How to achieve Artificial Intlligence?
How to Achieve Intelligence?
By ANONYO SANYAL
In an earlier post, we saw the Turing test, and how we define intelligence. It's time to dig deeper into achieving intelligence. Here we must be aware of the fact that Turing test actually is an operational definition of intelligence. Operational definition means a definition which is not merely a theory. It speaks of some tasks, you need to perform and some steps or operation to determine the nature of an item. Thus the term operational definition comes hand in hand with the Turing test as it helps us to determine the intelligence(nature) of a computer(item).
PIC.Turing Test Diagram Source:- https://en.wikipedia.org/wiki/Turing_test#/media/File:Turing_test_diagram.png |
Natural Language Processing:-
Natural Language Processing(NLP) deals with the conversational part of a computer. In NLP we focus on how a computer interprets and converses in English or any other human language. The first thing that comes to our mind is Chat Bots(Google Assistant, Siri, Alexa, Cortona etc.). Of course, chatbots are good examples but NLP is not fixed to only chatbots. NLP has plenty of applications. We can process comments about a product and say whether the user was satisfied or unsatisfied about the product. Even we can resolve the issues by using NLP in processing comments like whether it was the packaging or delay in delivery etc. In fact, Google is working on making phone calls using google assistant to make appointments at a barber's shop for an appointment. This has a great deal of application using NLP.
Knowledge Representation and automated reasoning:-
Knowledge representation and automated reasoning are two sides of a coin. Knowledge representation merely means to store information before or during a conversation and automated reasoning is using the stored information to answer questions and to draw new conclusions. If it does so then things like "Remember I kept my wallet on the table". If the system doesn't remember this then next time when we ask it "Where is my wallet?", it won't be able to reply "On the table". Then it won't be intelligent and the interrogator will conclude it can never be a human being. Thus using Knowledge Representation and automated reasoning is a must.
Machine Learning:-
As the name suggests it is about learning. It means to adapt to new circumstances and to detect and extrapolate patterns. In a nutshell, it means the program must be able to adapt itself to unknown situations even if it is not programmed to do so. Remember Knowledge Representation and automated reasoning is not the same as machine learning. A lot of work is being done in this field.
What is Total-Turing Test?
Turing test was designed intentionally to avoid direct physical interactions with the interrogator as he felt it was unnecessary. But if we say that acting is also intelligence then there must be physical interactions through a 'hatch'. There must be a video-signal where the interrogator can test the subject's perpetual activities. This test which includes both physical and mental interrogation is known as Complete/Total Turing test. To achieve Total Turing test the machine must have the following:-
Computer Vision:-
It will help to perceive objects. Say the interrogator picks up an apple and asks "what is this?" then the computer must be able to answer that it is an apple. Thus computer vision is important.
Robotics:-
Robotics as we know deals with moving objects. Say if the interrogator asks to pick up this book then the computer must do so. Thus robotics is also an important part to achieve physical interactions.
Are efforts being taken to pass the Turing Test?
Frankly speaking as of now, not much. Yes, humans are still far away from being successful in the field of AI. So it will take time to pass the test. But there is an award called the Loebner Prize, that one gets to pass a competition based on the Turing test. As of now in 2018 Mitsuku by Steve Worswick is the winner.
Comments
Post a Comment