Machine Learning

This morning in my CS 478 class, we talked about artificial intelligence. It was a very stimulating discussion. Probably the highlight of the semester. We came up with a couple of interesting ideas:

What is intelligence? There is something unique to humans (and perhaps other life) that defines intelligence. However I am not convinced that life itself is the definition of intelligence. Can intelligence be introduced into computational systems simply by incorporating biological medium into the system? Although machines have been built with rat brain components (http://www.sciencedaily.com/releases/2004/10/041022104658.htm) that does not necessarily make the system intelligent. Even if the brain tissue introduced a sense of spontaneity to the system, artificial random number generators are a zillion times better at producing an even distribution than any biological matter which is prone to scarring and bias. Perhaps intelligence has to do more with creativity or adaptation.

One of the biggest problems with artificial intelligence is that computer systems as we now know them, even the most advanced impossible computer (the turing machine) only respond to input. 50 years ago, we would have thought that airplanes that fly themselves would define intelligent. Now, however, we see them as machines that take input and do what we tell them to do. To a certain extent, biology also only responds to input, and is therefore no better than computers. My friend Mark is working on designing a set of stimuli that can make a set of neurons grow left or right at a junction. There is no intelligence in the neurons at all. They only respond to input. However, humans use the neurons to do more than to simply respond to input.

Singluarity (http://en.wikipedia.org/wiki/Technological_singularity) is an interesting topic that predicts that if we can design computers to design smarter computers, they will eventually take over the human race. Yeah, it sounds like it Science Fiction. And let's hope it stays that way. Apropo, Google, Nasa, and others have teamed up to create a "Singularity University" to create disruptive technology (see http://singularityu.org/). However, I think that the biggest problem with singularity is an understanding of the problems above.
Artificial intelligence or can be simulated on a computer by giving it an objective function. The computer explores the outcomes using an iterative or optimized pattern, deciding on the choice that will give the best outcome based on the objective function. However, this system is not really intelligent because it only responds to input. As computer scientists, we can give computers objectives in the form of an objective function, but it is impossible for computers to shoot for anything higher than the objective function, so the objective function we give machines is also their limit. I can imagine that a machine could randomly create a new objective function, but how are they to weight the value of the objective function?

As humans, there is some overarching objective function (defined by God or by the natural laws of existence). We try to model the objective function as humans, and base our actions off of our model. For example, we may discover that playing WOW brings us pleasure. We might build our objective function based on the model that we should do everything we can do to spend every minute on WOW. As we learn and discover more about the overarching objective function, we change our model to reflect our understanding. When we discover that playing WOW is not making us happy, we might develop a new personal learning objective that would incorporate sleeping and eating into the model that was previously only based on playing WOW. This iteratively progresses, and we get better as we learn.

Interesting thought.. Perhaps this is the reason why the example of Christ is seemingly impossible to obtain. The overarching objective function is a comparison between us and Christ. We are to become "even as He is." If there were any lower standard, we would be limited by that standard. However, because the standard Christ has set is infinitely higher than we could ever hope to reach, we are unlimited in our potential.

And that's what sets us apart from computers.

The end.

Abandon All Hope

There are some very rewarding parts of being a CS student. Like the privilege of taking classes from Dr. M who can smell good dry-erase markers. Instead of trying the markers out on the board, he can simply take a whiff, and instantly he knows if they are good or bad. That's cool.

However, it's also a lot of work. This weekend, I spent a disproportionate amount of time writing a kernel priority-based task scheduling algorithm. In C. Bleh. When I wasn't doing that, I was sleeping, cleaning the apartment for spring cleaning checks, or working on my side project -- virtual memory implementation. In C. Bleh too.

Good news, however. I got into MIT. Now I have to decide between UCSD, OSU, and the MIT/Harvard MEMP program.

How to Have a Successful All-Nighter

It's 12:35, and I am just now realizing that I am about to embark on the second all-nighter in a week. I have also realized that there are probably a lot of struggling people out there who might not know how to pull off a successful all-nighter. So here's my little handy-dandy guide to a successful all-nighter.

Step One: Accrue Tons of Homework and Projects. All-nighters are hard, but they're easier if you have a motivation to stay up. Sometimes it helps to find pictures of your professors to whom you owe homework, and to tape them to your monitor, or fingers if you aren't using a computer for your homework. In a pinch, I guess a couple of good movies would do for a less-productive all-nighter.

Step Two: the Sugar Hop. I avoid caffeine like the plague. Partially because I get belly aches when I drink it. Instead, I prefer the more natural alternative: sugar. Not only does sugar raise your mental capacities and make the evening more enjoyable, it also helps you keep your eyes open. My favorite source of sugar is candy or soda.

Step Three: Get Comfy. But not too comfy. Change into some good all-nighter clothes, like sweats or shorts and a t-shirt. Avid all-nighter experts insist that proper all-nighter gear is essential for a successful all-nighter. Avoid wearing pajamas, or anything that might be embarrassing if your roommates caught you in it the next morning when they are waking up for school.

Step Four: Sustenance. I personally believe that the biggest cause of premature all-nighter abortion is lack of sustenance. It is vital to give your body the nourishment it needs to keep on going the whole night long. Novice all-nighterers should consider breakfast fare such as bacon and eggs, which alerts the body that it needs to be awake for the next while. I have personally established a standard all-nigher cuisine: Western Family Mac and Cheese from a box. Not only is it only 30 cents (on sale) at the store, but it is easy to make, and has noodles, so it's healthy.

Step Five: Musik. You gotta have the tunes, or the ticking of the clock will drive you crazy.. eventually even to bed. Don't listen to your momma's music. That'll put you right where she is - in bed. Listen to the music of champions - like The Used. Or your favorite other music that'll keep you awake AND put hair on your chest.. rock on! However, be courteous to roommates and wear headphones.

Step Six: Be Healthy. Granted, the 40-hour day isn't the best thing for your immune system, but try to take precautions against anything that will have an effect on your body for longer than a week. Don't lick spoons that are sitting in the jello your roommate with mono is eating. Make sure that you are consuming some vitamin c if your sugar source doesn't contain vitamin c. Do some exercise... pushups every hour on the hour is a great way to keep up your stamina and build some muscle. Gimme Twenty!