I have a bit of a fascination with a great many academic topics but two of my favourites are Artificial Intelligence (AI) and Psychology. In particular, I am especially interested in Artificial General Intelligence (AGI) or Strong AI. One of the things that frustrates me about many AI texts is that they aim too low - they simply want to design a system that is capable of going about a particular task in an "intelligent" manner - flexibly with little to no human assistance. This is a laudable aim in and of itself but in my opinion offers no insight into the greater Strong AI problem; the reason is simple: Strong AI isn't just able to go about a single or even multiple tasks in an intelligent manner, it isn't constrained by the idea of a task: the most important aspect of Strong AI is that it itself decides what its task is.
Yes, Picasa has an eerily accurate facial image recognition algorithm; yes, computers are more and more often exhibiting behavior that even a few months before seemed to exist only in the realms of science fiction. Kurzweil predicts the singularity from a perspective of computational power where computers will exceed the abilities of human minds. I'd argue that whilst this computing power would enable such a capability to exist, without the appropriate algorithms that can redefine themselves sufficiently, such a singularity will not occur. After all, in many fields of human endeavor computers already outstrip humans: first this was in highly organised repetitive tasks such as addition, etc - it has since extended to code compilation - complex algorithms for computers which execute much faster than the equivalent human process. More recently the power of statistical and pattern recognition systems has allowed the public a view of the forefront of algorithm development - however, despite the seemingly "magic" ability of these algorithms and their ability to learn information, they have made little inroad into the depths of Strong AI. Why?
First, they are all only able to perform a task defined by another - increasingly complex tasks, naturally - but the task is still defined by a human: the algorithm designer. Some algorithms are able to adapt and change their own behavior but these are still performed within the bounds defined by the programmers' desire and imagination (the more limited of the two). Some curios of self defining systems have been created at a small scale in the lab but these lack direction of drive... after all, what drives us but ambition?
So, what do I want to do with this blog (aside from produce possibly random drivel)? I would like to explore the world of algorithms that define their own ambition in life, that are adaptive and are unconstrained by the imagination of the programmer.
I hope some people decide to come along for the ride.