How AI could help some of society’s toughest problems
https://www.technologyreview.com/s/612054/fei-fang-carnegie-mellon-artificial-intelligence/
In this article, “How AI could help some of society’s toughest problems,” Charlotte Jee explores Fei Fang’s use of game theory and technology in the real world.
Fang’s work focuses on employing Artificial Intelligence (AI) for social benefits, such as protecting crucial national infrastructure, reduce homelessness and prevent suicides. To achieve this, Fang conceptualizes different high-risk situations as a “zero-sum, two-player game against two adversaries,” and focuses on randomizing patterns so that it is difficult for the other player to pinpoint the weakest points of the system (Jee, 2018).
In 2013, Fang developed a system to protect 60,000 passengers on the Staten Island Ferry in New York City. She found that in the same area, more ferries were traveling between Staten Island to Manhattan than US Guard Patrol boats. Fang identified this circumstance as risky and deficient. There were cases where one patrol boat would only follow one ferry for its entire journey while leaving the rest of the ferries out in the open, unprotected. To prevent this, Fang and her team developed an algorithm that randomizes the patrol routes based on potential high-risk areas, making them unpredictable. This algorithm was found to be pretty successful; it “reduced the likelihood of a successful attack by half,” according to professional mariners (Jee, 2018). Using a similar approach, Fang employs her AI system for other situations such as illegal animal poaching in Uganda, detecting illegal mining sites and even to predict crop yields to prevent hunger.
This article reflects the concepts and theories related to the game theory discussed in class. Every decision we make, big or small, involves some process of analysis. As discussed in class, when we are making a decision that involves another person, we deliberate the pros and cons (the payoffs), attempt to gauge what the other person is going to do, and finally, choose the best option that is beneficial to us. Likewise, Fang’s game-theory implemented AI estimates what the threats are and how the attackers (high-risk individuals or groups) will act. Fang’s AI system uses the strategy of randomization to maximize the payoff. This is so that the other player, in this case, attackers, poachers, etc., cannot predict the AI’s actions, making it hard for them to gauge their payoffs. In a case like this, there would be no Nash equilibrium strategy other than avoiding the play. The attacker would either win or lose. In response, the AI system would lose or win, respectively. Hence, Feng’s conceptualization of a “zero-sum” game.
Although the system has shown to be successful, it is hard to say whether randomization will always result in the maximum payoff, or whether it will always be the best strategy. What if there are times when randomization provides other players the opportunity to attack? What if the randomization misses an extremely high-risk area? While game-theory is a familiar and latent concept that occurs in our everyday lives, this article shines a light on how we can use concepts such as the game theory to benefit the society as a whole. It raises questions about the possible future implications of game-theory in technology, as well as its limitations.
Jee, C. (2018, September 12). How AI could help solve some of society’s toughest problems. Retrieved September 18, 2018, from https://www.technologyreview.com/s/612054/fei-fang-carnegie-mellon-artificial-intelligence/