View previous topic :: View next topic |
Author |
Message |
ex4
Joined: 13 Jan 2006 Posts: 1062 Location: indonesia
|
Posted: Tue Dec 28, 2010 1:10 am Post subject: |
|
|
hehehe, theory vs reality vs simplicity |
|
Back to top |
|
|
hgrueneis
Joined: 04 Apr 2009 Posts: 902 Location: A-4786 Brunnenthal
|
Posted: Tue Dec 28, 2010 3:22 am Post subject: AI |
|
|
Meanwhile I have a feeling that the reason for the mayority of people to not want AI in Machines
is because they are afraid that the machines they own or not own could be more intelligent than they are or see their source of income compromized.
Set one example: One programer makes his money by writing programs for gadgets af any kind.
It takes the person a certain time and she or he can charge something for it.
That same person would be out of luck if an AI computer could create the same or even better program in 2 seconds.
AI small scale in microcontrollers or large scale computers scare people in general.
If everybody in this world is opposed to an IBM set of mind,
then IBM or other Companies that have an inquisitive mind and research the possibilities of AI
will rule this world of dumbed down people who can only follow exactly those innovative companies rules.
Movies like 1984, Westworld and other predictions have scared people.
We do not need to advance to voice programmable, self altering and advanced learning intelligent computers
but we certainly can try with a joint effort to get a little bit accomplished.
If its in dhe end just a collection of sniplets for different sensors and algorithms for different applications that make it easier for all.
An AI computer can consist of hundreds of networked controllers.
Like Per said: A LINUX set of mind. For some that make money with their programing idears it may be too much to ask for.
Pretty soon the possibilities for hobby enthusiast to build their own boards will be over anyway because the placing of components will become increasingly difficult in a homebased environment with limited possibilities.
I can picture an AI computer like an emulator that can be pluged into a board in place of a microcontroller and compile an intelligent program for that board and type of application with just a few directives.
So, should we use the time we have left? ....or not.
Hubert |
|
Back to top |
|
|
ex4
Joined: 13 Jan 2006 Posts: 1062 Location: indonesia
|
Posted: Tue Dec 28, 2010 3:31 am Post subject: |
|
|
is it a big request if we'd like to see some simple AI program to be implemented in ATMEGA AVR series?
ATXMEGA is big money here |
|
Back to top |
|
|
hgrueneis
Joined: 04 Apr 2009 Posts: 902 Location: A-4786 Brunnenthal
|
Posted: Tue Dec 28, 2010 4:06 am Post subject: |
|
|
The talk was of Atxmega for speed but it does not need to be. If you build a network of processors as you know, they do not have to be the same kind. You just need a common way of communication in between them.
The result is the important thing.
It could be a concept where each MCU has a dedicated function like in a multitasking system with Masters, Submasters and Slaves like a ring or mesh and each has its own address. It could even be a parallel bus system with a few control lines, kind of like TWI, only much faster.
An Atxmega192A3-AU can be ordered online in the USA for $7.99 at a quatity of 1. It is a matter of finding the right distributor. Some Atmega cost more. It also has the advantage that it can be overclocked (at least now), which does not work so good with Atmegas. The advantage of the A3-AU version is that the pin to pin center distance is 0.8 mm.
That makes it for a lot of hobby enthusiast fairly easy to produce a board.
The trend that it is not available in a DIP version should make some people think again. Once it goes into BGA then I am lost too because I do not have the equipment to produce a board for that.
So there is a lot of flexibility right now.
Hubert |
|
Back to top |
|
|
AdrianJ
Joined: 16 Jan 2006 Posts: 2483 Location: Queensland
|
Posted: Tue Dec 28, 2010 5:34 am Post subject: |
|
|
I dont have any problem with letting the machines "take over", if by that you mean taking some of the drudgery out of coding, and maybe even some of the systems analysis.
The "just a few directives" is what I think you will have most trouble with.
How many man-years of effort does it take to program a baby to be intelligent ? And its already a human.
Self-optimising code to do things like picking pathways, and navigating to local minima of solutions, eg the sensor calibration problem, are probably possible, but it still seems to me you might as well do that with the power of a PC, then run the optimised code in an AVR ( or any other micro, for that matter ). Fast and powerful as AVRs ( and I include Xmegas ) are, they are still several orders of magnitude slower and less capable than a PC. Provided of course you dont cripple the PC with some kludge like Windoze. _________________ Adrian Jansen
Computer language is a framework for creativity |
|
Back to top |
|
|
Per Svensson
Joined: 03 Oct 2004 Posts: 235 Location: Gothenburg, Sweden
|
Posted: Tue Dec 28, 2010 3:55 pm Post subject: Artificial Intelligence? (AI) |
|
|
Hmmm?..Perhaps it was a mistake to use "AI" in the topic line.
it seem to scare some readers away, and offend others.
AI is really no black magic, and neither is it extremely complicated either, but of course it means NEW THINKING!
