Archive for the 'artificial intelligence' Category

Computer Vagaries.

December 24, 2008

I was reading this article on Slashdot the other day and I couldn’t help but think about how many times I’ve seen this before.  I had actually started writing a post on this back in July, but I didn’t have enough ammunition to make my point.  This article provides ample evidence of how a user’s experience with PCs is not uniform between individual boxes.  This is not just dependent on operating systems.  Time and time again, I have experienced this.  For example, I could not get iTunes to install on one of my Vista boxes.  On an XP machine, Twitter doesn’t work.  On my iMac, I can’t get Firefox 3 to install. My experience as a hardware fanatic is one thing.  But what I wonder at is the experience of the average user.  A non-geek could buy a computer with Vista on it and try to install iTunes and fail.  Another user might be banned from ever experiencing Twitter and never be the wiser that the problem comes from the vagaries of one machine.

The latest example concerns my daughters’ computers.  I bought identical parts for them to make my life easier.  The free Nero program that came with the DVD drives works on one of the computers but not the other.  I have reinstalled windows twice in trying to resolve this.  It simply won’t run.  Yet, it runs fine on the other one.  I have a workaround.  Sonic works fine on the machine that hates Nero.  So I use Sonic.  It’s no big deal, but it bugs me.  Again it is supposedly identical computers behaving differently.

Digital isn’t supposed to be like this.  This is analog behavior.  The OP from Slashdot was about how computers and cores really aren’t the same from machine to machine. Each box develops its own idiosyncrasies.  I’m frankly amazed that computers work at all given this divergence.

But from an AI perspective and from the genetic algorithm perspective, this is crazy.  You’d have to develop using several different boxes simultaneously to allow for the divergence.

I was talking about this with Jason, AI researcher and the creator of Underworld Hockey Club (and also a Friday Night Party Line panelist) and I thought his comments were insightful:

“[O]ne interesting thing with genetic algorithms is that they learn with the computer.  If I train a checkers player on the cluster in the lab, that player will not be as good when I run it on my own computer.  It’s not stupid, but it’s not as good.  It’s because it’s tuned to the specific compiler & floating point operations of the cluster.  This is why we need online learning.  Genetic algorithms simulate evolution, but there aren’t any mature methods to simulate learning during the life of an individual.  A common belief that I share is that evolution contains 99% of the knowledge we need to survive, but without the 1% from learning, it is completely useless.  Look at deer, for example.  Baby deer (and other quadrupeds) are able to walk within minutes of being born.  Clearly, the basic constructs of coordination are tied to DNA (baby deer don’t flail stupidly until they figure it out), but the deer need learning to make sure that their programming works with their specific bodies.  One student in the lab is looking at neuroplasticity, which is the study of neural networks that can adapt to their environment in realtime.  It’s really interesting stuff.”

Advertisements

Weizenbaum and Artificial Intelligence.

March 15, 2008

In the Wall Street Journal of all places, I read that Joseph Weizenbaum had died. He created the ELIZA computer program that simulates human interaction. While not a sophisticated program, it is always mentioned (and always will be mentioned) in discussions of artificial intelligence and Turing tests. With his simple program, Weizenbaum immortalized himself as an AI pioneer.

Even though artificial intelligence has its critics, it is already ubiquitous. Just today, I called UPS and spoke at length with a computer. It used voice recognition technology quite effectively to identify my package number. Even now, I am writing this with NaturallySpeaking 9.5. This voice recognition program is inherently based on artificial intelligence algorithms.

But artificial intelligence is not voice recognition alone. AI “perceives its environment and takes actions which maximize its chances of success.” Weizenbaum, clearly a gifted man, gave up computer programming and the field of artificial intelligence altogether later in life. In a way, he was better off than two recent leading authorities in the field. However, based on what I’ve read about him and his work, he really felt that humans shouldn’t rely on machines for decision-making. Of course, now, we do that every day. Pilots use AI to fly airplanes. People rely on AI in their cars without even knowing about it. At some point, people will rely on AI to make decisions about their lives. In one’s PDA, one will have a virtual psychiatrist/business planner/personal coach always at one’s fingertips. I suspect that Weizenbaum would not approve of this, but I think it’s better than, say, relying on Astrology for that same advice.

Weizenbaum was particularly put off by the fact that when ELIZA came out, people really took it seriously. Some people really couldn’t distinguish a simple pattern recognizer from a human being. If you have played with ELIZA, you may find this hard to believe. But remember it came out in 1966. No one had much experience with such things back then. No one had much experience with computers at all. This is how far ahead of his time Weizenbaum was.

At the same time, I think his dismay at how stupid people can be was misplaced. Artificial intelligence, in the form of an interactive program designed to pass Turing tests has not progressed much in the last 42 years. That is not to say that artificial intelligence has not progressed much, it has. However people have not built an interactive program designed to mimic humans with much more efficacy than ELIZA does. There have been some recent attempts and perhaps this is now becoming vogue once again.

Instead of worrying about people who can’t distinguish a computer program from a real person, think about how much these people can be helped. Look how many of the videos on YouTube are made by people who cannot see the consequences of their actions. A quick check with one’s personal digital mentor, might prevent the disastrous outcome from the typical “hey y’all watch this” YouTube adventure. Or Weizenbaum, who was clearly smarter than I am, could be right and such AI development could lead to a Terminator-style apocalypse. I suspect, however, no matter the outcome, such AI is inevitable in time.

Ultra Hal Assistant and Microsoft Speech Recognition.

January 24, 2008

I’ve been messing around with Microsoft Speech Recognition in conjunction with Ultra Hal Assistant.  At some point in time, talking to your computer will be a great way to get things done.  That point is not now.  I have to give the folks who make Ultra Hal some credit for trying something that is essentially impossible.  However, I grew frustrated with the assistant almost immediately.

You can use Microsoft Speech Recognition to enter voice commands in the text box.  It doesn’t work very well.  I’m also playing with Dragon Dictate right now and that is a usable program.  MSR is nowhere near Dragon Dictate in terms of accuracy.

I abandoned the speech recognition part and just typed in the chat box.  I did get it to launch a program, but the rest of it was just inane computer banter.  I think a person would have to be high to get any enjoyment out of this.  In terms of usefulness, it’s far easier to click icons.  This program is a start, but it’s going to have to get a lot better before I’d try to use it again.

Right now, it’s hard for me to imagine that programs like this will be commonplace anytime soon.  Then again, when windows first came out, I thought it was just a gimmick and that the command line was far more useful.