According to MIT, machine learning is defined as the ability of machines to learn without being directly taught and to develop to accomplish new and unique tasks. Previously, AI required an immense pool of data and strong computing power to produce results. Nonetheless, with today’s significantly advanced computer processing and vast datasets, AI has finally surpassed these technical limits.
Recently, researchers at Berkeley AI Research have unveiled technology that is able to generate original content ranging from changing the season of a landscape photo to realistic human faces. Shattering the preconception of a mundane hunk of metal that could only calculate complex computations, AI has proven itself to be capable of creating authentic-looking images by mimicking a substantial data pool.
With these new advancements, AI has also ventured into areas that were previously untouched by machine intelligence, such as defeating the best humans in many popular games like “Jeopardy,” chess, poker and backgammon.
All of these games have been previously believed to require a human touch of creativity to play since they require players to come up with strategies and deep analysis of the opponent’s moves. The ancient Chinese game of Go, which has a vastly greater number of permutations than a game of chess, is considered one of the most difficult board games to play and impossible for a computer to master. That is until Google’s AlphaGo program was created.
In March 2016, AlphaGo beat world champion Lee Sedol four out of five times, marking a great milestone. Most notably, AlphaGo’s 37th move of Game 2 shocked most go grandmasters since the move was so unorthodox and was initially believed to be a blunder. European go champion Fan Hui said: “It’s not a human move. I’ve never seen a human play this move. So beautiful.” Move 37 was key to AlphaGo’s victory.
By feeding AlphaGo an extensive stream of expert gameplay and setting up various versions of AlphaGo to play itself, AlphaGo was eventually able to deviate from textbook human moves and create its own playstyle with moves, such as move 37, which, according to AlphaGo’s software, only had a one-in-ten-thousand chance of being played by a human.
So can AI be creative? The answer is yes, but with limitations. AI can create new content but does not understand its creations.
According to the New York Times, since truly emotionally impactful art comes from the human imagination, AI is still far from being able to grasp the underlying message of a creative piece. Even though AI can produce art indistinguishable from that of a human, AI generates these art pieces from human data and is still incapable of understanding the meaning of that data.