Researchers at the Massachusetts Institute of Technology and Cardiff University in the U.K. found that artificial intelligence (AI) bots can develop prejudices by learning from each other.
The team created a game in which AI bots must decide to either donate to a bot within their own group or to a bot in another group, factoring in donation strategies and the bots' reputations. As the game progressed, the bots became increasingly prejudiced against bots from other groups. The bots devised new strategies by copying each other, mimicking approaches that offered a better short-term reward.
Cardiff's Roger Whitaker says, "Our simulations show that prejudice is a powerful force of nature and through evolution, it can easily become incentivized in virtual populations, to the detriment of wider connectivity with others." Protection from prejudicial groups can cause individuals to create further prejudicial groups, resulting in "widespread prejudice" that is difficult to reverse, Whitaker says. Lower levels of prejudice occurred when a greater number of subgroups were present in a population.
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found