Ever since AI became a thing, a common fear was that these machines would one day surpass our capabilities and take over. In 1997 IBM’s supercomputer, Deep Blue beat the World Chess Champion Gary Kasparov at the time. Ever since then, AI has been beating us at what we used to excel at, and started achieving goals that we thought were unachievable. Here are a few of AI’s greatest success stories:
#6 Beating the world champion of Go
Beating humans in Chess was all about calculating possible moves. With the 2500-year-old Chinese game of Go, calculating wasn’t an option, as the game has more possible positions than there are atoms in the universe.
Last year, Google DeepMind’s AI software AlphaGo played world Go champion Lee Sedol. In order to win, AlphaGo had to get good at the game – and it did. By playing itself millions of times, AlphaGo bested Lee Sedol in a game often seen as the “Holy Grail” of AI. Recently, Google’s DeepMind has announced it will train artificial intelligence to take on Blizzard Entertainment’s StarCraft 2.
#5 Dominating Atari games
Old Atari games used to be extremely fun for us humans, but we never really managed to dominate them the way AI recently did. DeepMind – a London-based AI company acquired by Google in 2014 – was recently presented with a few classic Atari games. At first, Google’s DeepMind wasn’t great but soon enough it started to play better than any human ever did.
The following video shows DeepMind playing Atari Breakout. At first, it seems clueless, but it quickly finds weaknesses in the game’s system and ruthlessly exploits them to beat it.
#4 Writing poetry
Although AI can beat humans in incredibly complex games, it still can’t communicate very well. Google’s Deep Mind wanted to tackle the issued and fed its AI more than 11,000 unpublished books. Using the information provided, Google’s AI created a few poems that may not be the best one’s you’ve ever read, but that will certainly blow your mind, taking into account a computer wrote them.
#3 Beating professional poker players
In April 2015, Claudico – an AI bot – played high-stakes Texas Hold’em with professional poker players. It taught itself how to play the game by playing against itself, and even though it didn’t win the tournament, it gave players a hard time. Following Claudico’s footsteps, Liberatus played trillions of hands with itself to be able to play four humans heads-up. After 120.000 hands, Liberatus won a total of $1,766,250 in chips.
Keep in mind Poker is a game of incomplete information, where bluffing can play a huge role. Both AI’s mentioned above managed to bluff successfully.
#2 Developing a scientific theory
Scientists and biologists at Tufts University reportedly programmed an AI that was capable of developing a new scientific theory with no help from humans. Essentially, its theory analyzed data using a trial and error method, and discovered how the regeneration process of a flatworm worked. This was, until then, a 120-year-yold biological mystery.
#1 Beating humans on IQ tests
So far, it’s clear that AI can learn, but some of the hardest tasks it ever faced were analogies, synonyms and antonyms, as one word can have a variety of different meanings. Yet researchers from Microsoft and the University of Science and Technology of China built an AI that could outperform the average human on IQ tests – and that means figuring out analogies and synonyms.
A sample study with 200 humans found that the computer outperformed the average adult, and that it reached the level of people with masters degrees.
If you liked this article, follow us on Twitter @themerklenews and make sure to subscribe to our newsletter to receive the latest bitcoin, cryptocurrency, and technology news.