The future of computing, Linux and AI.

[AICollector]

About 20 years ago, there was a supercomputer. I cannot remember much of it, save for the fact that it had a whopping 235mb hardrive.

Now look where we are today, with 1 terabyte HD's becoming more commonplace. Miniature computers designed for mobile use that have the power of a fully fledged desktop not more then five years ago.

Now, lets look into the future here. Tasks that take up a large amount of cycles now will be more or less pittance in a few years.

But now, the warning;

AI technology (recognition of audio, visual etc) is strangely lacking in the world of Linux. AI Tech is, sadly, within the realm of propitiatory software.

So first, lets look at face and gesture recognition.

http://lifehacker.com/software/featured-windows-download/add-face-recognition-login-with-bananascreen-277812.php

That leads to a free face recognition login system for Windows. I can understand why that hasnt been implemented in Linux systems, simply because face recognition is far too resource hungry for now.

Then we have other systems, gesture computing, I would hope most of you are aware of Microsoft's Project NATAL? I would personally love to have an natural langauge- agent based computing model to replace the desktop.

This brings me back to my other point; speech recognition. There are really only two good speech recog systems, Dragon NaturallySpeaking and Window's built in version. The former is much much better then the latter.

I can only feel that speech recog to be of infinite use on any computing system. How many times, as a writer, have I found it so much easier to verbilize a story, rather then sit down and write it? There is CMU Sphix, of course, but I do not beleive it is ready, nor do I think it is being developed anymore.

My point is; technology, as we all know, is advancing, and AI tech (semantic systems, recogntion, etc) is becoming increasingly important. I would hope Linux entire would take advantage of such technologies, for I think it'd be such a waste if it didnt.

I recognize KDE's NEOPNUK intergration, and GNOME's yet-to-be-developed semantic spaces. I also recognize the increasing importance of automation of both operating the computer and of providing information to the user. I simply beleive that technology, like the one mentioned above, is quite possible and there is little reason for it not to be intergrated into Linux in the future.

[nickcollie]

oh I agree entirely. except for one point. why do I need speech recognition when i type faster than I can speak. processors and the human brain are the quick bits of the equation. speech is just a bottleneck. And, hey, my kids can taxt message faster than they can speak and that is using a rubbish keyboard and two fingers....

Lets find a better way of interfacing...

[philcamlin]

i like ubuntu and thats what im staying with

we should show this to windows users :)

[pookiebear]

The fuzzy logic stuff for artificial brains is just like programming in the fact that it is mathematical in nature. You just have to layout the right math. When a few more breakthroughs in taht side of the math happen, then you will see some real stuff developed by the open source guys and the closed source development teams.

[AICollector]

(EDIT; Sorry for the long post, this is something close to my heart so I have a tendancy to rant on about it, please, dont be put off)

Well, lets look at some of the tools avaible.

OpenCV, a free computer vision library under the BSD license.

Using this library, it is possible to get face recognition system.
It's just come to mind that a video call or other webcam activity could be used in a odd (but fun) way, Logitech did it with some of their cameras, take heads, faces, eyes...and tweak them. Ever been greeted by a talking shark?

Wordnet 3.0

Wordnet is a machine-readable dictionairy with semantic access.


DBPedia;

an online machine readable version of Wikipedia.

Wolfram/Alpha

Mathmatical system that can take apart many many queries


Google Sqaured

A project still in the labs, but if you type in something like 'dog breeds' it will list, with pictures and other assorted information, infomration about dog breeds.

Siri
A commercial shoot-off from DARPA's Project CALO (Cognitive Assistant that Learns and Organizes) and is coming to mobile phones (Android included)

I would also like you to note the KDE app 'Digikam' and the desktop netural ImgSeek accept content based queries (both are image managers, and by content based queries I mean you can draw a little picture and it'll respond by finding a match, ImgSeek is in the repo, by the way)


There are many FOSS AI programs online, especially in Sourceforge/Freshmeat, if memory serves, Richard Stallman himself started with AI.


@nickcollie; you may not find speech recognition that useful. It's more like a quirk of my mind that I tell stories better then writing them, and you and I cannot doubt how useful it would be to the handicapped.

Same with Speech Synthsis, really. We have a fairly good voice with Orca, but recent advanceds in this field give us perfectly naturally sounding voices.

My point really is, whilst the field of AI uses Linux for most (if not all) of it's work, it almost never gives back unless it's an old, discarded item. AI takes a lot to produce and there are very few FOSS programs still functional today. There is almost 100% propitry software here, and of course, these projects don't last long unless they become really spectacular.



Now, lets have a bit of fun and look at potential applications for this technology in, say, Ubuntu.

1. Computer recognizes me by my face and logs me in.

2. The Computer understands my command in natural langauge format and executes it (sort of like an interface layer between the user and the command line to make it more friendly) like 'Format external harddrive'

3. Computer knows my habits and knows that I like to use Firefox, thus it prepares to launch it, shortening the time between me 'offically' launching it.

4. I move away from the keyboard and go downstairs to get a snack, the computer can see I am not in front of it and thus changes my IM status to 'away' whilst not popping up notifications until it can see me again. (though it could just as easily track my status by monitoring keyboard and mouse input)

5. I use Facebook alot. The computer finds that those who use Facebook alot tend to download a particular program. The computer recommends this program to me. (here, this program displays Facebook updates as notification bubbles on the desktop)

6. I attach another monitor to my computer, my computer is aware of what I do when I attach another monitor (which is output another desktop to it) and automatically does so.

7. I'm on an IM client, but I am not in front of the computer. Someone tries to talk to me, but here the computer, detecting I am not around to respond, activates a sort of smart auto-responder/chatbot to converse with them.

8. Speaking of chatbots...let's have an offical Ubuntu-sanctioned Psychiatrist with us; Meet ELIZA, she won't judge you, she'll listen to you pour out your heart, ask questions which assist in sorting out your own feelings and thoughts, and, best part, dosent charge by the hour! ELIZA was the first chatbot, and I (and many others) have noted her 'talking therapy' actually helps when trying to organize your own thoughts and feelings. She's only about four lines of code in Perl (and a database). How I wish someone would make her into a program for Ubuntu!


I fully beleive that intergration of such technology is plausible and indeed, encouraged because it can make our computers so much more useable and so much more friendly (In a few studies I've done, should the computer be given a 'personailty' (IE, voice, the 1st person viewpoint of pop up boxes "I'm sorry, I was unable to burn that disk due to a fault with it, perhaps you'd care to insert another one and I'll try again?") the user expereicne is more enjoyable, there are other studies avaible about this, one or two focuessing on productivity (They found it more productuve, by the way) The automation of tasks and the increasing benefits of such technology is something that should not be ignored. I hope that Ubuntu, leader among the user-friendly distros, will see this, as will GNOME, KDE and the other distros)


One final note; The FOSS system is very good for developing AI, and any such FOSS AI system can take advantage of this; http://en.wikipedia.org/wiki/Artificial_intelligence_systems_integration

I hope you can see just how troubling that is within the world of closed source software. FOSS is perfect for that because developer A can hook into Developer B's software. Closed source AI has a tendency to reinvent the wheel at times, if you know what I mean.