- InsightsInsights
- CryptocurrenciesCryptocurrencies
- Stocks
- White Papers
- IndustryIndustry
- GeographyGeography
Artificial intelligence is one of the most misunderstood areas of tech and it's no wonder really. There are so many films about artificial intelligence surpassing human intelligence and taking over the world, usually in pretty messy and violent ways. There's something innate in humans that causes us to be suspicious of ultra intelligent things and maybe we should be after all. It turns out that there's some artificial intelligence out there that's so bright it's started outperforming humans. Here are some of AI's most important defeats.
Chess was one of the first things that artificial intelligence ever beat humans at. It was 1976 when this happened for the very first time, but the one that is most reported was twenty years later. Garry Kasparov was considered by most to be the greatest chess player of all time, so IBM, the makers of an artificial intelligence called Deep Blue, wanted to see if their machine could beat him at his own game. From six games, Deep Blue managed to win twice, which might not seem a huge deal, but when you consider than Kasparov was the best chess player in the whole world, even one defeat would have been impressive. Following on from this though, IBM knew that they could do better. They invited Kasparov for a rematch a couple of years later, after they'd made some changes to their programme which was now named Deeper Blue. In this match the game finally beat Kasparov, forcing him to resign in the sixth and final match. Speaking of the match Kasparov was deeply impressed with the machine, stating it was like playing a whole new kind of intelligence.
For most people, learning the rules for Texas holdem isn't too difficult and many will understand after just a few games with a patient teacher, or a little practice online. However, most of us would imagine that a robot could struggle to learn them and certainly we would assume that the best human players in the world would beat a robot each time. Yet, in the case of DeepStack though, this isn't true at all. DeepStack has routinely beaten some of the best poker players in the world and it isn't the only AI to do so. In 2017 an AI called Libratus beat four professional poker players simultaneously in a game of Texas Hold'Em. Not only can an AI not tell if a person is bluffing, but unlike chess, where the AI can see all of the pieces on the board, poker is a game of imperfect information. This means simply that the AI cannot see what is in the other player's hand, meaning it has to work on guesswork. Working on guesswork involves genuine intelligence rather than just clever programming, making this feat even more incredible.
The need for AI is becoming more and more apparent and whilst beating humans at chess and poker is cool, there are some applications that might be a little more meaningful. For example, Microsoft has created speech recognition software that is so effective it can transcribe real time speech with an error rate of less than 1%, compared to the human error rate of around 6%. Not only this but AI can now comprehend better than us too. In a university level comprehension test, Alibaba's AI was asked to read around 10,000 questions and scored 82.44% in a follow up test. Humans scored 82.304% in the same test, which although close, is still lower.
