Robots – to Parliament to think "for us"

A very unusual question was asked to the residents of the EU in terms of whether they would like to be controlled by artificial intelligence. A survey of scientists revealed interesting trends.

It was hosted by IE University’s Center for the Governance of Change in Madrid. Scientists figured out what such a replacement would lead to.

Most of all this would be wanted by residents of Spain (66%), Italy (59%) and Estonia (56%). Residents of the UK (69%), the Netherlands (56%), Germany (54%) and the USA (60%) responded negatively.

IE University’s Change Management Center conducted a similar study two years ago. At that time, only 25% of the Europeans surveyed spoke in favor of using artificial intelligence (AI) in political decision-making.

Many researchers drew attention to the tendency of decreasing trust in politicians. Back in 2019, the Cambridge University Press published an article in the European Political Science Review by a researcher at the University of Zurich, political scientist Eri Bertsu, which claims that people’s distrust of politicians has become common in democracies and is often used to explain current political developments.

Researchers at the Pew Research Center (USA) concluded that many citizens see the need to reform the political systems of their countries, regardless of the degree of satisfaction with the government (studies were conducted in 2019 and 2020).

The poll was conducted at the end of last year; 4.1 thousand citizens of the USA, France, Germany and Great Britain took part in it. The results of the 2019 survey also show a low level of public confidence in political systems: 68% of those surveyed noted that it is very important to restore the level of public confidence in the federal government.

According to some experts, replacing politicians with artificial intelligence systems seems attractive only at first glance, but in reality there are many problems associated with this. Any system that uses machine learning can initially be biased, according to the specialists of the American magazine IEEE Spectrum, published by the Institute of Electrical and Electronics Engineers. An example is the case where a system that reviews loan applications may refuse a client, explaining this decision by income and age, when in reality it is based on race. This bias can occur because it reflects the correlation in the data that was used to train the AI.

Back in 2016, The Conversation, an international scientific media network, published an article that asks what decision would artificial intelligence make when faced with a moral dilemma? The authors of the article come to the conclusion that even if there was a parliament full of robots in the world, it would still require “an agency with live employees who would define ethical standards.” Thus, the main problem is that the robot must be programmed with an agreed set of ethical standards to enable it to make appropriate judgments.

The New York Times cites comments from experts who work with AI systems and points out the biased decisions made by robots. Stanford University Computer Science professor Daphne Koller cited the example of radiology, when artificial intelligence analyzes X-rays and must make a diagnosis based on data from several hospitals.

The danger lies in the fact that the algorithm can learn to recognize in which hospital the image was created, and if in some hospitals the percentage of fractures is much higher than in others, then the diagnosis will not be based on an X-ray image, but based on a set of data and that which hospital the scan was performed at. Thus, the use of artificial intelligence in political decision-making can create certain problems.

Most politicians are negative about the idea of ​​automating their work. “Deputy work cannot be replaced by any artificial intelligence due to the fact that a deputy is obliged to express the moods, opinions of all his voters, understand the development trends of the country and the world, make appropriate adjustments every day, know the history and practice of his state, national character,” the leader said. the communist party. At the same time, he recognized that artificial intelligence, as a tool, can help a person.

Experts believe that it is quite possible to teach the machine what parliamentarians are doing, but there is no point in this.

“The problem is that even if we hypothetically assume that we currently have a cognitive architecture that can be taught to write laws, and there are a sufficient number of examples of how to write laws, the learning process of such a system would be much more expensive. than to take a person, give him a salary and put him in charge of the duties of a deputy, ”- said the director for science and technology of the Agency for Artificial Intelligence.

…



Source link