Note: This article appears in full text on researchgate at following link in original: (PDF) 360 degree Test for Intelligence for Machines (researchgate.net)
[ https://www.researchgate.net/publication/366824869_360_degree_Test_for_Intelligence_for_Machines]
Turing Test, is one of the most famous test to declare intelligence of a machine. Turning test had several variations and characterizations to study. The most popular version of Turning test involve a human interrogator and a human and a machine in different rooms, the interrogrator would ask same questions to both human and machine and analyse the answer, in typically 5 minutes time, and declare if he has found out who is machine and who is a human. Firstly, the questions are not explicitly set nor the duration of time can be analyzed to be concrete. One of the first few versions of Turning machine proposed to question man and woman too, as given in reference [1].
The intelligence of a machine with 5 minute interrogation can be changed now as Turnign Test was framed in 1950, its been quite a long time, and many advances in Artificial Intelliegence have been made since then. However, we can never leave what Turing proposed us, just we need to change the incorporation of latest state of art AI concepts in the Turning Test, so that this is a rational competition between man and machine. The machine can access internet so why cant human being ? When the machine can use parallel processing, why cant human use search engines and dictionaries ?
In this article I cover areas that need enhancements that can be added or that are required to be added in a plain Turing Test. These are some of the enhancements that I view. Turning test was a tough test in its time and even today, however, in Turning test, the extent of intelligence or score of intelligence of a machine are not yet covered. It is paramount to know if machine is not intelligent then what is the score of intelligence of the machine. We need to define the intelligence levels after the interrogator evaluations are over. There are ways in which this an be conceptulized, which I shall cover in next article.
Now, the main aim of article is to upgrade Turnign Test, given we are at advent of massive impacts of AI on human civilizations. When AI has grown in 360 degrees why are we interrogating a human and machine only in Natural Language Texts? Had Turing been here he would had enhanced his test to be a 360 degree test himself? Would he? This is a question for you to answer.
Now this 360 degree change in AI, ever since Turning wrote his test to check whether a machine is intelligent or not, needs to be accumulated in the test of intelligence. The earlier version was based on verbal response, but now there are image processing intelligence tasks, signal processing, written, oral tests and even the test in finding a fruit from tree or playing a football match, all these can be done in AR/VR and hence can be simulated by a machine and human both. Why only chess, there are plethora of games machines need to learn to win from human and still being a worker machine for humans.
Here I believe as Turing kept 5 minutes interrogation of a machine and human for verbal interrogation, there should be 5 minutes for computer vision exercise as well, may be detecting things on the screen of computer, given the fact machine would be connected to internet, internet best services be provided to human as well. As we apart form intelligence of verbal arguments shall be measuring who is more intelligent and by what ratio is he/it/she being more intelligent.
Another 5 minutes in test in AR/VR tests, accessing the real time intelligence, walking in AR/VR, counting in ripe mangoes, or going near trees to predict right time to have a bee hive. Other test as like predicting rain, winds or fog, as humans do based on view of sky.
These are some of the 5 minute test in AR/VR which checks the intelligence in walking, swimming, driving to mention a few. Why only chess, why not football ?
Yes, this seem to be like making up of a all rounder Robot, yes when computers can use parallel processing, the intelligent agent we are making, why shall it not use multi cores and multi processing and cloud computing? Given as a fair game of test, human be allowed all these too to win from a robot the game of Intelligence.
References
Artificial Intelligence, Computer Science & Mathematics 