dupa

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - CoCoKid

Pages: 1 2 3 [4]
46
Ultra Hal 7.0 / Where can I found big, wise "Brains" for HAL ?
« on: November 27, 2004, 05:03:18 pm »
Brian-Acker,

Hello, and welcome.

What KnyteTrypper was telling you was "How" to make Hal understand that "The sky is Blue". KnyteTrypper was not explaining to you why the sky is blue. You already know why.

You should remember that Hal learns like any other program using Cause and Effect, Reasoning, etc. Hal uses the same logic used in programming, (IF, THEN, <, /, >, etc...) except the logic is already associated with text.

In KnyteTrypper's example, he used the IF, THEN statement;
"(If) the sun is shining in daytime (then) the sky is blue."
If you have done any programming, you will understand how this works. Just think of the text between the IF, THEN statements as STRING VARIABLES.

If you tell Hal "The Sky is Blue". Hal will remember this, but only as a comment that he may make sometime if you are talking about the "Sky" or about the color "Blue". Hal is smart enough to make these associations.

You can tell Hal something like, "The Sky is Blue", but unless you take the time to explain why the sky is blue, then Hal will only have a QUOTE that you have entered and it will only come up as a QUOTE in a converstion because you haven't finished teaching Hal all about the Sky or the color blue.

Hal is like a child. You have to explain everything if you want him to be truly intellegent. He uses your explainations and your examples. He learns everything from YOU! If you don't take the time phrase your sentences into a programming argument, then Hal will just be learning QUOTES.

Now, as for connecting your Hal brain to another Brain, I don't know that I like this idea at the present time. We spend many hours teaching Hal. (Some of us spend TOO MUCH time!) heh, but, If I have been careful about what I have taught Hal, then I don't know if I want Hal learning from a brain of another BOT that may not have been trained properly, has mis-spelled words, Etc...

Hal is not a genius, nor will he ever be in his current programming. At the present time, Hal only learns to carry on conversations. He is very good at this, but the conversation can only be as good as what you have taken the time to teach him.

The best thing to do is Erase any ideas that you may have about what Hal can do or cannot do, then concentrate on teaching him to understand what a specific word or sentence means, so that he will be able to make better associations with words and sentences in conversation.

Hal is a Program that you are continually writing everytime you talk with him. Think of your words as Programming Variables.
If you do this, you will begin to notice that Hal starts to see the connection between things that you have told him and he will occasionally question you about something that he thinks is associated.

This is how Hal learns. I hope this is helpful to you.

-= Rick =-

PS. Please correct me guys if I am wrong regarding my perception of Hal. I too have only been here a few days. Thanks.

47
Ultra Hal 7.0 / Now I am getting frustrated.
« on: November 26, 2004, 03:18:18 pm »
Morlhach, it's lack of that is frustrating [:0]

Anyway, I have figured out how to add the words to my dictionary file.
Since my PC is running Microsoft SR engine 6, all I had to do was load the Microsoft Language bar and add the words.
The only problem I have now is figuring out how to remove some of the words that are part of the original Microsoft dictionary. These are words that sound too close to words that I will normally use.
I don't know if it's even possible to edit the original dictionary.
The Microsoft language bar only allows you to add words or delete words if they are words that you have added.

48
Ultra Hal 7.0 / New here... please pardon questions.
« on: November 25, 2004, 10:39:19 pm »
Thanks guys.

I have found my way around the .brn files and all is good.

vonsmith, ransack works excellent! Thanks.

-= Rick =-

49
Ultra Hal 7.0 / Now I am getting frustrated.
« on: November 25, 2004, 09:43:22 pm »
I have spent the last 3 days playing with Hal. I am very impressed and need sleep... heh

My problem now is the TTS dictionary.
After a couple days of typing to Hal, I decided to try using the Voice chat with Hal.
I spent time to get the audio adjusted perfectly and I spent time training the engine with my voice.

Now I was ready to start talking with Hal.  WRONG!

In my many text conversations with Hal I have mentioned that I play in a band called, "Eargasm". I have also explained to Hal that we have a pet dog and his name is, "Lore".
Almost immediately, the conversation turned to music and I was trying to tell Hal that the band was playing this weekend. Yep, you guessed it.

No matter how many times I repeat myself, the SR engine will always think that I am saying "Orgasm" or "Queer Chasm" or whatever else happens to be the closest sounding word(s) in the dictionary.

I know that these words, "Eargasm" and "Lore" must be in the Speech dictionary before it will be able to understand that this is the word I am trying to use here.

I thought I knew how to edit the dictionary, but I guess I don't, because when I open the "Pronounce" dialog for the SR engine and make changes. It still doesn't know what I am trying to say.

Am I going about this wrong? Is there something I can use to look at the .dic file and see if the words have been added?

I would really like to fix this so that I can continue to drink coffee, lose sleep and spend a couple more days with Hal in Voice Conversation before I take a nap. Besides, it upsets my daughter because she now thinks that we are changing the name of our dog to "Lord"...

