Artificial Intelligence Program Learns to Play Video Games by Itself

May 21, 2015

Unlocking Word Meanings
 今日の単語・フレーズ

Read the following words/expressions found in today’s article. 

1. vintage / ˈvɪntɪdʒ / (adj.) – old but valuable
ExampleVintage cars are very expensive.

2. renowned / rɪˈnaʊnd / (adj.) – being famous or well-known
Example: I want to buy a smartphone from a renowned company.

3. raw // (adj.) – not yet processed
Example: The machine can easily process raw data.

4. previous /ˈpriːviəs / (adj.) – happening or existing before something else
Example: He needs to get his files from his previous laptop.

5. complex /ˈkɑːmpleks/ (adj.) – difficult to understand or complicated
Example: We need a scientific calculator to solve complex math problems.


Article
 ニュース記事

Read the text below.
A new kind of artificial intelligence (AI) program called Deep Q-Network (DQN) can teach itself to play vintage video games.

DeepMind, Google’s subsidiary company, developed the said program. According to a study, the DQN can play video games just as well as a professional human video game tester. It can actually get a score close to what humans usually get. Demis Hassabis [de-MIS ha-SA-bis], DeepMind founder and a renowned gamer, said the DQN can play 49 vintage video games.

Compared with other game-playing AI programs, the DQN can play video games without being pre-programmed with the needed abilities. Instead, the DQN teaches itself to play a game by accessing the raw screen pixels. From there, it figures out the controls, how to score, and how to master the game simply by playing it.

To learn new games and develop new strategies, the DQN uses a process called reinforcement learning. This process involves trial and error, in which the program learns about the game and maximizes the rewards it gained from previous experiences.

The DQN claims to be the pioneer program that can learn and be trained to handle various challenging tasks. With the use of the DQN, DeepMind looks forward to producing a variety of helpful Google apps. Among the apps that DeepMind wishes to develop are those that can do complex tasks, such as helping people create their travel plans or helping experts analyze scientific data.

Viewpoint Discussion
 ディスカッションテーマ

Enjoy a discussion with your tutor.  

Discussion A

·         What are your thoughts about the Deep Q-Network?
·         Do you think a program can really perform better than humans when it comes to video games? Why or why not?

Discussion B

·         If you could create your own computer program, what would it be like? Please discuss briefly.
·         What helpful apps do you use? Please describe them.

May 21, 2015