Artificial Intelligence Project

Discussion in 'Intelligence & Machines' started by Baal Zebul, Jun 18, 2004.

Thread Status:
Not open for further replies.
  1. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    Jiri, i am never keen on sharing information. Especially not on a open forum.

    But, our AI.

    We have parameters, that is all we start our AI with. It does not need the ability to walk, it will be started without the ability to kick the ball in the soccer game.
    The parameters are Mental/Physical values. A mental value is Empirical, telling the ALF when something has been proved.
    A physical value can be Gravity or wind.

    The object within the visual stream is stored as models and textures (or possibly just images).
    However, the visual stream is merely a subject for the imagination method and all it needs to imagine something is to use its parameters.
    Imagine a blue ball, 15 Kg in weight, A radius of 10 cm for instance but with this method above it would be autonomous.
    It would be able to "remember" the spin of the ball, the direction of the ball.
    It will "forget" unnecesary data. The trees in the background might be unnecesary in some cases. So with its grouping method and its relation system it would check if the trees might be of any use or if it can be removed.
     
  2. Google AdSense Guest Advertisement



    to hide all adverts.
  3. leda Registered Senior Member

    Messages:
    77
    Hello Baal. I'm a computational linguistics. You have made a fatal error. We are in fact all born with an innate ability for speech. The grammar we each use is mostly inbuilt, and only slightly learned. There are a number of switches. By being exposed to a particular language, these switches are thrown into certain configurations (verb following/veb preceeding as a simplified example). This, combined with our INNATE grammar, the grammar you as a new born baby came ready equipped with gives us our rules. Exposure to language gives us our vocabulary and sets the switches, but it's not the whole story.
    It's a very interesting field, and one that cannot be reduced to your simplistic approach. I suggest that you read some books on Chomsky or Head Driven Phrase Structure Grammar. Also an amount of research into how syntax relates to semantics may be useful, as well as a look at some of the philosophical issues of meaning representation.
    How can you solve this problem when you don't even know what the problem is?
     
  4. Google AdSense Guest Advertisement



    to hide all adverts.
  5. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    leda, "switches" ?
    You can call it switches but i prefer to call it parameters.

    A parameter is a default value (inaccurate) or a threshold (most likely visual).

    Humans are probably also initiated with parameters but hey, they are not started with the ability to speak. They have the ability to learn how to speak. Just as my AI.
    What many others (to be fair) think is that if they start a politican with enourmous information and fancy speech then it will be truely intelligent.

    Some seem to think that by moving the pre-programmed code to a database makes it intelligent.

    leda, when you were born could you ask "What is that i am seeing over there?". You could just have said "I am as stupid as my inventor and therefore i have to ask what it is i am seeing" or "To err is human, that is what my inventor did! Mind if i ask you a question? But you have to tell the truth cause i cannot filter away the lies."
    I created my Ai in my own image. If the AI "professionals" cannot understand how their own brain works then something is terrible wrong. I have even moved beyond that and started analysing how other humans, birds, cats, dogs brains work. Therefore :

    Jiri, mind if i make you immortal?
    I have a new feature, mind it call it UJB ("Unlike Jiri Behaviour".)
    The feature makes it possible for the ALF to "read between the lines" (giving it the ability of assumption (if logical of course) )
    I am not trying to be mean, i am the worst ever when it comes to comming up with names. But since i came up with it whilst talking to you i see it as fair that you get some credit.
    It does not need any new parameters or senses to function, it was already included in the design but it was not connected. Mind if i name it after you?
     
  6. Google AdSense Guest Advertisement



    to hide all adverts.
  7. G71 AI Coder Registered Senior Member

    Messages:
    163
    I hope you are not talking about the same AI here.

