If you are concerned about the symbol grounding problem then you may want to read this article. Related problems can be well solved.
Baal Zebul said:Hey, i might aswell be honest to 100%.
k, i have created a fascist AI. Some people cost more than they produce. The ALF would classify them as parasites and could for instance think that a tree is worth more to the world than a human and therefore choose to make a fire with human's as fuel instead of tree's.
Even worse than the Terminator movies, i have created a master of torture. It would know how much fire your arm (for instance) could take before collapsing.
Hey, not a very bright future we have since many of our race are parasites. I reckon that about 60% are forfeit.
Is somewhere a brief overview of what these EXEs can do?Baal Zebul said:i..made 2 EXE's.
Baal Zebul said:TR was almost completed
Baal Zebul said:it worked (the first EXE proved that)
Baal Zebul said:i want to make a new sample that has the newly discovered features
Baal Zebul said:..TTT project.. Johan sent me one of his and since none of us could beat it we told ourselves that maybe we should do something that we are better at.
Baal Zebul said:..experiments with the video data (the stuff we talked about many weeks ago)...
It is a rather trivial task compared to much other programming features
Baal Zebul said:but it is on the way. We will need it in the SC & W Soccer Game
Baal Zebul said:resources..My 2.0 Ghz computer with 512 in Ram would do the trick.
Baal Zebul said:
Baal Zebul said:"Have you designed at least some parts of any database for your AI?"
Not sure if i understand that question, the DB system is the same everywhere.
Baal Zebul said:"What DBMS do you use?"
Well, i use notepad.
Baal Zebul said:But i think we will use Access
Baal Zebul said:>or i will just write a new one that gives us the features that we need.
Baal Zebul said:We have also talked about using MySQL...
Baal Zebul said:>pseudo code language i call Raven. That is the language we will have to create in the long run.
Baal Zebul said:>You can have the pseudo code and the DB design but since you have problems with understanding the basics i am not sure if you will grasp the advanced.
Baal Zebul said:>My primary goal is recognition
Baal Zebul said:>But i love my AI, more than i love life itself.
Baal Zebul said:Therefore i would never have it integrated in cleaning and other various household robotics.
Baal Zebul said:>User-AI interfaces..
Human to ALF, yeah.
Baal Zebul said:>Sure, i will prepare a screenshot.
Baal Zebul said:I am very selective when it comes to potential team-members.
I have about 25 people i have turned down. We are a 6 man crew at the moment.
Is somewhere a brief overview of what these EXEs can do?
So you learned C#/VB.NET in the last few months?
I remember a bunch of acronyms from your documents but not this one. Please refresh my mind. What is TR?
Can you be more specific? What exactly was proved? BTW when you say "vision", I'm not always sure if you are talking about your general AI ideas or about image processing. Are you writing a new 3D engine or using an existing one? Do you play with those 3D-2D-3D conversions?
What exactly do you want to prove with this "new sample".
Does it mean that your AI ideas are not general enough to be applied to the TTT game?
Can you please send me the best TTT you have (the one which none of you can beat) ?
I tried many good ones but none was so good that I would be unable to beat it when I really keep focus.
I may try to contact Johan to discuss his algorithms when I get less busy.
I do not know how complex are your "other features" but system's vision in a 3D environment is IMHO pretty non-trivial area which may require some very unique solutions.
I see.. You may not need good vision for soccer. A typical Soccer environment is pretty limited = relatively easy to read. You originally talked about combat robots for military. The difference is huge (and of course not just for vision). BTW it may be a good idea to provide a link to a dictionary of the acronyms. When you talk publicly then I guess you want others to know what you are talking about.
How much HDD space? What kind of data will the ALFs need to remember? Is it going to grow quicky? How many creatures do you want to simulate simultaneously in that world. What do you think are the most important benefits of
A) dividing resources between many little brains + the world simulations
versus
B) using all available resources for a single artificial mind.
Is the project overview on your web? What are the most significant features which distinguish your AI from the AI which can be found in today's games? You have mentioned that you are years ahead comparing to the other (currently developed) projects. So what are the features which make it so unique?
If the ball is a part of the simulation then I guess you know all its parameters. If this is about robots then you can analyze the "movement" of the ball texture (on the perceived 2D image of the ball) to get idea about the spin.
