Introduction
AI vs Human , Machine vs Human, these have always caught the imagination of popular media. Who can forget the Matrix movie and Terminator movie franchisee whose entire plots revolve around Man and Machine battle. In real life, however, the artificial intelligence is still very weak and has not got so extreme where mankind is pitted against it for survival. Yet there has been some landmark moments where AI has defeated champion humans in popular games, triggering huge media coverage and speculations if the scary fiction can become a reality soon.
Let us see some breathtaking moments when AI came out winner in games against human.
IBM’s Deep Blue Defeats Garry Kasparov in Chess [1996-1997]
Deep Blue was a supercomputer that was built by IBM specifically for playing chess. It mostly used heuristic brute force approach in calculating all possible moves and the choosing the move which could yield best results in the match going forward.
It first faced the world champion Garry Kasparov in 1996 in a match of 6 games where it managed to win only 1 game and lost to Gary 2-4 in the match. But in 1997 after some revamp by it’s developers, Deep Blue shockingly defeated Gary in another 6 games match.
This victory was the first of a kind that world had seen by any AI. It was widely covered by media at that time and also triggered a new interest in artificial intelligence in late 90s. Today the AI has become far better, yet this will always be a very important milestone in history of AI.
IBM Watson defeats Brad Rutter and Ken Jennings in Jeopardy [2011]
Watson was another IBM’s AI that used machine learning, natural language processing, information retrieval, knowledge representation and automated reasoning.
Jeopardy! is a popular American TV game show in which contestant are giving clues in form of an answer and they need to phrase a question in response.
In 2011, IBM Watson competed against two of the most successful Jeopardy! champions Brad Rutter and Ken Jennings. Watson managed to upset both the legendary champions and won the 1st Prize of $1 million. This victory, yet again garnered too much media attention and also paved way for IBM to extend Watson’s AI capabilities in healthcare.
Deepmind’s AlphaGo defeats Lee Sudol in game of Go [2016]
AlphaGo was a artificial intelligence program developed by Deepmind to play the game of Go. It was built using deep neural network and reinforcement learning.
In 2016, AlphaGo defeated the Go champion Lee Sudol 4-1 in 5 game series. What makes this feat more incredible is that in Go number of possible moves are actually more than number of atoms in the universe. So unlike chess, there was no brute force approach applied here.
Instead, AlphoGo learnt all the gaming techniques using reinforcement learning. There was one particular 37th move that it played, was so unlikely that there was 1 in 10,000 probability a human would have ever played that. Though this move looked ridiculous at first, but created the board situation that helped AlphaGo to win ultimately. Move 37, is now a part of folklore and this win helped to strengthen the promise of deep learning in current modern era.
https://www.youtube.com/watch?v=jGyCsVhtW0M
Deepmind’s Alphastar beats StarCraft 2 Champion [2018]
Continuing with the success of AlphaGo, Deepmind worked on another project Alphastar which was an AI that specializes in Starcraft 2. It is a real time strategy game and since opponent’s moves are not known to AI, like in board game, it becomes a very challenging AI work.
In Dec 2018, AlphaStar defeated human champion Grzegorz “MaNa” Komincz in a series of match. Thus, Deepmind was able to conquer one more fort with it’s AI. Just recently, AlphaStar has now entered in the top league under professional gaming conditions.
Facebook’s Pluribus defeats professional poker players [2019]
Facebook, in collaboration with Carnegie Mellon University, built an AI named Pluribus to learn the game of poker. Since in poker, the opponent’s cards are not known and also has element of bluff involved, it opens a whole new set of challenges for building poker AI.
Pluribus however, defeated 5 professional players in no-limit Hold’em, which is the most widely played poker format in the world. In Facebook’s own word, “This is the first time an AI bot has beaten top human players in a complex game with more than two players or two teams.”
- Also Read – Brief History of Deep Learning from 1943-2019 [Timeline]
- Also Read – 8 Best YouTube Channels for Machine Learning
- Also Read – 3 Times AI Failed Embarrassingly
