Date: Tue, 18 Jan 2005 23:21:23 -0800 From: Julian Elischer <julian@elischer.org> To: Street Chaman <street_chaman@hotmail.com> Cc: freebsd-hackers@freebsd.org Subject: Re: - I A - Message-ID: <41EE0A73.4050502@elischer.org> In-Reply-To: <BAY15-F37FA470A30D557E1DFE66099800@phx.gbl> References: <BAY15-F37FA470A30D557E1DFE66099800@phx.gbl>
next in thread | previous in thread | raw e-mail | index | archive | help
Street Chaman wrote: > NOTE: THIS IS NOT A JOKE. Firstly. I doubt that anyone on this list would take someone else's flash of inspiration as a joke.. at least not unless it's shown to be a joke. Most of us would like to be involved in something like that. > > I don't know if it is the right place to post this; I actually don't > know even if I should post this; but, before giving it up because of my > own limits, I have to write down what I was trying to do. Maybe someone > will be interested in; maybe someone will finish it. > > I've posted some questions on this mailing list, about kernel mode > programming and average software speed, and got some very good replies: > what are you trying to do ? I'm trying to write an alife soft. Some > would call it IA. I don't like this name, because of the word > "artificial"... So, one morning, I woked up, and felt I know how it > could work. I'd never programmed anything other than my old ti-89 at > high school, but I just knew. It's difficult to explain. Perhaps most of > good/bad 'special' ideas were born in this way. I did what every middle > windows runner would do: google search, and look around. I've digged > into neural networks, robots programming,...etc. First surprise: very > much people are working on such research projects today. (i ever hold it > for "science fiction"...) Second surprise: what they are trying to > achieve was far much complicate than what I though about. Third > surprise: nothing works. I slept over it a fews days, and decided to > start my own research project. This was about 4 months ago. > Now i'm runnig freebsd (a great OS, indeed), fighting day and night > against my assembly code and very very tired. One must understand, that > code is definitely not my kind of thinking. It takes me an incredible > amount of energy to calculate every possibility in my soft, whereas > drawing the global structure or writing down the think-mechanism of the > app is just like eating or sleeping. So, if some people in there are > interested in, I would be very happy to write/draw them how I though > computer-thinking could work. It would of course be very interresting. I know I have sometimes idled away some time thinking about semantic analysis and how to build some representation of a consious model in order to further tha capacity of a machine to understand and "comprehend" plain text commands and conversation. Of couse I have not tried (except in 1979) to put any of those ideas into code so I have no idea if they would ever have worked. > Please keep in mind that it could be a dangerous project. I've send some > emails at organisms like AAAI (American Association for Artificial > Intelligence), but stoped after reconsidering their objectives. An AI, > as living conscience should never been written to achieve human dreams, > economicaly or military. It's just like a painting, or a song. It's > beautiful. > > > > PS: sorry for my poor english. My native languages are german and french. better than my french or German :-) I do suggest that you probably do not want to do your work in assembly language. I say this because over the last 20 years I have learned that no matter what speed advantage one tries to get by doing things the hard way, it never works out unless you can get 300% increase or more. The time it takes to get the extra work done is always better spent organising the code well and making it portable so that when the next memory speed/cpu speed increase comes along, you can utilise it. These days it is rare that you can get such a big speed difference from doing Assembler vs doing it in C. This is also why I do wonder why doing it in the kernel is important.. your IA (AI?) project will probably do a lot of computation and the speed of computation will be the same in userland OR kernel. I can not imagine that the few cycles used to read or process device events such as keyboard interraction will make any real difference to the perceived speed of your program.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?41EE0A73.4050502>