Mikail

(no subject)

New version of djuzeppe online AI engine is ready. Now you can teach Natasha bot from main page.

http://project.comex.ru/ai/ai.php

Now bot know mostly russian, but you can teach her english language.

There are three inputboxes in teaching form. In first put keywords, in second - answers, in third - tegs/themes. Like in flickr you can assigne tegs/themes not to pictures, but to keywords and answers.

Divide keywords and tegs with one comma ",". Answers divide with two commas ",,".

Alicebot

I decided to randomly chat with the older version of ALICE. Not wanting to sound crude, I will simply state that the programming is outdated. Does anyone know of any other AI programs I can chat with for free?

Penrose

Could anyone explain Penrose's argument in his Emperor's New Mind book why (paraphrase) machines will not be intelligent? I (gasp) read the book and didn't get it (didn't agree with it at least). Funny how the man never mentions probability theory.
  • tvn

Genetic programming help.

I am trying to create simple program to learn genetic programming - basically I am trying to evolve the solution of "or and and x y z and and not x not y not z" , abc or !a!b!c.

Some of the settings I have is chromosome population : 400 , generate 100 , fitness is calculated by having each chromosome result test agains 8 cases (x = 0,1 , y=0,1, z=0,1). Crossover rate is 90% and mutation = .1. I also have some simple function to make sure the chromosome size doesn't get too large (max size = 70).

My program is able to solve easier problems such as ab~c or ab+a~c and so on but can't solve the original problem. Any suggestion on what I should try ? For instance: I can cheat a bit and tell it about my original solution but not sure what tips I can give to it.

Thanks in advance.
  • peetm

HAL

I hope this has never been posted before, but ...

When do you think HAL will be achievable (however it/he does it, and whatever it/he is)?

Droping the visual/aural/oral processing - do you see HAL as becoming a reality any sooner?

Cheers
sam

Sic transit gloria Minsky

Yesterday morning (Tue 14 Apr 2003), Wired News ran an article on Marvin Minsky's recent talk at Boston University, lambasting the state of the field in artificial intelligence research. I must say, if I weren't convinced that Dr. Minsky has lost the plot, I am now. I'm sure I will wibble more on this subject later.

"Surprising and disappointing." - Stuart J. Russell, UC Berkeley
"Marvin may have been leveling his criticism at me." - Rodney Brooks, MIT AI Lab

Not exactly the kind of blurbs one wants for one's next book...

Here is the Slashdot discussion thread on Minsky's talk. Any comments?

(Cross-posted to my LJ)

--
Banazir
  • Current Music
    P.O.D. - Sleeping Awake
Daz2016
  • daz71

My Virtual Self

I've combined an AliceBot with a Biovirtual 3d head to create a virtual me.
Currently I'm hosting VIRTUAL DAZ at www.pandorabots.com but would like to eventually move the AI to my own system and possibly have it give more specific answers.
I'm going to promote Virtual Daz more on my home page and in my journal, see if I can get some sizable chat files to help me develop him more.
Mostly Harmless

New Member

Hi! My name is Richard. I just joined yesterday. i have been a fan of AI and Alicebots for quite a while now. In fact I have one (known as HALBot) here (PandoraBots).

I am an AS student (Yr 12) studying Physics, Music Tech, Maths and ...Computing. Bye for now.

Rich.

ai

Computer programs are, by and large, one-dimensional. They proceed linearly in chain fashion from one instruction to another.

Thinking, however, appears to be done multi-dimensionally, considering several aspects of a question at one time.

It is possible that this can be emulated in artificial intelligence programming.

However, thought processes first start with related things, (relationships of objects, objects being patterns) in no particular order, then proceed to order them in a semi-logical flow according to "weights" of value, importance, or relevance.

Thoughts may generally proceed in a linear fashion, but are also multi-threaded, capable of being redirected at any point by additional input. You can sort of think about what you are doing while also considering the weather outside. Can computer programs do this? Yes, of course, especially if they use parallel processing (more than one CPU).

I think that it might be quite possible to emulate thinking if a program could be developed which ran several or many processes simultaneously, with the various processes in communication with each other, which is what occurs subconciously in the human mind.

The interaction between the processes should probably be similar to moderating a discussion group. One process might "feel" that what it was working on should take priority, while another would have a different "opinion," thus requiring a "master control program" to make arbitrational decisions when contention develops. All based on a system of priorities, weights, and values.

Thus a true ai program would tend to be rather a group of programs operating as a community.