    Names used in the dev-related documents should make as much sense as possible for new members of the dev team. Names used in the user-level documentation should make as much sense as possible for new users. The other names are not important. Use whatever names you like. I don't really care. I care about practical AI algorithms and about ways how to organize related data. Especially about the unique ones. Unfortunately you do not seem to know what that really means in the real world + you cannot answer questions which I consider important (whatever the reason is). So, for now, I just wish you good luck with your non-fascist projects.
     
  8. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    You know that i try to make as little sense as possible to those outside the team (etc, you).

    Fascist AI, it will be logical, it would see just see values. Fascism to me.
    My image, it does not care about anyone, it process data just as i do.

    But i will change the UJB term.
    Any suggestions?


    btw, i was avoiding some of your questions because they were so inaccurate (especially when you were talking about DB)
     
  9. G71 AI Coder Registered Senior Member

    Messages:
    163
    Leda,
    Does the UG theory (or any other theory/article you know) include detailed description of the underlying universal language model which could be effectively used when searching for shortcomings of a new artificial language (which is being designed for the User-AI communication)?
     
  10. G71 AI Coder Registered Senior Member

    Messages:
    163
    Baal,

    Do not bother sending the TTT. I already talked with the author. It's far from being that clever. Many can beat it easily. It's one of his very first computer programs. Baal, we started to talk in 2003 and the more details we discuss, the more obvious your confusion is. Even a developer from your "own" "team" confirms that you are pretty much confused about the complexity of various concepts. I recommend you to stay away from big claims. You need to spend more time studying the field and the technology.
     
  11. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    I already told you, Johans tic-tac-toe player was not clever so i decided that we should skip TTT since it would be hard for us to test the intelligence level because we could not even beat his.

    We are named SC & W, "S" does not stand for Strömbom and there is no "J" in it either.
    Unlike you Jiri, he understand the concept, just not how to convert it into code.
     
  12. G71 AI Coder Registered Senior Member

    Messages:
    163
    Baal, some of your statements are better to be ignored. But let me tell you this: The fact that some of your friends "understand the concept" the same way as you do does not mean that it's a valid AI/AGI concept *AND* the fact that someone has a valid theoretical concept for AI/AGI does not mean that it can be developed using the current (or the near future) technology. At some point, an experienced guy may tell you something like "yeah, that seems to be a valid AGI concept!". But it's useless unless you can code it + keep in mind that many ideas which initially appeared to be promissing failed badly. Many pages on the Internet are filled with "nice looking" but practically useless high level ideas. One of the key differences between you and serious AI developers is that you are designing important parts of the system using black boxes (without a good ideas how to make it work IRL). Serious development is about using very real components. If you do not have relatively clear ideas how to turn your black boxes into real components (+ good algorithms on the other levels all the way up) then you cannot claim you really solved the hard AI problem. Stay away from your shallow strategies of trying to confuse people. The AI community members need to be honest with each other. We have enough real problems to focus on.

    You could evaluate it relatively. There are competitions of all kinds of artificial players and you could let various versions of your players to play against each other (and measure various parameters) in order to get some ideas about the progress. BTW do you think that the authors of the Deep Blue system could beat it when playing a fair game? I guess you want all your AI systems to become better than humans, don't you? So how exactly are you going to measure the intelligence level of the other project? If you cannot give me relatively clear answers then do not bother to answer at all. I know, you think I cannot read between your lines but I see it differently. I think the problem is that you do not have a good understanding of concepts related to your "between lines" stuff. The way how you discuss your AI project on various places (including this one) is kind of chaotic. If you want to discuss it then try to create a good documentation. Break the issues down, provide links and try to move forward systematically (like Dennis did). BTW I do not think Dennis presented a good AI idea, but he is trying to move forward very systematically + he is playing with very real components. That's what counts.

    I noticed your attempts to learn some stuff in the other threads. Check the http://www.howstuffworks.com. You may find some answers there.

    BTW I just got an update on CAIO. Unfortunately, the CAIO AI language development does not go very well. I cannot go into details because I signed related NDA but, basically, the authors did hit some unexpected and very critical problems. The CAIO development is currently not moving forward. The July deadline is lost and the future of the project is uncertain.
     
