May 18, 2015

Book review: The Quest for Artificial Intelligence

This weekend I've watched two movies about smart robots. The first was called Chappie and it tells the story of a robot named Chappie, who is the first robot with the ability to think and feel for himself. Chappie is therefore much smarter than any other human. The second movie, Ex Machina, is also about a smart robot with the same abilities as Chappie, although the robot in Ex Machina is looking more like a real human.
The question is how realistic are those movies? Will we soon see robots that are as smart as Chappie? To be able to answer that question, we have to take a look at the history of artificial intelligence, or AI. And that's why I read the book The Quest for Artificial Intelligence by Nils Nilsson.

The Quest for Artificial Intelligence was published in 2010, and is therefore up to date with the latest histories of artificial intelligence. But AI is a field that these days is moving fast, so some of the latest achievements are not included, such as the algorithm that learned how to play the game Breakout.
Nils Nilsson knows what he is talking about. Among other projects, he has developed the famous A* search algorithm. If you have ever played a computer game where the characters are controlled by the computer, the game is probably using the A* algorithm to help the characters find their way around the map.
The early history of AI begins hundreds of years ago when the old Greeks dreamed about self-propelled chairs. Then it continues with Leonardo da Vinci, who in 1495 sketched designs for a humanoid robot in the form of a medieval knight, and ends with a mechanical duck.
But the real history of true artificial intelligence begins after the Second World War, with a series of meetings. At these meetings, researchers described early attempts to highlight features in images and how to program a computer to play chess. The first meeting was called "Session on Learning Machines," so the word artificial intelligence was not yet invented, until someone suggested the word. Everyone were not convinced, but then most researchers began to use the name artificial intelligence. 
"So cherish the name artificial intelligence. It is a good name. Like all names of scientific fields, it will grow to become exactly what its field comes to mean."
What happened after these early meetings was that the quest for artificial intelligence began at the same times as the computers improved. But everything didn't go smoothly. Nils Nilsson has named the downs in the quest for artificial intelligence "AI winters." What happened was that governments around the world sponsored researchers to develop AI algorithms (generally for military purposes). But since AI is a difficult topic, these algorithms didn't always work.
When they didn't work, the governments decided to decrease the funding, and the researchers had to endure an AI winter with little or no money. Then the computers and algorithms improved, the governments were yet again excited, then the algorithms didn't work as promised, and another AI winter happened.
According to Nils Nilsson, what these AI winters led to was scared researchers. The naysayers around the researchers could give comments like:
"Most people working on speech recognition were acting like mad scientists and untrustworthy engineers." 
So the researchers decided to develop simple algorithms that would actually work as promised, and ignore the more complicated algorithms that didn't always work as promised. For example, the transcription of spoken sentences to their textual equivalents is now largely a solved problem. But that is not true intelligence. The computer can't still understand natural language speech (or text) so someone can have a dialog with a computer, like in the movie Her. The latter is generally called strong AI, while the former is called weak AI. So most researchers have throughout history focused on weak AI to not lose any respect. They were saying:
"AI used to be criticized for its flossiness. Now that we have made solid progress, lets us not risk losing our respectability,"
So if you are wondering why artificial intelligence is not as intelligent as in the movies Chappie, Ex Machina, and Her, the answer is that most researchers have focused on weak AI. The emphasis has been on using AI to help humans rather than to replace them. Yes, a computer can beat a human in a game of chess, but that computer is not intelligent. Deep Blue is considered to be the machine that's the best chess player in the world, but Deep Blue doesn't know that it is playing chess. 
"Does Deep Blue use artificial intelligence? The short answer is no. Earlier computer designs that tried to mimic human thinking weren't very good at it. No formula exists for intuition. ...Deep Blue relies more on computational power and a simpler search and evaluation function."
One idea here is to actually have computer chess tournaments that will admit programs only with severe limits on computation. This would concentrate attention on scientific advances. 
The last chapter in the book is called "The Quest Continues," so artificial intelligence is still far away from being a solved problem. We may have algorithms that know how to drive a car, how to paint a painting indistinguishable from true art, and how to compose music. But we are still far away from robots like Chappie.
This short text is far away from being a summary of the book The Quest for Artificial Intelligence. The 700 pages are filled with facts and anecdotes. But don't worry, even though you will need math to develop AI algorithms, the book includes some math to explain a few algorithms, but it is more a history book than a math book. So if you are interested in the history of AI, then you should read it.

No comments:

Post a Comment