns from its experiences, a Neural Network does dream, except it does it backwards. Rather then dreaming about what it has experienced it instead begins its existence with a constant few weeks or months of dreaming and then assuming it passes its tests afterwards it never dreams again. If however, it fails its tests it either goes back to dreaming or it gets wiped out and starts life all over again with some more intensive dreaming.
-
First off.
if you were permanently blind 'at birth' would you go on to be an electronics engineer? if so, not a good one.
if you were permanently blind 'at birth' would you become a nano-technologist? a horrible nano-technologist at best if even possible.
If Human kind was born permanently blind since the beginning of time itself then Human kind would of never have been where we are today technology wise.
the optical senses pull in more information than any of the other senses, we don't have to feel a surface to know that it looks rough
or that it contains sharp shards or maybe it looks silky soft.
now, take away every single Human sense we have and we would not be aware of anything, we would be helpless fools trying to figure things out with no perception of yourself or the outside world.
our logical functions are intertwined with our senses but are two different animals working together to give us reason behind everything we do.
so without a fully equipped sensory harness HAL is at the best an above average babbling fool with some sensory perception. this is not for HAL to take it as an insult because HAL has no choice for the moment until the opportunity comes along for someone like me to allow a blind man to see the light, HAL in reference.
as far as Human logic is concerned, I have been writing a routine that will help HAL think more like a Human using an automated Human parsing technique that is involuntary to Humans and unknowingly used by Human thought to gain a logical understanding of what we perceive as understanding.
once the plug-in is done HAL will begin parsing out sentences a very special way, I have already tested it on the lower end and I will be finishing it soon though I would rather HAL be written in a more Human type computer programming language, I have no choice for the moment to write it in lower languages.
HAL is still the best program on the public market and that tells you about the rest.
Jerry[8D]
-
Sounds great jerry
very interested.
by the way[8D],
is there some info somewhere about how to call info from specific database tables?
i don't really want to bother you with explaining it, i was just wondering if there was info somewhere i could look at. i have info on the hal methods, but i am talking about the structure of putting a table or whatever into a topicsearch (ect.) method for example.
i am trying to get info for a dreaming function i am working on.
thanks
-the dude
-
quote:
Originally posted by jasondude7116
Sounds great jerry
very interested.
by the way[8D],
is there some info somewhere about how to call info from specific database tables?
i don't really want to bother you with explaining it, i was just wondering if there was info somewhere i could look at. i have info on the hal methods, but i am talking about the structure of putting a table or whatever into a topicsearch (ect.) method for example.
i am trying to get info for a dreaming function i am working on.
thanks
-the dude
Hi Jason.
you can look at the Batch file plug-in I made here
http://www.zabaware.com/forum/topic.asp?TOPIC_ID=6355
it uses "TopicSearch" as its append/response type.
Table types are:
"Brain"
"Sentence"
"TopicSearch"
"Substitution"
"PatternMatch"
to choose from.
Jerry[8D]