5 artificial intelligence wins against humans

Humans have invented games and sports over the centuries that have become part of the culture of many nations. What few people imagined is that, long after, even the great champions would be overcome by machines.

Next, our colleagues at Tecmundo - Mega Curioso's "big brother" - selected robot victories, pre-programmed computers or advanced artificial intelligence against the human race, which for many pessimists and apocalyptics is just the beginning of the revolution. of the machines. Check out:

1) Chess

The dispute between IBM's Deep Blue computer and Russian chess player Garry Kasparov is a milestone. It was the first time that it was clear that computers could surpass even the mind of a human champion in one mode. In fact, Kasparov won the first best of five in 1996. But he was defeated in the second race a year later by an already updated version of Deep Blue.

The Russian accused IBM of cheating and using human interference in moves considered creative and spontaneous, but the company denied the accusations and retired Deep Blue before a third round of matches. Deep Blue did not use artificial intelligence, but rather a brute force calculation method that analyzes the possibilities, predicts the opponent's responses and only then suggests the best move.

2) Go

But the most felt defeat came in 2016. The AlphaGo program, which was developed by Google's affiliate DeepMind, defeated Go World Champion South Korea Lee Sedol. There were four wins in five matches, a milestone that hurt even the pride of the local population.

The challenge here was much greater. Sedol was considered unbeatable and the Go is extremely complex, requiring not only calculations and probability. Unlike DeepBlue, AlphaGo is a complete artificial intelligence, self-taught and very trained against machines, other professional players and even against itself.

3) Jeopardy!

IBM again defied humanity with Watson, which also beat the creators themselves. In 2011, the supercomputer defeated Ken Jennins and Brad Rutter, two of Jeopardy's biggest competitors in history. This is that gameshow that has been playing on United States TV since the 1960s and involves general knowledge in an original formula: the participant needs to formulate the question that hits the answer provided.

Watson won the competition in both editions and IBM donated the $ 1 million prize for charity. The victory was important not only to show a machine beating two very different humans at the same time. It also consolidated Watson, which was still a little unknown to the general public, as a platform that is now very well used in health, education and many other areas.

4) Poker

Do robots know how to make poker face? Libratus artificial intelligence, programmed by researchers at Carnegie Mellon University, proved so. In 2017, she won a Texas Hold'em tournament against a host of human contestants entitled to broadcast on Twitch.

Libratus's trump card was not just the calculation, but the processing of each match to learn from mistakes and not falter again. Winning poker is even more incredible because it does not involve viewing an entire board, but fewer cards and the ability to deal with bluffs from the opponent. Dong Kim, one of the losers, said it was like playing against someone who knew which cards you had in your hand, so impossible it was to beat the machine.

5) Text

In 2018, artificial intelligences beat humans twice in reading. The first was in text interpretation tests, with a system that mixes algorithms and databases. Alibaba, Google and Microsoft beat humans in the Stanford Question Answering Dataset, created by the US university. It gathers questions about Wikipedia entries and serves more to train these AIs to "read" and "write" answers.

The second win in the industry came from artificial smart LawGeex, which beat 20 lawyers in a contract review test. The machine achieved 94% accuracy by doing everything in 26 seconds against 85% of humans in a task that took 92 minutes.

5 artificial intelligence wins against humans [video] via TecMundo