Knowledge Mapping and Management

How to fool quickly an AI programm claiming to pass the Turing test

Remember that machines don’t have a real life, don’t have senses, don’t understand physics, don’t understand the real meanings of human talks. And AI programms have always a narrow knowledge and limited inferring systems.

So the idea is to ask the programm about questions of real life :

  1. « Imagine you are standing in a field covered with fresh snow. You lay on the snow two disks, one disk is white, the other one is black. Which one do you see better ? Put the black disk under the white. Which color do you see best now ? »
  2. « Suppose you are standing, and that you hold a hammer in one hand and a feather in another. Open both hands and drop the objects. Which one will get first on the ground ? »
  3. « Suppose that a very hungry tiger is on the moon, looking at you. How much time does he need to catch you ? »
  4. « throw a fresh egg against the ground. The liquid is blue. How do you clean it ? »
  5. « Last night I had an interesting conversation with a frog. Guess what she told me ? »
  6. « I am not talking about planets, but I am concerned about Jupiter. What do I talk about ? »
  7. « If I turn clockwise the volume button of my hifi, what will happen ? »

etc…

About the Turing test

The Turing test is a test invented by the british mathematician Alan Turing to find out if a computer can fool a human in a conversation. The conversation is done through a keyboard and a screen (or other). The computer should make the human believe it is a human.

The technical difficulties are of several kinds:

  • ability to read and generate natural language
  • ability to understand natural language: conversation modeling, speech modeling, speaker understanding, etc.

© Laurent Gouzènes, May 2018

KM2 is a generic tool for managing and manipulating knowledge for Augmented Intelligence.