Google’s DeepMind team has developed artificial intelligence that can generate policies that suit everyone. Although artificial intelligence gave a successful result, it did not fit democracy according to DeepMind, but according to political scientists, it did.
The debate about artificial intelligence replacing the governments of countries has been going on for a long time in the world. While the issue of how ‘equal’ artificial intelligence will treat people and how ‘ethical’ is a big debate area, special news came from Google’s artificial intelligence initiative DeepMind recently.
Google’s DeepMind team has managed to train an AI system to come up with a ‘one size fits all’ policy for distributing public funds in an online game. In the training of artificial intelligence, 4000 real people and computer simulations in an online game played by 4 people were used.
AI has developed a common policy that does not conform to either of the two accepted policies:
In the game in question, players start the game with different amounts of money and decide how much to contribute to help a pool of public funds grow. In the end, they get a share back for their contribution. Players also voted for their favourite policies to distribute public money.
Analyzing this game for education, artificial intelligence finally determined its own policy. The AI redistributed public money according to how much of each player’s starting money contributed. In this way, he tried to reduce wealth inequalities between players. It also gave almost nothing to ‘freeloader’ players who didn’t contribute half of their starting funds.
The players, who voted on the policy of artificial intelligence along with the policies they developed, chose the policy created by artificial intelligence as the majority. Other options for this policy were ‘egalitarian’, which distributed funds equally regardless of how much each person contributed, and ‘libertarian’, which distributed funds according to the proportion of each person’s contribution.
Artificial intelligence, although successful policy making, does not quite fit into democracy:
The DeepMind team explained that although artificial intelligence managed to prove itself, it could not meet democracy as follows:
“Democracy is not just about winning, implementing the policy you like best – it’s about creating processes where citizens can meet and argue with each other as equals.”
While the DeepMind team stated that artificial intelligence ignores the needs of minority groups (those who vote for other policies), there was also a counter-argument. Mathias Risse of Harvard University shared that this is acceptable among political scientists. Risse emphasized that there are problems in modern democracies such as economic elites depriving the ‘majority’ of their rights.