Also, I must stress, I am no AI expert by any means myself, but have gathered enough knowledge to realize how powerful AI can be IN CERTAIN CASES. Most of the times, traditional algorithmic methods is to be preferred when processes are well known and thereby can be properly modeled, but in some cases it is hard or impossible to grasp the theory behind the processes. In these cases AI comes in as perhaps THE ONLY way of resolving a task. This is particularly true when many parameters are simultaneously involved in a process.
If we take this RC-aircraft example for instance. Yes it is possible to measure all parameters at fabrication time, and then keep them constant. But process parameters often vary slowly over time.
- Temperature changes.
- Fuel is consumed and thus, weight is lost over time.
- Load is added (cargo)
- Controller is moved from one aircraft to another.
In all these situations it is nice to have an adaptive system that "understand" what is going on.
In some (read "many") cases, nature is so hard to understand that a model of the process is impossible to describe mathematically.
AI will most likely solve that task without a model, as long at a good evaluation is at hand. In many cases it is much easier to check the EFFICIENCY of a process than trying to figure out HOW it works.
The drawback is often that it takes TIME to find the solution, but the big advantage is that we can find a multidimensional solution that is guaranteed to converge. ("Nature ALWAYS finds a way").
My own interest was born during the design of an autopilot for sailboats.
(I have tons of theories about this process for those who are interested.)
Such a controller has to deal with a constant process (the boat itself), but also with external factors which are impossible to predict and that varies over time. (wind pressure on hull. Hull tilt due to sail pressure, Wave of all sizes and frequencies, Rudder gain, Steering motor latency, battery condition. Etc etc.) You can not build a stiff and solid model that takes all these varying parameters into concideration. You need some kind of intelligence behind it.
So the reason for YOU to join this group might be:
1) I want to learn more
2) I want to spread my own insights
3) I like the idea to build up banks of knowledge which is formally formulated. (Software banks free to all users)
4) I already have some routines that might fit into this interest group
If all these arguments still fall flat - at least it is satisfactory to widen your own understanding about AI.
You never know when skills will come in handy...right?
Regards
/Per |
|
Back to top |
|
|
Per Svensson
Joined: 03 Oct 2004 Posts: 235 Location: Gothenburg, Sweden
|
Posted: Tue Dec 28, 2010 4:07 pm Post subject: Artificial Intelligence? (AI) |
|
|
Adrian made an interesting point here:
[i]"Self-optimising code to do things like picking pathways, and navigating to local minima of solutions, eg the sensor calibration problem, are probably possible, but it still seems to me you might as well do that with the power of a PC, then run the optimized code in an AVR ( or any other micro, for that matter ). Fast and powerful as AVRs ( and I include Xmegas ) are, they are still several orders of magnitude slower and less capable than a PC. "[/i]
My opinion here is that in many cases it is NOT the lack of CPU power that limits the use of AI. It is rather the time it takes to see what happens after a parameter change. When you apply a parameter change to a child. (You instruct him) It often takes days before you can observe the effect.
In many physical processes you find the same limitation.
You let the AVR spend 1 second to calculate a new set of parameters for the ships Autopilot, but then must wait 25 seconds before this heavy boat respond to the change. A 3GHz PC would NOT help in this case...
/Per |
|
Back to top |
|
|
six1
Joined: 27 Feb 2009 Posts: 553
|
Posted: Tue Dec 28, 2010 4:32 pm Post subject: |
|
|
you're talking about intelligent. so what is intelligent?
Scientists have found that man has made decisions before he starts to think about deliberately.
So explain me intelligent...
Humans had and have always one Problem: They think about it and will never reach the core, because they don't realy understand. All the way it's a little bit like trial and error.
Trial and Error is the only way in time for Humans to discover... and this is the way i'm learning _________________ For technical reasons, the signature is on the back of this message. |
|
Back to top |
|
|
hgrueneis
Joined: 04 Apr 2009 Posts: 902 Location: A-4786 Brunnenthal
|
Posted: Tue Dec 28, 2010 5:52 pm Post subject: |
|
|
six1 wrote: you're talking about intelligent. so what is intelligent?
Intelligence is the ability to use the available knowledge to solve a task.
Knowledge by itself is not intelligence. Most intelligence tests are actually for the most part useless.
I know academics personaly who are by no means intelligent, they for the most part studied everything and recall from memory. If you present them with a problem you will get no intelligent answer.
Many people have a lot of stored information in their brains but lack the ability to logically combine it to come to an intelligent solution.
Most of todays generation is not thought to solve problems, they are given the solutions in school. "Memorize this because it will be in the test!"
Conclusion: Memory is not intelligence, the first sentence applies.
Dumb example: It is green, has sticks, comes out of the ground so it must be a plant. This might be called an intelligent (or not so...) statement.
The so called intelligent solution always depends on the number of variables and their relative weight in the process. The more known variables you have, the more things or processes you can identify and solve.
Does that answer the question? (at least partialy?)
