What is Intelligence in AI?

 

Nowadays, the word 'intelligence' can be more associated with machines than with human beings. In this article, let's discuss in detail how this word became more related to machines than people.

 

In 1950, Turing pioneered the idea of machine intelligence, and that was the start of the artificial intelligence concept. Turing was great at thinking in such a different way. At that time, ENIAC computers only performed basic mathematical calculations. Let’s consider, if a non-living thing started to think, how it would communicate with others? If human beings are able to think, we can judge whether someone is correct or not by his response, either through verbal or non-verbal cues. If someone responds correctly to the questions, then we say he is intelligent. If someone wants to respond, we have to ask some questions regarding something. If the other one answers properly then we conclude his thinking is correct else either way. So Turing came up with the famous Turing test to communicate with machines.

 

 

There will be three players. One is called the interrogator and he asks the questions to the other side of the screen. On the other side, two will be there. One is a human being and another one is a machine. If the interrogator fails to identify who is a machine, then the machine is smart enough to win the interrogator. Hence human beings can think. This is what Turing finds and nowadays this concept helps in the AI field to find the difference between robot and human-machine by captcha. Here is the example shown in figure 1.

Fig 1. Sample Alan Turing Test.

 

 

By the Turing Test, we conclude that yes, the machines can think. In 1980, John Searle said just a machine can answer like humans we cannot say it's intelligent enough. A machine can be treated as intelligent only if they understand what they are doing. Without understanding a person or machine could not be as intelligent. In order to prove his statement, he used the Chinese room argument.

 

Let’s consider in a room there is a person sits with a rules ledger. He is not able to speak or nothing about the Chinese language, but he receives a Chinese input and he processes the input with the rules ledger to find out the response and he gives back the Chinese output properly. So the man who gives the input thinks that the person who inside the room is a native Chinese speaker. This way by changing some symbols anyone can pass the Turing test. So Searle concludes a machine cannot be an intelligent one unless it understands the process. Figure 2 has shown the pictorial representation of the Chinese room argument.

 

Fig 2: Chinese Room Argument

 

However, the word intelligence remains in the AI world today. Since we can say the machine is not completely intelligent enough, the AI have two different classifications. One is Strong AI and another one is Weak AI. So far, all the applications in the AI world developed comes under Weak AI since it only has taken the input and did some process to produce the output. Once the computer understands the process of what they do, that becomes Strong AI. In my view, Artificial Intelligence should not replace the human being, but it can leverage and assist human beings to perform better in the years to come.