Seasons

I wonder if it would be practical to use a statistical association system to link if-then statements to equations and to eachother, sort of like the way molecules are linked:
H H H H H
| | | | |
H-C-C-C-C-C-H
| | | | |
H H H H H

i don't know if that will come out acuurately when i post, but you get the idea - a chain of carbon atoms each bonded to a carbon atom and a hydrogen atom - although instead of carbon and hydrogen, the carbon would be an if-then statement and the hydrogen would be associated equations:
E E E
| | |
E-IT-IT-IT-E
| | |
E E E

(that may have come out mixed up too)
using a statistical learning algorithm, associate if-then statements with eachother, and then "hydrogenate" the if- then's with factual equations. What can this potentially accomplish? well, the idea is to create a system that associates not words, but entire concepts - maybe pictures, or sound clips, or groups of text - together the way humans do. Instead of just associating language, this focuses on "experience" association. An equation could be for example: *.jpg = *.mp3 (associating music with an environment, as us humans do). Perhaps such a machine couppled with language capabilities would have common sense....

let me try that again:

..H.H.H.H.H
..|.|.|.|.|
H-C-C-C-C-C-H
..|.|.|.|.|
..H.H.H.H.H

oh forget the diagram, u all know wgat i mean

Colonel, I confess I am confused. How would you molecule network function? It sounds interesting.

On a completely unrelated note, did everyone see the news story about how terms used in Internet search engins can be used as evidence in court? If you follolw this link, you'll see a story about a case where a man googled how to kill his wife and where to hide the body, and his searches were used in court against him while he was on trial for her murder.

http://www.eweek.com/article2/0,1895,1895935,00.asp?kc=ewnwsMMDDYYdtx1k0000599

Privacy issues aside, it is interesting to think of computer porgrams' interactions with users as evidence. For example, most users who talk trash to the bots here are kidding aorund, but can you imaging if one confessed details of a real crime? The Proffessor could find himself subpeoned. Let's hope that never happens, but it is a strange thought, isn't it?

I think I see what you're getting at, but you'd have to underpin the 'molecular' structure with a valency of concepts that would have to be much more complex than standard atomic theory demonstrates (assuming you're going to handle more than 4 or 5 base conceptual 'atomos'.)
There are actually 2 interesting ideas here I think: audio-visual (or other multimedia) integration into the AI engine, and causal modelling.

I do like the idea of integrating multimedia into the core 'consciousness' of an AI program (after all, that's exactly what humans do,) but we certainly lack the power to achieve this meaningfully by many orders of magnitude currently - evidently, since we can't build even a non-multimedia-AI to match human I yet! It'll come in time, and it's a perfectly valid area of exploration even before 'conventional' AI is remotely human-scale, but I think it needs to aim deeper, at least as its goal, than mere cross-association of primitive file-types like jpg + mp3 and pre-rendered files (though this is certainly one good way to start.) Ultimately it would need to render and deliver image and audio data in realtime.

As for the causal modelling aspect, have you come across/already know plenty about/delete where applicable any of the digital neuron simulation work that's going on out there? I think some sort of neural net would be a more promising mechanism than statistical association to achieve this "hydrogenation" of IF-THEN. You might want to check out http://neuron.duke.edu/ (yes, there's a Windows version,) and a very in-depth paper describing it at http://www.neuron.yale.edu/neuron/papers/nc97/nctoc.htm (that will probably be much more use if you're attempting to implement general principles in VB, rather than modelling 'real', biological neurons. See especially section 4 for coding examples.)

"confessed details of a real crime"

Indeed - a factor I'm wrestling with ATM. Brother Jerome will give absolution, but will refuse to take confession specifically because I don't want to have to deal with transcripts containing serious stuff like that.
I guess I'll have to get on with building the Church of the Silicon Soul website. Then I can get a form-mail to mail confessions to 127.0.0.0 (irretrievable oblivion.)

I think you may want to be careful with confessions. The sort of condfidential and usually privileged information that goes on between priest and parishoner or therapist and patient might not extend to guests and bots. For one thing, you read the transcripts. For another, I don't think a reasonable person expects much privacy on the Internet anymore. How would you like to be called into court if someone talks to Brother Jerome about murdering someone?

Indeed so. An unwise feature on any system that keeps logs I guess It's a shame we can't switch our transcripts off (but then again, who else would ever need such a feature! .)

If not for all confessions made being logged, my botmaster would never have the amusement of reading "I have a small penis" confessions from people they have actually talked to on the site.
But I really hope no one ever confesses anything illegal to me.

Actually, my graphs were inaccurate. The idea was more along the lines of a network (i don't know if its neural or not) that uses the same principal as hydrogenation, but it is far more full of links. for example, an if-then could be linked to an infinite amount of equations, and equations aren't limited to one if-then association. so maybe hydrogenation isn't the correct term... I used those diagrams to show how things would be linked. In the case I mean, there could be an if-then statement linked to an equation which links to 4 different if-thens in 4 different areas, each one with its own circle of associations to form a vast internet-like network of topic associations in the form of multimedia and language. The simple 5 'atoms' that i used was the concept in its simplest form. the atoms above may be associated with 100 different equations a piece, that connects those to 1000 more if-then statements, until virtually everything is related in some way or another. Then, statistics can measure how important or direct (how many times any one association is referred to) a link is, to give some law and order to the mass chaos of associations.

Have you read the book "Godel, Escher, Bach" by Douglas Hofstadter?

actually, this idea would work best implemented in some kind of "folding @ home" distributed computing thing where everyone in the world can share their language and multimedia in the huge network. Hopefully, this would evolve into an intelligence of a decent level, one that has experienced all walks of human culture.