Artificial Intelligence
I’ve been writing this post on the shuttle to and from work. It has been in the works too long now, so here it is, in complete, but at least started. Hopefully, I’ll address further key points in future posts.
Humans are infinitely smart because we are intelligent. We can apply what we know about the world around us to solve problems that we have not encountered. That is the key to our ability to adapt. That is why we are such a successful species.
Artificially Intelligent entities that are man made cannot solve problems outside of their problem domain, and even if a problem exists with in the problem domain, if the designer of the system did not anticipate for a particular edge case, then the entity will never be able to solve that given problem. In short, the entity isn’t intelligent at all, it is only smart with respect to it’s given domain, and even then, it is certainly less smart than the individual or team that designed it, as they could not possible anticipate for all possible scenarios in a given problem space.
Of course, the Gold Standard for artificial intelligence is not how well an entity solves a problem, but how well it fools people. The Turing Test is that Gold Standard, and has been for over 50 years. The goal of the test is for the given entity to appear human enough for a judge to be unable to distinguish between it and a real human. The rules are strict, and in all cases, the judge interacts with the entity under question remotely via teletype, IRC or chat. Again, we have a defined problem space for the entity to operate in, and as long as the creators of an entity are careful to consider all possible edge cases, the entity can expect to do fairly well within that domain. Of course, the entity is not intelligent, only smart.
There are four things that define true intelligence. A recognition of a superior entity, curiosity, the ability to gather information to satisfy that curiosity, and the ability to use that information to solve problems from a variety of domains.
The first two elements are essential to the idea of self awareness, but they will be discussed individually below.
All humans understand the concept of a superior entity. The idea ranges all the way from the concept of a God or gods, to parents or managers at work. On some level, in all we do, we are accountable to something greater than ourselves. Society holds us accountable for our behavior, managers hold us accountable for our work, spouses and children hold us accountable for our actions within the home. Being accountable to something outside ourselves is critical to understanding what it means to be intelligent, if only because it affects what we do in so many ways. Even the most driven and best intentioned of people sometimes need the extra boost that can be derived from knowing that he or she is accountable in someway to something.
In one sense, this concept of superiority is implied by the mere fact that the artificial entity is acting as designed by it’s creators. However, the artificial entity does not learn from these actions. It does not have the capability to deviate from the directives of it’s creators, and therefore cannot learn from those experiences. In a sense, it’s creators understand that they are the superior entity, but the artificial entity has no knowledge of this and lacks the ability to both discover it, or do anything about it. It cannot be obedient, because it does not know what it means to be obedient. It cannot disobey for the same reason; it does not know what it means to disobey. As an artificial entity, it was designed to operate within parameters defined by the problem space as it’s creators understand it, and it will never operate outside of those bounds, severely limiting it’s potential to learn.
Curiosity is essential to intelligence. It could be said that curiosity is the very definition of intelligence. Humans have yet to create an artificial entity that asks “why”. If an artificial entity does not ask why, it will not be motivated to learn. If there is no desire to understand the world as it perceives it, and challenge those perceptions, it can never expand it’s knowledge base. This may be the single most important trait that an artificial entity can have, if it can be programmed. It is curiosity that leads any entity to ask, “Where did I come from?”, leading it to wonder about a superior being or intellect. It is curiosity that leads an entity to wonder what would happen if it chose to disobey that superior entity. Of course, no artificial entity can be curios, because, as has been discussed, it will have been created to operate with in certain bounds defined by it’s creator, bounds often ultimately defined by a problem space.
Artificially intelligent entities are often given means to interact with their environment. Such devices may include cameras, temperature sensors, humidity sensors, microphones, etc. In all cases each of these sensors has a specific job, and gather information in specific ways. Of course, all this information means nothing to the entity. It only gathers the information because it was programmed to do so. It doesn’t know why, in fact, is incapable of knowing why. It can use the information it gathers only in set, predefined ways, and is not capable of using it beyond how it was designed to use it. There is no real curiosity about the surrounding environment, the entity does not gather information to satiate a curiosity, it only gathers what it has been programmed to gather, and generally discards it when finished with it.
Finally, any information that an artificial entity gathers means nothing to the entity beyond it’s immediate application. It gathers information about the world around it, acts on that information in a predetermined way, and then forgets. If faced with a similar situation again in the future, the entity must regather information, re-evaluate and re-decide. A truly intelligent entity remembers what it has learned and can apply that knowledge to future similar problems, even if that problem exists outside of the originally defined problem space. A truly intelligent entity has the capicity to think, not mearly follow the steps outlined by a given algorithm.