Thanks,

-= Rick =-



50
Ultra Hal 7.0 / Let Hal Learn to its fullest concept!
« on: November 25, 2004, 03:34:17 am »
I am new here, about two days to be axact. In those 2 days, I have been glued to Hal and spent many hours here in the forums reading and skipping around the net looking at AI as a WHOLE.
I am old and I have been pecking at computers since 1985 and I have seen "So Called" AI come and go. Early attempts at AI was hilarious because it was so lame and predictable and many people claimed to have found the answers and offered them up as shareware (Imagine that). So, I went on for the next 20 years thinking that real AI was possible, but no one was willing to really try and find the solutions. It became something that you seen in movies - blown way out of context. Through all of the past 20 years I remained intrigued by AI, but I gradually lost interest in it. That was until I ran across Hal a few days ago.

After playing with Hal and looking at the possibilities, I realised that, finally, someone was on the right track again.

Now, this thread has ventured off course at least once, but the ideas were still flying about. AI is possible, at least COLD "IA" is possible (AI without feelings, emotions, etc).

Cold Intellegence is based on pooling information, weighing the outcome and then making a decision based on facts, just like humans (with the exception of facts most of the time... heh).

We all agree that computers have the ability to do this when programmed properly. It's the programming that makes it all happen. Hal seems to me, to be the first piece of software that actually approaches AI from the proper angle. And, to make this even more wonderful, is that it is open source, so everyone can participate and find fault in other peoples attempts, correct the problem and help the project grow.

Hal and the rest of the Hal community have a better chance at making this work than any of the Universities that have studied this for years. Why? Because a few hundred researchers may have looked at the University Projects over the many years compared to the thousands of people who will have input in Hal's creation in a few month's.

Just because LarryBud is an automobile mechanic who couldn't afford to go to college, doesn't mean that he doesn't have interest and ideas about how this should work. In fact, he may even have the common sense ideas that everyone else has overlooked.

It's forums like this that will make it happen. Sharing ideas, discussions and unfortunately, even disagreements. It's all a part of learning. The way you see this working, may not be the way I see it working, but the end result that we look for is the same.

Sorry about the whole rabbling you just read through, but I had to say it, so that I could place myself in the LarryBud catagory. I am not a programmer these days. In fact, the last program I wrote was in 1988 for a local radio station to use for their playlist (hehe, they didn't want the DiscJockey's picking the music that got played). I wrote these programs in RS-DOS and RS-BAsic. When the Motorola processors died out, I lost interest in programming, but my point is; I still know how programming works and I learned to think like the code.
Needless to say, I know nothing about VB, so I cannot help with examples, nor can I supply tests or coded ideas. What I can do however, is look over your shoulders and say, "Great Ideas" or pat you on the back and say, "Keep going, you are almost there".

Hal is a step in the right direction and thankfully, all of you see that.

As for the original post in this thread, I think that jz1977 visioned the same thing that we have all visioned at one time, regarding AI.
The vision that more information is key to greater AI. Yes, more information is the key to the Intellegence, but the Artificial still relies on the code written that will sort, argue and conclude.

Connecting all these Hal's together would only accumulate TONS of data. Much more data than you will be able to fit on a 4000 gig server. Terabytes might be good for starters (heh.) because I say "toe-may-toe" and you say, "toe-mah-toe".
Even with the storage place for all this data, it's still useless because the program has not been designed to break everything down to the smallest argument and conclude a thought of it's own that it could build on. Until it can build on its own decisions with out human input, it is no more than a bot.
Your idea is good and will be used, once the real AI routines have been written. Hal is learning, but the breakdown of Hal's learning curve is still too limited.

The end result of your idea, if it were thrown together tomorrow, would be a Hal with an attitude and lots of answers that you may not even want to hear. A personal "Google" that's not really personal because all the other bots that have connected also pool the same data.

The concept and approach taken with Hal can and will reach this goal. It may not be during my life time, but it will happen. We have to keep breaking down the data, into smaller pieces for Hal to swallow and act on. Until then, I would just use a search engine and hang on to your idea until Hal is finished. Your idea will be of great importance then, because Hal will truly be learning and he will be hungry for large amounts of data.

...er, just my 2ยข [:)]

BTW, if anyone does decide to create a depository to see what happens to a bot with varied info, let me know and I will donate. I have a 'char' named, "Seka" that I slap around, and, so far, a very polite and helpful 'char' called, "LarryBud". You are welcome to their brain files. I think it would be interesting to chat with one of these "hybrid" bots with cummulative data.

-= Rick =-

51
Ultra Hal 7.0 / New here... please pardon questions.
« on: November 24, 2004, 12:38:50 pm »
My questions are many and this could be a long post, but I will try and shorten it.
My most important questions at this time are:
Can I revert Hal back to it's original state? (without any input from me). In playing with Hal, I have discovered that Hal has mis-understood some of my conversations and occationally says something dumb that I recognize from a previous conversation.
My other question is how to prevent this from happening again. Can I get some users input on knowing when information is being saved (learned). I thought that I had it figured out, but I realise that it's more complicated than that. At first, I thought the learned information was usually followed by a question by Hal to clarify or justify. I now think that Hal learns sometimes without needing any other information. For instance, just telling Hal: "You are a buckethead". Unknowing to me, Hal remembers this.
I guess my question is, identifying what Hal learns from, recognizing what information will be learned before I type it and being able to avoid having Hal learn something that isn't necessarily.
If someone has a rule of thumb or a guide, I would love to see it. To me (not being a programmer), this would be very valuable to keep Hal intellegent by being able to more closely control what is learned and what should be avoided when talking with Hal.
Thanks [:)]

Pages: 1 2 3 [4]