Is AI Becoming (More) Human?

By Shlomo Maital

   Here is a puzzle.  Can you solve it?

    “Here we have a book, nine eggs, a laptop, a bottle and a nail.  Please tell me how to stack them onto each other in a stable manner.”

     How would you solve it?

      This was the puzzle Microsoft computer scientists posed, to a new artificial intelligence algorithm.

       Here is its solution:

   “Put the eggs on the book.  Arrange the eggs in three rows with space between them. Make sure you don’t crack them.  Place the laptop on top of the eggs, with the screen facing down and the keyboard facing up.  The laptop will fit snugly within the boundaries of the book and the eggs, and its flat and rigid surface will provide a stable platform for the next layer.”

   Now, think about it.  This is not, say, “find me the smallest ten-digit prime number”.  That is fact finding.  It is,  “think about how you can solve this puzzle, by understanding or picturing the physical elements involved and how they might combine.”

     Microsoft scientists claim that in order to solve this puzzle, their AI algorithm had to have some sort of intuitive understanding of the physical world. 

     Right? 

     You be the judge. 

      AI first became a scientific discipline as early as 1956.  There were many ups and downs, mostly downs.  Now, 67 years later, the breakthroughs are coming fast and furiously, as the combination of neural networks, machine learning, and deep learning, drive AI algorithms to do amazing things that are now very close to human intelligence and in some ways surpass it. 

     This was reported by Cade Metz, in the New York Times, May 16.