If you’re going to teach a computer how to play Jeopardy!, how would you do it? Would you teach it all the rules of language that you can think of or find, all the common sense nuances that are inherent in language and idioms and slang terms that come with it? Or would you feed the computer millions of examples and let the computer figure out its own rules? In Smartest Machine on Earth, the programmers of Watson chose the latter. Why? Well, another, similar program utilising the “recipe” technique took six years to program and new rules are still being added to it. Also, when initially using that technique with Watson, when a test was run Watson was only able to answer with a 10% accuracy rating. Machine learning had to be employed. An anecdote to prove the point existed in the form of the program used by the U.S. Postal service to read addresses on letters in order to send them. Each person writes a letter in the alphabet differently. The example used in the documentary being ‘A’. It can be anywhere from a lowercase ‘a’ to an uppercase ‘a’ to an upside-down ‘V’ or a simple triangle. No human can write a rule to tell the computer that a letter is ‘A’, but a human can give the computer millions of examples of the letter ‘A’, and let it create its own rules and find patterns that a human cannot to use to identify the letter ‘A’. Machine learning is an innovative technique, utilised in voice recognition, and even language translation. It’s flexibility is what makes it so valuable, and so usable, as well as being the program that Watson needs.
Alex Wabel takes Machine Learning to a whole new level, creating a program that translate spoken text from one language to another. It’s an innovative new application that can be used with multiple languages, including Chinese and Japanese, notoriously difficult languages to translate due to language nuance created by tonal changes. As Alex takes his application abroad, people are amazed at its abilities, from restaurant workers, to the shopper, and managers. Based in Machine Learning, this new application is revolutionary, and full of potential, able to help facilitate communication between people who speak different languages who otherwise would not be able to understand one another, while broadening the social and business web worldwide, and removing language barriers.
Watson may seem complex, but his program is built upon a foundation of a very basic language: on and off, true and false. This is the foundation of binary and binary code, which is then the basis of the programming languages used to create Watson. Programming language is a reflection of the age-old concept of true and false. If I have three apples and he has four apples, then we have in total seven apples, which is true, we don’t have eight apples, which is false. It’s how we know that nine is greater than two, but negative nine is less than negative two. True and false forms the basis of our perception, which we then transform into hypotheses, then into laws and rules. These laws and rules then become the basis of language, and in the case of Watson it the the programming language that allows him to run.
Many computer and robotics experts are fascinated with the concept of AI. Sci-Fi films depict robots that are able to think for themselves and understand their surroundings, that are capable of feeling emotion. Many films depict robots as threats to the human race, as powerful beings that surpass human power, like HAL of 2001 Space Odyssey. The new advances in AI, especially the program used with Watson, can seem threatening to humans. However, as shown with the anecdote of Deep Blue, a computer whose purpose is solely to play chess, Watson may seem intelligent, but he can only play Jeopardy!. Watching make silly mistakes in Smartest Machine on Earth simply because he doesn’t understand gender perceptions nor care about them, it’s easy to see that Watson isn’t a threat. Computers do not have the human ability of language and object recognition, they have no associative memory. You can give a computer the entire compilation of the holy books of the many religions of Man to read, but the computer can pull no meaning from the words. Computers and Robots aren’t a threat to humanity, they are tools with great potential to aid humanity.
http://www.pbs.org/wgbh/nova/tech/smartest-machine-on-earth.html
No comments:
Post a Comment