We have seen massive progress in the field of Artificial Intelligence (AI) over last few years.
By harnessing advanced machine learning capability using neural networks, We have been successful in replicating the power of a human brain.
Deep Blue computer developed by IBM turned heads when it defeated world champion Garry Kasparov in 1st game of a six match series. Human brain eventually won as Garry won the series (4-2) but this was the first step towards realising the advent of AI.
In October 2015, A computer program developed by Google’s Deepmind defeated the champion Lee Sedol (4-1) in 1 of the most difficult board games called “Go”.
This was a shot in the arm for AI as “Go” is an extremely difficult game and requires complex strategising skills. Just to put in numbers, there are more possible moves in a game of “Go” then atoms in the universe!!! This is massive number. Number of legal positions to choose from are approximately 200000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 (2 followed by 170 “0”).
The program was trained with 30 million positions played by human experts. This input laid the “smartness” proficiency of the program. Using this information, the program was able to evaluate and assess based on best games played by experts.
Another huge breakthrough came in August, 2017 when a program developed by Open AI (Cofounded by Elon Musk) defeated 1 of the most celebrated DOTA game player in 1 on 1 game combat.
This was path breaking because unlike “Alpha Go”, this program did not observe or train itself on human played games, Rather, it played against copy of itself in order to train to an optimal level. In matter of a weekend it was able hone its skills to a pro level and defeat DOTA champion Dendi.
In October, 2017, Deepmind released upgrade to its “Alpha Go” program named “Alpha Go Zero” which had self learning ability. It easily defeated the previous version of program that had trained itself based on human game play observation.
Henceforth, a machine program is mature and capable enough to outsmart a human brain without any intervention or dependence on them. These are exciting times in the field of AI and self driving autonomous car is 1 of the symbolic practical applications we see today thanks to Tesla.