Bet On The Bot Or Humanity: Who Wins?

August 1, 2019

It’s no secret that Artificial Intelligence is progressing rapidly, and we’re near the start of a new era. Machines already win at complicated strategy games, and even more – they can master driving. But there are still things that we, humans, are better at. Let’s find out whether a human brain can still stand a chance when competing with AI.

One glorious representative of our species took down IBM’s Miss Debater during a showdown at the company’s sync conference. The champion in question is Harish Natarajan who won the 2012 European debate and was a grand finalist in 2016’s World Debating Championships. He battled the AI over preschool subsidies for about 25 minutes which included 15 minutes of prep time, 4 – opening statement, 4 – rebuttal, and a two-minute closing summary.

Miss Debater formulated its arguments from a database of 10 billion sentences, drawn from newspapers and academic journals. While Harish Natarajan presumably pulled arguments from his brain. After both parties made their points and the audience of 700 journalists and top debaters from Bay Area Schools crowded Natarajan the winner. This was the first won for the human race and AI in a while.

For decades, human beings have played games versus computers. For example, Google’s DeepMind project created AlphaZero. The AI program becomes unbeatable at the most complex strategy games on the planet: chess, shogi, Go. The AlphaZero is trained to analyze situations by breaking the game down into tiny parts and visualizing all possible moves. Lately, the AI has beaten the world’s best chess-playing computer program. The AI taught itself how to play chess in under four hours. Experts note: if the system like AlphaZero can learn all the moves in Go, then it has the potential to replace dozens of jobs.

Elsewhere Elon’s Musk non-profit AI startup called OpenAI has achieved a remarkable feat. Every year the Valve Corporation hosts the competition for expert players of Dota 2 – a popular online multiplayer game. The competition brings together players from all around the world, who are fighting for the Grand Prize. The AI was trained by the engineers of Elon Musk. It competed one-on-one against one of the world’s best Dota 2. Dendi, who plays Dota 2 professionally, was surprised that an AI could outplay a human. He said that playing against the AI felt almost like playing against a human. The incredible win was achieved after just two weeks of real-time learning by the AI.

The International 2018 champion OG lost to Artificial Intelligence too. OpenAI co-founder and chairman Greg Brockman explained that AI is powered by deep reinforce learning, which means they didn’t code it how to play. Engineers coded it how to learn. Only in its 10 months of existence, it’s already played 45,000 years of Dota 2 gameplay.

Besides, engineers from Carnegie Mellon’s University created AI which battles against world-class pro poker players. The program, Pluribus, achieved superhuman performance at multiplayer poker, which is a recognized milestone in AI, and in-game theory. According to the creators of Pluribus, the technology could be used to solve a wide variety of real-world problems that, like in poker, involve actors who bluff, or hide key information. First, the program practiced against itself, learning little by little how to use poker moves to best its advantage. But its major strength is the ability to use mixed strategies, the same thing humans try to do. It’s a matter of execution for humans – most people just can’t play in the perfectly random, and consistent way.

Going back to Harish Natarajan, he thinks, that debate is more complicated for a machine than games like Go and Dota 2. He said that debating involves processing large amounts of information to construct relevant arguments, clearly explaining complicated concepts. The most important thing is to make those concepts matter to the audience. Natarajan also concedes that a day when AI will be better than humans at all those three tasks will come.