Computers that we can talk with are now less science-fiction and more science-fact. Does that mean computers are now ‘intelligent’? Are they capable of thinking for themselves?
Many people think something called the Turing Test could be used to find out if a machine is capable of ‘thinking.’ To pass the test a machine has to make people think they are talking to a human. John Searle, a philosopher, came up with a thought problem about this called ‘the Chinese room.’ He said it shows the Turing test is a poor test for intelligence.
Imagine you are in a locked room with a letterbox to post messages in and out. You can speak English, but can’t speak or write Chinese at all. On the wall of the room, there are instructions in English showing you how to reply, in Chinese, to any question in Chinese.
So if someone posts you a note saying 你最喜欢的颜色是什么?, you would know to reply with 蓝色 just because the instructions say so. Although you’ve given the person on the other side of the door an answer, you don’t know what the question was, and you don’t know what you’ve answered!
‘你最喜欢的颜色是什么?’ means ‘what is your favourite colour?’ and the reply, ‘蓝色,’ means ‘blue.’ What if your favourite colour is actually red? Because you don’t know Chinese, you don’t know you’re being asked for your favourite colour, and even if you did, you don’t know what the Chinese word for red is. (It’s 红色.)
This is why John Searle thought the Turing Test had a flaw. Even if a computer could hold a conversation with a human, like a human in the Chinese room, it would just be doing so because it was programmed (following instructions) to act that way. It doesn’t necessarily mean that the computer is intelligent, or has a mind of its own, any more than you, in the locked room, could understand Chinese!
Clever critterness… but is it a deception?
More AI stuff →
Go deeper with cs4fn →