    Last edited: Jul 9, 2004
  13. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    You speak to the student when i have 2 senior programmers in the team (plus vincent, but i cannot count with him for a while)

    Jiri, i started programming when i was 7, All i see is code.

    When me and my ex-team made 3d games, we changed the game type and story all the time. Most got bored and it contributed to us falling apart, but we developed a technology for terrains that was before its time. We could have huge terrains, it is just now adays that Battlefield has reached our size on terrains. (Novalogic's Polyvox terrains are larger but that is only because they load the same terrain over and over again)

    Jiri, i know how you see the problem. You are a scientist, you reach your results from trial and error. (Structured trial and error)

    I on the other hand do not believe that something should be tested till it is perfected.
    You think that trial and error should perfect it.
    "Your brain cannot simulate a whole world" (or something like that you told me once)
    Well, i guess we will have to see about that.

    And please, i already told you in the mail, i did not attempt to learn, i intended to test a theory. Etc, i asked about Ice. They told me nothing more than i already knew. But that answer is not enough to me.

    I have no NDA and they did not get the project.
     
  14. G71 AI Coder Registered Senior Member

    Messages:
    163
    Baal,
    Why don't you invite the most experienced professional from your team here? That might get more interesting than talking with school boys like you or Johan. You keep mentioning those years ;-). What you should realize is that there are many others who have many more years of experience and much more significant things on their resumes and they still have hard time to solve some of the very challenging AI problems (and please do not talk about you being better because of your "thinking out of box" IF you cannot go into related details). Some of the AI legends are still living and working on those hard problems. Having some senior developers and x years of coding experience (with games) is good but it does not mean as much as you seem to think. I'm not saying that you cannot succeed. I'm just saying that what you are talking about is not the key to the success. I'm not gonna comment your notes about "trial-error" and “perfection” even though there would be things to clarify. Just design a prototype which can be practically coded, discuss the ideas if needed, share all the unique ideas if you can and if you have any and try to move forward with the practical development. Make sure that ALL the key components can be coded using the available tools. Dive into the design related details. The details are often what makes various things doable/not-doable. Spending time with too general AI discussions and personality related exchanges is not really gonna move you forward.

    You can likely find the theory related details on the howstuffworks.com. Many things are explained on that server so that even amateurs can get relatively good ideas about the concepts.
     
    Last edited: Jul 10, 2004
  15. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    listen, i am at vaccation now. I cannot answer your post ATM.

    1 and a half week or so and ill be back.
     
  16. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    managed to get some time to reply...

    Jiri, there was a time when you had "limited" time, you seldom replied the day after, guess that has changed because now you hurry to reply every day. I liked it when you did not have any time because it fitted my schedule too. But now, well. I am not going to discuss with you any more. It is up to you if you want to believe in me or not, unforunatly we live in a free world.
    Tell me when you feel that you can recreate Vermer, Goethe, Wagner's brains.
    heck, even tell me when you can recreate a dogs brain.
    Tell me when you are closing in on recreating Einstein.
     
  17. kmguru Staff Member

    Messages:
    11,757
    Gremlins in Sciforums? I got an email notification on 7/24/2004 8:43 AM that says:

    Baal Zebul has just replied to a thread you have subscribed to entitled - Artificial Intelligence Project - in the Intelligence & Machines forum of sciforums.com.

    I came here to reply but found the post missing....
     
