Artificial Intelligence

Artificial intelligence is the usage of transistors in a microprocessor to mimic the actions of neurons in a human brain. According to ComputerWorld, “artificial intelligence is a sub-field of computer science. Its goal is to enable the development of computers that are able to do things normally done by people -- in particular, things associated with people acting intelligently… any program can be considered AI if it does something that we would normally think of as intelligent in humans.” Over time, as the concept of artificial intelligence has matured, several sub-categories of AI have developed. These include general and narrow AI, and within each of those, strong AI, weak AI, and hybrid AI.

General artificial intelligence systems are those which are intended to perfectly and completely simulate human reasoning on any particular topic or task. Think “JARVIS” from the Iron Man movies or “HAL” from 2001: A Space Odyssey. Narrow artificial intelligence systems include those which are deigned to intelligently and efficiently carry out a specific task or train of reasoning. Such systems include Google’s AlphaGo and IBM’s DeepBlue, both of which were designed to carry out specific tasks (in both cases, board games) very well. Each form of AI can be implemented through strong, weak, and hybrid methods. Strong AI is a system designed to perfectly mimic the firing of neurons in the brain. A strong AI system, when the first one is built, will theoretically be a perfect replica of a human brain. Weak AI is a system designed to just get the task done, no matter whether a human-style pattern of reasoning is used. In between these two forms is hybrid AI, where the exact methods of human reasoning inspire but do not totally inform the methods of reasoning used by the computer.

AlphaGo, Deep Blue, and Watson are all proof of the potential AI has to become a permanent fixture of the world of the future. AlphaGo and Deep Blue are very effective implementations of narrow artificial intelligence. As The Atlantic points out, AlphaGo is able to “improve—and it is always improving, playing itself millions of times, incrementally revising its algorithms based on which sequences of play result in a higher win percentage.” Because AlphaGo is able to constantly improve its own algorithms, it is intelligent in a way that a static computer program could never be. By continually improving itself, it mimics very well the way in which humans practice sports and study for tests in an effort to improve their own algorithms. Watson is the first impressive implementation of general hybrid AI. While it does not come close to the level of JARVIS or HAL, it can perform a wide variety of logical and intuitive tasks very well. General artificial intelligence systems are currently very good at logic and computation. The key breakthroughs will come when such systems acquire intuition, as sense of morality, and the desire for self-preservation (the scary one!). Once general AI takes on these characteristics, it will be able to rival the power of the human brain.

The Turing Test is a good indicator for narrow AI systems, where the test can be adapted rather well to the specific task the AI system is meant to carry out. However, when it comes to general AI, the test doesn’t hold up as well simply because it cannot test enough variables to accurately determine intelligence. Since perfect general AI will work just like a human mind, it would follow that general AI should be able to beat a Turing Test every time. Once we reach the point where biological and electronic computers become indistinguishable, or perhaps even inseparable, we will have come to the singularity. Ethically, there is no problem with the singularity in general. On an individual basis, certain computers are bound to act unethically, just as certain people are bound to act unethically. Such a dynamic is necessary for the proper functioning of society.