Artificial Intelligence is all over pop culture.
It’s been the subject of science fiction for decades, Isaac Asimov’s many books, HAL 9000 in 2001: A Space Odyssey, and more recently, Ava in Ex Machina.
But what is artificial intelligence?
Is it a machine’s ability to think? A sort of consciousness? What even is consciousness?
These aren’t easy questions — and philosophy rears its head a lot in AI development — but scientists have some ideas.
A commonly accepted way of loosely defining AI is “any man-made device exhibiting intelligent behavior,” according to Dr. Barenholtz of the Machine Perception and Cognitive Robotics lab at Florida Atlantic University. Again, intelligence is another one of the open-ended questions like consciousness.
Some people think calculators are smart or intelligent, but they can’t learn or function outside of their pre-programmed purpose.
This leads to one train of thinking in the AI community, that the key to artificial intelligence lies in creating a machine that can learn and change its own function. Not needing to be pre-programmed is a large basis for artificial intelligence.
If it could learn from trial and error, than there would be nothing to program anymore, and it would grow and be intelligent because it would be like us. Already this logic is being applied by researchers such as Google with Deepmind.
Deepmind can beat many Atari games by trial and error and being rewarded when it wins. Breakout was one prominent example, but it mastered 48 total according to The Verge. One programmer Tom Murphy, has an A.I. that similarly beats Nintendo games, one being Tetris. The interesting part being that computer actually pauses the game once it is about to lose, and never unpauses. As Murphy says, “The only winning move is not to play.”
Another train of thought of what constitutes an A.I. says that simply fooling a person into believing they are talking to another person is enough to be considered AI. This idea became famous from the Turing Test.
Basically a person is supposed to message someone or something through a computer and try to tell if it’s a computer or person they are talking to. If the computer fooled them into thinking they were talking to a human, than the computer is artificially intelligent.
The counter-argument, because there always is one, is that a program can manipulate symbols to follow patterns to fool a person and still not have an understanding of what it is doing. And therefore not be intelligent.
An example to solidify this counter-argument is the chinese room idea. If a person who has no knowledge of chinese and is given a question in chinese, he or she could make a correlation with english characters and chinese ones to make response in chinese to the giver of question without any actual knowledge of the language.
Cleverbot is an example of a this concept. The machine is not intelligent, it simply learns from human conversation what words and groups of words go with others, giving the illusion of conversation.
A.I. has many leaps and bounds to go before becoming a reality rooted in society, even as computers become smarter, people’s standards for them become greater as well. But, until we can figure out how we as humans think and have consciousness, we won’t be able to recreate it.