  18. G71 AI Coder Registered Senior Member

    Messages:
    163
    Baal,
    When I want to respond here (or work on a particular private project), I simply do it when it fits my schedule. I thought the others here do the same thing. AI development is relatively important to me, but it's not my top priority, which is one of the reasons why I do not know when I'm gonna be done (whatever the "done" means when it comes to AI). It's more fun than race to me and I do not go for "recognition" as you do. I have:

    - some practical AI ideas which seem to be promising and which can really be coded using the currently available tools
    - decent dev skills
    - several variants of some parts of my AI DB design
    - draft design for some of the key interfaces
    - relatively small amount of code written
    - very limited time for the AI R&D

    There is still some work on the theory evaluation to do. A significant part of the initial coding is not very interesting from the AI perspective but it's still pretty much design specific so it's worth to spend some additional time with getting the theory more stable before writing lots of code. Also, there is a big difference between making some complex algorithms just working vs making it working + very dynamic/flexible (open for the future extentions/improvements/changes/script_configurable etc). Well done initial design can save a huge amount of the future (next versions) dev time. AI is one of MANY things I'm working_on (/interested_in) and, as I mentioned, not the top priority. You are, on the other hand, saying things like <i>"I love my AI more than i love life itself"</i> and <i>"I have a new </i>[AI]<i> concept </i>[for]<i> human replica </i>[doable on]<i> normal PC"</i> etc + you are (reportedly) working on it for some time + you are talking about being experienced team lead with decent coding experience and having some senior developers in your team.. So I guess it makes sense to ask how far did you get this year with your AI. Anything to download/test? Any unique/practical ideas to share? Any design document online? Brief project plan/schedule online? Where are the screenshots we talked about? What can the current version of your AI system do? And it would be great if you can keep your posts here as technical as possible. We are not just some AI comics readers. You are talking with researchers and developers here.
     
  19. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    kmguru, i posted some new calculations on how much processing power would be needed to re-create a dogs intelligence and how much for a human's. rough estimations were that 0.5 GHz would be neccesary for the dog and 10.4 GHz for the human and that was if everything ran at the same time, which of course is impossible. That complex materials don't exist in this world. But as i wrote, the human brain was too hard to calculate to actually back up.

    Jiri, have to go away on vaccation again for 10 days or less. don't have time for everything ATM.
    School will not get you anywhere, that i have known for long. So of course i have to do something else.
    In the begining i was more interested in posting documents and i often liked to make diagrams on how the brain evolved. Now i think about it all the time so when i am preparing to code something i come up with something new. That slows me down, but gives me a better concept.
    I think i have 1.0 complete now after one year.
    no time, ill post more when i get back. Plus my AI design will be complete, i might send you a copy. Just have to hear if the others are pro that. the C and the W in SC & W.
     
  20. Stryder Keeper of "good" ideas. Valued Senior Member

    Messages:
    13,101
    Leda,
    I know this input is a bit late but you mention that children have preformed switches to learn language from a young age, however it's not like that. It's been proven with "Wild children" children that have been outcast by their parents and left to grow up in the wild (woodlands etc). Such children have proven to lose their ability to comprehend language, this is because during the Interphase of new braincells, the cells have to be stimulated to generate a communications array with other cells. Without stimulation the array doesn't form and it causes the child to potentially not speak.

    The stimulation occurs from the Ears that need to be able to hear, the reverbiration of the ear drum then signals the sound to the developing Temporal lobe which is also the preportion of the brain that controls vocal patterning.

    Therefore the conclusion is that a child has to "Overhear" parents and other people talk to build up the ability to talk themselves or otherwise potentially become "Mute" for lack of communication.
     
  21. NanoTec Registered Senior Member

    Messages:
    30
    As a usual latecomer I cannot respond to all that has been written,


    Attempts to gain insight or information from nothing are inherently random.

    Predicting behavior or guiding actions based on previous data and past events does show promise. The difficulty remains defining the scope of the data required for better accuracy.

    The trick when developing a learning algorithm is to have a large set of data with a known set of true results that can be obtained from it. Absolute truths.

