Friday, November 25, 2011

The trick of knowledge

Computational agents can be considered intelligent to the extent that they utilize human-level knowledge in their behavior. How to do that is the great difficulty. I submit that the trick of knowledge is going beyond mere possession of the facts of knowledge to the ability to know how to apply knowledge. So, if we want to encode knowledge in a form that is useful to computational agents, that encoding must also include an encoding of the knowledge of how to apply that knowledge. Sure, we can hard-wire that latter knowledge, but that may be difficult, error prone, and probably much less flexible or adaptable to evolution of the environment. And even if we are successful at that hard-wiring, that hard-wired knowledge must be properly parameterized to be used in a complex environment.
 
It is worth noting that even the knowledge of how to apply knowledge needs its own knowledge of how to apply that knowhow, and so on seemingly ad infinitum. Clearly at some level there must be hard-wired knowledge. Picking that level is a central challenge, but does highlight the need for a rich knowledge-based infrastructure.
 
In any case, the trick of knowledge is not in what you know, but in your ability to apply that knowledge. Maybe that is the essence of intelligence itself.

Wednesday, November 23, 2011

Truth of statements and truth of existence

There are two forms of truth that we have to deal with:
  1. Truth of existence. Does an object or phenomenon exist in reality?
  2. Truth of statements. Is a statement true or false?
In real life both are equally relevant. We are surrounded by and part of the physical world. The world of language and statements is but a subset of our world, but a very important subset and the subset that is the primary focus of what separates humanity from the rest of the physical world.
 
But inside of a computer, where reality is kept at a distance and almost literally is a separate and distinct universe, truth is concerned mainly with statements and the notion of existence outside the realm of inside of a computer is itself a mere statement. In other words, the computer can know about the real world only to the extent that we seed its pool of statements with statements about the real world that we know to be true. We must define what is true of the real world.
 
If truth of existence means anything inside of a computer, it is simply as statements making assertions about the real, outside world, statements which a computer cannot evaluate per se and whose truth can only be obtained by human or other physical input in statements whose only justification is of the form "because we, agents of the outside world, say it is so."
 
In short, beliefs about the outside world are true inside of a computer only to the extent that we external agents have correctly encoded our external truth into machine-cognizable statements of truth. We must define truth about the outside world. Unfortunately, we may be wrong or may make mistakes when doing such encodings, so there is the risk that a computer may not start out with a true understanding of the outside world.
 
Eventually, ultimately, as we embellish computers with sensors and the ability to directly "learn" from those sensors and human documents and other artifacts, it may be possible for a computer to directly "learn" at least some aspects of truth in the real world. But, once again, truth inside the computer will be limited by pre-programmed assumptions about how sensors and human artifacts work. After all, how can a computer "know" whether our manufactured sensors accurately convey "the truth of the real world" and don't distort this "truth" in some ways of either malevolent or accidental nature, sometimes even despite the best of intentions or maybe because of the worst of intentions.
 
Could we construct the ideal criminal or achieve some ideal sense of evil, either by accident, negligence, or by intention? We may even create evil simply as a test case, but will we be able to control it? Or maybe someone may create evil because they do seek to control it, or maybe someone might create an intentionally uncontrollable evil solely in the pursuit of chaos.
 
In any case, achieving alignment between the truth of the world and the truth of statements within a computer is a tricky business to be sure.