Loading...

Loading...

06. Artificial Intelligence

音声の再生について

再生コントロール部分の歯車ボタン より,再生速度を遅め(0.75倍速),通常,速め(1.25倍速)から選択することが可能です。また,「3秒戻る」,「3秒進む」ボタンで再生位置を細かく調整できます。

本文テキスト
  • 06. Artificial Intelligence
  • Computers are designed to follow instructions given by humans.
  • For example, if we want a computer to classify triangles into right triangles, equilateral triangles, etc., we can tell our computer (1) to determine the angles of all the corners, (2) to check if one of the angles is 90 degrees (a right triangle), (3) to check if all the angles are 60 degrees (an equilateral triangle), and so on.
  • Then our computer will be able to classify triangles.
  • There are limitations to this approach, however.
  • To classify triangles, we can come up with a set of clear rules, but there are a lot of cases where we cannot easily do so.
  • Imagine you want your computer to classify images of fruits like apples, oranges, bananas, etc.
  • It is not easy to prepare a set of rules like those for triangles.
  • To handle subtle cases like classifying fruits, scientists developed artificial intelligence.
  • One important type of artificial intelligence is the artificial neural network which was inspired by the structure of the human brain.
  • Our brain is a huge network of neurons.
  • Each neuron's function is simple, but a large number of neurons work together to do complex jobs.
  • An artificial neural network consists of numerous nodes that are connected to each other.
  • Just like neurons, these nodes are designed to work together to perform complex tasks.
  • In a typical artificial neural network system, nodes are arranged in layers.
  • The nodes in each layer are connected to some or all of the nodes in the next lower layer.
  • When a neural network processes information, the nodes in the lowest layer process it first and each of them recognizes a basic piece of information.
  • For example, in a neural network system that handles images, each node in the lowest layer usually recognizes a dot in a particular location of an image.
  • Nodes in higher layers can recognize more complex features like curves and straight lines.
  • Nodes in still higher layers can recognize even more complex features such as the appearance of fruits.
  • Artificial neural network systems have to be trained.
  • For example, if we want to create a system that can classify images of fruits into apples, oranges, and bananas, we need to prepare a large number of sample images of these three kinds of fruits.
  • We also have to tell the system which kind of fruit is shown in each image.
  • Then we have our system study the sample images.
  • The system classifies the images and compares the results with the human classification.
  • If the system's classification is wrong, the system adjusts itself until its classification becomes as close as possible to the human classification.
  • A neural network system adjusts itself by weakening some connections of nodes and strengthening other connections.
  • This way, certain features recognized in a lower layer are emphasized or ignored in a higher layer.
  • As the system makes adjustments by studying samples, it becomes more sensitive to more important aspects and less sensitive to less important aspects of information.
  • The idea of artificial intelligence has proved to be successful.
  • Now computers can do useful jobs without clear rules.
  • A good example is voice recognition.
  • Your smartphone may be able to recognize spoken words.
  • It is also becoming possible for artificial intelligence to interpret X-ray images to provide a diagnosis.
  • Many engineers are now working on ways to use artificial intelligence to make our lives better.