    If you are modeling it after a human being; it being mentioned as an example:
    It begins not with a blank slate but with a genetic template,
    It makes one creature sentient and another unintelligent.
    It doesn’t matter if one considers DNA: software or data, the two are interchangeable.
    On top of this are years of development (see Stryderunknown)
    The body learns to interact with the environment.
    Truth expands from (warmth, hunger) to (touch, taste)
    The environment for a human being is naturally an exponentially large set of data.
    Only later do these things become related to abstract concepts.

    Forever it must be said “, but it’s not the whole story.”

    I can assure you that in your quest to gain help, and that is why you began this;
    providing information almost always to your benefit.
    There seem to be many people here who are genuinely interested in what you have to say, you only need to give them an opportunity to comment constructively on your project. All you need is some step-by-step analysis of the programs decision-making process –examples of input and output. It need not be graphical just polished to ease understanding and genuine.
    To imitate my work is the greatest flattery I can receive.

    I was told once all conquests came from three factors Gold, Glory, or God
    If your after the Gold, its in the training of the AI not the software
    The Glory, its in promoting your idea and giving the credit
    God, (or morality) learning about yourself and humanity

    Now I ask, respectively
    Are you serious about the bow and arrow AI? …if its what you need…



    “What color is the blue ball?”
    The computers task seems to be to outsmart the questioner
    Is there being a question asked: (? and what)
    What is being asked? (is)
    What is a “color”? (nonverbal concept)
    of what? (the “adj” ball)
    is “adj” a property of “color”? (then it is part of the answer)
    No amount of guessing will produce the correct answer for the right reason. (Perhaps your iterative process needs explanation.)
    The real test is if the program can identify a color from a non-color adjective,
    And due to your constraints, “without knowledge”, it is incredibly random.

    Chartreuse? :bugeye:
     
  22. G71 AI Coder Registered Senior Member

    Messages:
    163
    Baal,
    So where are the calculations? How did you get the 10.4 number? I think it depends a lot on a particular design so what kind of performance testing did you do with your system? You were planning to develop some criteria for measuring the intelligence level. What's the status of that? How do you define "thinking out of box" for the purpose of AI? Do "C" and "W" have any AI web sites? BTW Testing should be performed by people who were not directly involved in the development. If the system is capable of solving relatively complex problems then I could probably find some time for free testing. Even thought I'm primarily a developer, I'm a pretty good tester. I would be surprised if I cannot find any problem in your system. And of course I wish I could finally see any testable result of your work on AI. I have seen just too many general statements and no real thing. Do you have any test scripts written? If it's gonna work relatively well then I could probably also do some free code reviews. Code reviews currently make nearly 30% of my job and I sometimes "feel" bugs just by looking at the 2D shape of the code. If your response time is limited to 12 sentences (or so) then I would prefer if you just focus on addressing the questions. I do not care if I have to wait 2 weeks or so for some answers. I'm busy with many other things. I'm just interested in getting answers at some point which is often not the case when I talk to you. BTW my discussions with the other AI developers are so different that I often wonder if you really are who you say you are. So show us finally some practical results if you have any and let's hit some details. The too general AI discussions are already kind of boring to me.

    Leda (or any other linguist interested in AI),
    please contact me at G71ai@aol.com before the end of September. I'm looking for some help with evaluation of a new User-AI communication interface. Thanks.
     
  23. Baal Zebul Somewhat Registered User Registered Senior Member

    Messages:
    388
    Stryder, that behaviour is well covered in my design. Of course "fogetting" the ability to speak is rather extreme and that is not featured in my AI but the whole mechanism that leads to it is already included.

    Jiri, i spent the last 11 days observing nature (or more accuratly, its inhabitats)
    During the day i made discoveries, during the night i documented them.
    I now how some work to do:
    • Correct a common phrase by Darwin
    • Correct Freud
    • Prove modern math (related to physics) incorrect

    And i will not answer any questions about that, because:
    sM & oM > iM

    And no Jiri, my team are not AI developers. But you already know that.
     
Thread Status:
Not open for further replies.

Share This Page