Hubert |
|
Back to top |
|
|
luizabbadia
Joined: 11 Dec 2007 Posts: 112 Location: rio de janeiro
|
Posted: Tue Dec 28, 2010 6:29 pm Post subject: |
|
|
@ Per Svensson:
My friend I think you are doing right ,bringing new topics to this wonderful forum.
In my case , I donīt have a 8051 processor thatīs why I donīt read the 8051 forum topic but that doesnīt mean that the 8051 topic is not important , for many other people it is.
The best reasons I found to expect your topic come alive one day can be listed bellow:
1- I would like to know about and try in my spare time.
2-If I , by any chance, loose my time because of hardware limitations,or anything else , reaching nowhere with this , well...this is exactly were I am now, so I would loose nothing.
3- Remember NASA president talk " If we had mediocre plans we would only achieve mediocre results " , so letīs rock.
.
.
.
n - One cientist asked Faraday when he discovered electricity "For what does electricity serve for?" his reply was " For what does a newly born child serve for? "
So , please go on with your ideas , bring your knowledge , and thank you .
Ah! Happy New Year for all !!!!!! |
|
Back to top |
|
|
Kiedro
Joined: 08 Mar 2006 Posts: 81
|
Posted: Tue Jun 07, 2011 9:30 pm Post subject: |
|
|
Sorry for having read this topic just a couple of minutes ago - half a year later than its original posting!
Per, you are absolutely right! A forum for "kind of advanced" up to "experimental" computational methods IS TIMELY also in the Bascom camp.
I also agree that we need more "openmindedness" - the success of Linux and more recently Arduino should make us alert! "Hiding knowledge" was the privilege of medieval monks - it is not anymore a success recipe of the time we are living in. If a programmer has found a clever new way to solve a problem, he should share it with others just for the sake to promote global knowledge. Money is not everthing - humans do cooperate (and not only compete). Sharing knowledge may (or even may not) pay back in "virtual credits" such as reputation - even without it has value. I think that on a long run this is a better model than the current model which assumes that everything can be viewed from a market perspective.
Natalius |
|
Back to top |
|
|
Per Svensson
Joined: 03 Oct 2004 Posts: 235 Location: Gothenburg, Sweden
|
Posted: Tue Jun 07, 2011 10:19 pm Post subject: |
|
|
Yes Piedro!
Open software and now also open hardware seem to gain terrain these days.
We have many clever guys in this forum, that is clear.
some are good at assembler (fast code)
others are good at math and algorithms
and some are skilled in electronic hardware.
Most of us are individualists it seems, and that is also good in some respect,
but we should focus these wonderful resources in a more efficient and still voluntary manner.
After this little "call for AI" I wrote a few months ago, that caught some attention, but very little action
from the members, I actually wrote a rudimentary AI library. It uses tha Genetic Algorithm principle to
find optimal function parameters ( multiple simultaneous parameters )
As I thought, AVR (Xmega128) have no problem coping with tasks like this even when coded in BasCom.
It can evaluate a population of 50 chromosomes, 9 genes each, in less than 50 ms.
Provided that the fitness function is not too heavy.
What I think is even more interesting is to offer a hardware platform to a reasonable price where our forum members can focus on
providing a better and better software solution to a given problem. (Much like Linux does, but for a given hardware)
The absolutely most important issue is however to find the functionality that is fascinating enough for people to chew on.
The quadrocopter is already taken...
Rotating LED-clocks is not advanced enough to get us going.
Humanoids involves too much hardware..
My "Baby" happen to be "the perfect Yacht autopilot". But of course, it is just a matter of my love to boats and sailing..
It is however a very challenging task. The more you look at it, the harder (and exciting) it gets.
Suggestions invited!
/Per |
|
Back to top |
|
|
Per Svensson
Joined: 03 Oct 2004 Posts: 235 Location: Gothenburg, Sweden
|
Posted: Tue Jun 07, 2011 10:21 pm Post subject: |
|
|
Sorry! - Kiedro is the spelling |
|
Back to top |
|
|
Kiedro
Joined: 08 Mar 2006 Posts: 81
|
Posted: Tue Jun 07, 2011 10:39 pm Post subject: |
|
|
Per:
Think about promoting Bascom AND having your forum!
Why not putting the GA into an application note? You may have to live with a (relative) lack of interest for some time - but on the long run it will grow! This is (trivially) because even the most powerful monk/politician/decision maker cannot stop progress by human creativity!
Natalius |
|
Back to top |
|
|
Paulvk
Joined: 28 Jul 2006 Posts: 1257 Location: SYDNEY
|
Posted: Wed Jun 08, 2011 2:21 am Post subject: |
|
|
Hardware is my area I love building things I struggle with the maths but have just got AVR Xplain ATxmega128A1 from Atmel, with both the Xmega and a 32bit AVR on the one board I would think this should be a good platform to work with AI, its low cost makes it attractive. |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum You cannot attach files in this forum You cannot download files in this forum
|
|