How Close to True Artificial Intelligence?
Matthew Keane
Computers are really amazing devices. They can store, transfer, and manipulate huge amounts of any data faster than you can blink. What would make this better, of course, is if the computer requires less and less input over the years. What if it required absolutely no input? What if instead of us telling it how to analyze a situation, it could do it on its own, like the many famous robots in Star Trek and Star Wars, Data and R2-D2 and C3PO? This is the realm of artificial intelligence, as various programs and robots are learning to process, think, and even feel as humans do. Right now, the computer that is being used to type this could be intelligence since it is catching all the spelling mistakes. This is seen commonly in many science fiction works, and if these are compared to the real world state of current technology, a picture can be drawn to see where these AI's are in terms of development.
Artificial....Intelligence?
When the term "intelligence" is used, it is usually define as how close a robot is thinking like a human [1], or doing what human's consider intelligent. In many ways, this has already been achieved. Pull up any chess game installed on any computer and crank the difficulty up to hard, and you will see an artificial intelligence. It is not as exciting as playing with Wall-e, but many players, including chess champions have a hard time defeating computers. One might argue that chess is a mathematical outcome. Computers, who can store and sift through hundreds of math problems in mere seconds, could predict all the possible moves for chess and thus be nearly unstoppable. It is not so much intelligence as just mathematical prediction, or what Angel Garrido refers to as "monotonic logic" [1]. These mathematical predicaments are not a "general intelligence" which all programmers are striving for.
This breakthrough was surpassed by a robot called Watson, who won a jeopardy tournament against two of the leading champions at the time. Jeopardy is arguable a much more demanding game, since it requires the use of riddles, puns and other puzzles that aren't necessarily based off numbers and algorithms. This comes above the clunky AI in many science fiction works, who become confused when someone mentions a slang word or does something "illogical", as such as Baymax from Big Hero 6 or the Terminator from Terminator 2, who cannot understand why humans cry. But some AI, such as Watson, learn from human interaction just as these robots do. Through this, they can understand entire languages and find answers to new questions, which requires a lot more than simple math logic to do. Angel Garrido calls this "fuzzy logic", and also has a number of different types of logics that an artificial intelligence must use [1]. Through this, talking to a robot may become as seamless as having a conversation with C-3PO.
Best to start this video about the five minute mark. It talks about robots advanced intelligence in the job market and other fields.