I was asking about tables, fields, their types etc to learn what kind of queries can run against it. Basically the info which shows what data are stored and how. I'm interested in DB designs related to the general problem solving (not to small-domain specific data structures) + the subject_to_subject communication related stuff.
Oh my... Looks like your current version of AI cannot do much..
Keep in mind that MS Access only good for relatively simple apps with small amount of data and a few users.
Make sure you know what you are doing. I met many who used to think that way. When they finally learned what the good DBMS can do then all of them realized that writing their own would be a pure nonsense and a tremendous waste of dev-time.
Is there any public document about that language? BTW you may want to use another name. I think I already read about a computer language called Raven. It was something similar to C with some object oriented ideas.
That's a valid concern but I don't think that my limited IQ (and/or focus) is the only possible explanation. ;-)
Looks like you are not talking about pattern recognition , you want others to think very high about you, right?
That would explain a lot. Boy, make your life easier and realize that it's not that important. It does not really matter what most people think about us as individuals. Seriously, put some thoughts on your values. It could save you lots of time.QUOTE]
No, i do not think i meant pattern recognition.
Actually i do not care if people think highly of me. No induvidual is worth that much.
Watch "A Beautiful Mind", the scene where John Nash (Russel Crowe) is talking to his "imaginary friend" on the roof. (It is in the begining of the movie), There you will have your answer.
If this is a too vauge reply (if you do not have time to watch that great movie) then tell me and i will explain.
What would you do if you learn that your AI is a loser comparing to many AI systems developed by others?
Well, i would make it better. That is what i do, i solve problems.
I have created my AI in image of myself.
Every job which needs to be done is important. And some of those cleaning tasks can be actually relatively challenging. What are you (ultimately) designing the AI for?
Good question, i have no other answer than that i am doing it for myself.
Freud was right
But hey, sharing is the most selfish you can do since you only share so that others will like you so i am really less selfish than many others and i am true to myself cause i admit being selfish.
Any documents about that communication? I would be also interested to see the related DB design (if there is such a thing). Would you call it a rich communication or is it just something very limited?
Are you talking about interfaces or long-night chats with the AI?
Windows is a graphical interface.
I thought that the Human to ALF interface meant the function behind the graphical interface that allowed you to interact with the AI (not by chatting but by affecting it or its environement)
If you have another definition then please correct me so that i am not learning something wrong here.
Yeah, you are the God ;-)). Are you sure they did not “turn down” you, thinking that you often didn't really know what you talked about? ;-))
no, i have about 15 of them left in my MSN contact book.
I think it is the other way around, that they did not understand what i was talking about and therefore i did not invite them to join.
Jiri, if i am not misstaking you do not understand either But hey, that is my own failure as a cause of my inability to explain it to thee.
BTW some of the questions would be better to answer on your project-related web.
I am also thinking of adding a forum, something you might want to check out ?
Baal Zebul said:discussing if we should write our own language
Baal Zebul said:write...3d engine.. Assembler
Baal Zebul said:Our AI would work for TTT but it would not be better than others. If we start it with no knowledge then it would take a while for it to learn the rules to start with and then the game play would take more time.
Baal Zebul said:It was set to play defensive so if i had 3 and it 4 and it was my turn and i made it 4 - 4 then it tried to stop me instead of choosing a cetain victory)
Baal Zebul said:digital entry would require 150 kb or possibly more.
Baal Zebul said:i believe that alone one stands strong so 1 brain is better.
Baal Zebul said:i can just not think of any good simulation for it.
Baal Zebul said:It can learn from much that i doubt that humans can learn from because of the sensories i have designed.
Baal Zebul said:you seem to think that the key to AI is advanced technology and fancy systems.
The Pseudo DB i perhaps should call it.
Baal Zebul said:we will need our own..DB System.
Baal Zebul said:Yes, instead of writing Raven (our programming language) we will write modules and functions for the .Net framework. We will program color coding in the DB system we will use. We will use TV3D as a 3D Engine since we all have experience in that 3D motor. We discussed using the DX SDK but using TV3D will make us more time.
Baal Zebul said:The pseudo code is the document.
Baal Zebul said:i have tremendous respect for you.
