Artificial intelligence algorithms are still bad at arguing
People like to argue, especially on sensitive topics. In fact, all policy in many countries is based on the ability of opponents to argue and get the audience round to own way of thinking. It is the ability to argue, make correct arguments and persuade is considered a stumbling block in the development of artificial intelligence. Thus, the brainchild of IBM Project Debater for several years tries to prove his point in the discussions. In the summer of 2018 the machine took part in the debate, where its opponent was a famous TV presenter Dan Zafir. The essence of the dispute was to subsidize the remote medical services industries, as well as space exploration. And Debater won a landslide victory. But recently, AI had to deal with a much more difficult opponent – Harish Natarajan, who won the world championship in discussions. Their communication took place in the framework of IBM conference in the United States, opponents talked about the benefits and harms of state support for children’s educational institutions. The result was a very exciting and complex dispute.
Artificial intelligence algorithms are not yet able to argue
The subject of the dispute was not reported in advance – it was announced right at the event. The main objective of AI was the belief of the participants that subsidies for educational institutions for preschool children is an important public task, and Harish Natarajan argued with this position. Each participant had 15 minutes to prepare – it was enough for AI to handle 10 billion news on the topic, and Harish made an approximate plan of his speech. The debate itself consisted of several stages: 4 minutes speech, the response to the attacks of the enemy and the final 2 minutes. Interestingly, Project Debater began its speech with a light joke that despite all the achievements of Natarajan, who is the champion in the debate, he certainly did not argue ever with the computer, so that the future is already here. As an argument in favor of subsidizing AI gave examples of what state support can provide decent support for children from problem families.
He is convinced that every person must have a moral desire to help those who really need it. Natarajan also expressed the opinion that the issue of subsidies does not imply improving the quality of education. He believes that it is very likely that this step will simply be the distribution of funds to all in a row, and low-lifes will remain without real help.
Harish noted that some children may not need to attend preschool. AI took his phrase as “My opponent believes that kindergartens are harmful.” But in spite of such verbal games and alerting, Harish resulted in a much stronger and more precise arguments than artificial intelligence.
The debate was judged by 416 spectators present at the conference. They voted with their smartphones – before the debate and at the end of it. In the course of his speech, Natarajan persuaded 17% of the judges, so he was the winner. The interesting thing is that artificial intelligence gave viewers more useful information. After a verbal duel Harish noted that the neural network appeared very worthy rival with whom it is difficult to argue.