Top 10 Times Robots Beat Humans at Games
#10: Watson
The first entry on our list features a computer system that was actually created with real life intellectual combat in mind, specifically an appearance on the iconic TV quiz show, "Jeopardy!" IBM’s Watson has grown leaps and bounds since its inception - today working at high levels in the medical field. But its biggest claim to fame occurred in 2011, when it won a famous "Jeopardy!" match against two of the show's strongest champions. During the second game of the match, the humans rallied and seemed to have turned the tide. But Watson fought back, and ultimately came out on top.
#9: Microsoft Speech Transcription
Top 10 Worst Microsoft Products Ever
Ok, so maybe there aren't many multi-million dollar word transcription leagues out there for competitive office types. But that doesn't make this technological breakthrough any less notable! October of 2016 marked the first time when Microsoft's AI program equaled the error ratio of a professional team of human transcription experts. We should mention that the program's success rate was only slightly better than its human counterparts, with the average being more on the level of "human parity" than any sort of blow out. Still, this AI success story served as a breakthrough when it came to improving how tools such as our mobile phones understand us on a daily basis.
#8: DeepMind Atari
Do you remember how it felt when you finally reached that high score at your favorite old school arcade game? Putting your initials on the leaderboard for the entire neighborhood to see? Well, how about if an artificial intelligence program could figure out over a hundred ways to make that high score look like beginner's luck? Enter Google's DeepMind technology, and its absolute dominance over a score of iconic Atari franchises. A 2015 report boasted how DeepMind cleaned up on even the hardest Atari titles, oftentimes performing at a level superior to nearly three-fourths of professional, human game testers. Want a hard figure for proof? DeepMind scored a whopping 999,990 points at "Ms. Pac-Man." Beat that one.
#7: Rubik's Cube Robots
5 Amazing Rubik's Cube Facts
Solving a Rubik’s Cube can take people’s hours. The human record, however, actually stands at just 3.47 seconds, held by China’s Du Yusheng. That’s pretty impressive . . . but the robots definitely have us all beat. In 2016, German engineer Albert Beer’s robot “Sub1 Reloaded” solved a Rubik’s cube in 0.637 seconds. That’s . . . fast. Two years later, even that record was blown away when MIT engineers let their robot strut its stuff on the speedcubing stage. Their “Rubik's Contraption” managed to solve the cube in just 0.38 seconds. Sure, these robots weren’t directly playing against individuals in this case, but it’s safe to say that they far outperformed we mere humans.
#6: Alibaba & Microsoft Reading Comprehension
Call this next entry a case of "close, but no cigar" for humans. In 2018, competing artificial intelligence programs developed by Alibaba and Microsoft both beat out human scores on a Question Answering Dataset by Stanford University. Alibaba’s program scored 82.44, and Microsoft’s 82.65, and compared to the human performance, at 82.304. Sure, it may not seem like a big margin, but both companies had plenty to brag about, given how far AI has come in terms of reading comprehension. The long lasting effects could be similar to Microsoft's speech transcription programs, in assisting computers to answer our basic questions with more clarity and accuracy.
#5: The Northwestern AI Challenge
What If Robots And Humans Were Identical?
If there's any sort of comfort to be gained from these stories, it's that sometimes artificial intelligence can fall into the same mental pitfalls as humans. Case in point: the computation model developed by computer engineers at Northwestern University that was a natural at visual problem solving. In fact, the program was so adept that it aced a group standardized test known as Raven's Progressive Matrices, ending up in the 75th percentile, compared to its human counterparts. The silver lining? That the program did tend to miss similar questions as humans who took the test, although Northwestern engineer Ken Forbus was quoted as being hopeful that future AI would combine the best of visual recognition and reasoning.
#4: Libratus
Ask anyone who's watch the World Series of Poker tournaments on television: this card game can be BIG business. So, it only makes sense that some sort of super computer would be developed to play Texas Hold 'Em games with the best of the best. Enter Libratus: an AI program that was born from the success of Claudico, a similar program that played against human competitors back in 2015. Libratus followed a similar path in 2017, but benefited from its enhanced design and ability to learn from its mistakes. The cream of the poker crop all tried to take down Libratus, and the AI didn't win every hand; but it's lead was absolutely unstoppable, and Libratus eventually came out on top.
#3: AlphaGo
There's something to be said about knowing when you're licked. This was the case with Lee Sedol, one of the world's strongest and most successful competitive Go players in the world. Go is one of those ancient games that's easy to learn, but difficult to master, especially when you're playing against a computer opponent. This was the situation Sedol found himself in back at this 2016 grudge match against DeepMind’s artificial intelligence program named AlphaGo. Sedol did manage to win one game against AlphaGo, but ultimately lost the best of five contest against the program. He would announce his retirement in 2019, stating, in part that AI influence in the game "could not be defeated."
#2: The OpenAI Massacre
Top 10 Times the Olympics Bankrupted a City
Ok, so we've seen artificial intelligence bots take on the world of Atari and ancient board games. But what about modern day esports tournaments? Well, the OpenAI Five project answered that question with their final public demonstration in 2019 when they absolutely slaughtered a human team on the battlefield of DOTA 2, also known as Defense of the Ancients. The human experts never had a chance, as OpenAI won match after match with relative ease. So definitive was the program's win that the development team decided to change their focus on AI/human cooperation, presumably knowing full well that no real competition lay out there for their victorious bot-monsters.
#1: Deep Blue
25 Bizarre Discoveries In The Deep Sea | Unveiled XL
It's arguably the most famous instance of man versus machine of all time, a historical moment that hinted at the vast potential of AI. Deep Blue was developed by IBM to play chess with the best, eventually challenging Russian chess master Garry Kasparov to a series of matches in 1996. Deep Blue would only win one of this first best-of-six challenge, but a rematch a year later saw the program definitively defeat Kasparov. The human champion was so upset at the loss that he accused Deep Blue of cheating, but the proof was there in the pudding, and marked the first time when a computer would defeat a human chess champion in a tournament setting.