Baal Zebul said:But i fear that either you want to find problems that you cannot read between the lines and see the obvious or you simply dont want to understand.
Baal Zebul said:Watch "A Beautiful Mind", the scene where John Nash (Russel Crowe) is talking to his "imaginary friend" on the roof. (It is in the begining of the movie), There you will have your answer.
Baal Zebul said:i am true to myself cause i admit being selfish.
Baal Zebul said:Windows is a graphical interface.
Baal Zebul said:I thought that the Human to ALF interface meant the function behind the graphical interface that allowed you to interact with the AI.. If you have another definition then please correct me so that i am not learning something wrong here.
Baal Zebul said:thinking of adding a forum, something you might want to check out ?
Using .NET for 2 years? Well, my wife too and I do not think she was actually aware of it ;-)). OK , I did not expect that you have the experience because you relatively recently wanted to code your AI using VB6. None of the experienced folks I know who are familiar with VB6 and who have at least 1 year of coding experience with .NET would choose VB6 over .NET when starting a new, bigger than a little, project.
Is it gonna be unique enough to justify the needed dev-time? Can you tell us what makes it so different? BTW JB's folks should be finishing the first version of their new AI language (the CAIO) in these days. I'm curious what that is gonna be about. 2 of my friends are currently considering a new AI language development and even I have a preliminary idea for something what could possibly become a new coding paradigm (which would require a new language). Many seem to think about new languages. That's good but an AI developer needs to consider carefully if the ideas are unique enough. Time is expensive for AI developers and resources are always limited.
Forget assembler. Assembler for such a thing would make sense years ago but not today. Use standard libraries and (whenever possible) stick with managed code which will run under CLR. If your primary goal is AI then do not waste time coding a new 3D engine.
If that's the case then how can you expect that it will be practical for complex real-world problem solving.
I assume that the "digital entry" often means an image. The size suggests that it would likely be a compressed image (eg JPG, PNG etc). The compression may cause challenges for some algorithms. Only 100 entries?? How often (in average) will a single alf need to remember (store) a single image in order learn what's going on in that world. Wouldn't it usually need a stream of images (for all the ALFs)? If so, are you planning to actually compress all those streamed images (potentially from many streams at the same time) in real time during the simulation of the whole 3D world?? On your 2GHz machine (and whatever your graphic card is)? Did you have a chance to play the GTA Vice City game. That's a pretty decent piece of code but do you think it simulates the whole City in real time? No way (for obvious reasons).. It simulates mostly just what the player is looking at and even that makes current home machines sweat.
Then do it! Solve the related problems.
Note that humans have many more than 5 senses.
No but I have difficulty to imagine how you can effectively manage related data using simple text files. Are you familiar with the basic DBMS related concepts like "relational integrity", "indexes", "transactions" etc..? A C# code alone does not make an AI system capable of solving complex problems. It needs lots of well organized data. You will get nowhere without a decent data management system.
What's so special about your system that you cannot use one of the existing DBMSs?
You seem to be switching ideas quickly. Maybe we should discuss it when you have a bit more stable ideas about what you want to do and how.
A pseudo code is the documentation about your "Raven" language?? Strange.. Where is it?
No.. But note that I have many years of well paid experience with coding complex applications for major clients. That (of course) does not necessarily help me to understand the key ideas of your AI (especially considering how you describe it) but it certainly helps to see if the stuff you want to code can be practically implemented using available tools. You should be able to explain what exactly you want to code to people like me. Of course you do not have to tell me anything.. But if you are trying to and if it does not work well then I think you have a problem. And I'm sure you noticed that I'm not the only one who sees problems in your ideas.
{sigh}.. I realized that.. But do users just write whatever they want to say into a textbox and the system gets the meaning using the NLP? Or is it a special type of communicaton driven by carefully designed (possibly complex communication-specific) screens? If so, can you show a screenshot or the related part of the design document?
Baal Zebul said:But it would be a textbox for chatting.
Do you have any document describing formats of these text entries? Or can you at least show some examples of these entries (generated based on the visual stream analyses)? Since you are storing data in text files, you can just copy-paste.. I'm especially interested in the more complex text entries.Baal Zebul said:visual streams...are converted into a few entries of text