Many believe that, Artificial Intelligence (AI) has the potential to revolutionize the future. During the Cold War, then-Senate Majority Leader Lyndon Johnson believed that “whoever wins the space race would gain control over the Earth”. Today, AI development is the new space race. Recently, Russian President Putin said, “Whoever becomes the leader in this sphere will become the ruler of the world”. Despite the hyperbole in this statement, the importance of technology’s power should never be overlooked and artificial intelligence is truly today’s hot topic.
Artificial intelligence is not some novelty that exists only in computer laboratories full of geeks and nerds. Actually, it has already stepped into our daily life. AI is responsible for putting a personal assistant in the phone. Siri might be the most well-known example. Although speaking to Siri still sounds silly and unnatural, one has to admit that it is really convenient to find out the nearest restaurant without taking off the gloves outside in a cold winter morning. Tesla car is another smart machine with AI: Tesla incorporate Autopilot into the driving system, which is an ease for drivers frequently spending hours on long distance. Moreover, Tesla can warn an alert to the driver while facing a potential danger of collision, which might save lives.
Since 2011 an advanced AI surveillance system, named as “Skynet”, has been introduced in China in order to protect the public security. (The name, Skynet, comes from an old Chinese saying, which represents justice, instead of the famous movie, Terminator.) The Skynet system is installed into the closed-circuit television (CCTV) cameras located in the public areas, such as airports and train stations. Skynet has the ability to recognize human faces and compare the scanned data with the database of criminal suspects at large. The police will receive an alert notification if a suspect is identified. Along with the Skynet system, other AI techniques have also been used to fight against crimes. Police in Suzhou, China use big data such as the usage of water and electricity of each household to predict crimes: If anomalies occur, the system will notice the local police. For example, if the water usage for a 3-people household increases more than 10 times, it might imply potential illegal activities. With the help of AI, in China the crime rate for eight types of felonies including drug trafficking, robbery and intentional injury dropped 42.7 percent from 2012 to 2016.
However, AI is far from perfect. An MIT research team showed that they can easily confuse Google’s AI classification algorithm: In the experiment, the AI classification algorithm suggests the turtle in the picture is a rifle with 90% probability. The concept used by the MIT team is named as “adversarial image”. It is a picture designed to fool the AI classification algorithm into labelling the picture as something completely different from what it shows. Moreover, this concept can be used not only in 2D pictures, but also in 3D objects by changing their surface texture. The MIT team also built a 3D turtle, which was also classified as a rifle with high probability by the AI algorithm. It would only lead to a false alarm if the algorithm classifies a turtle as a rifle, but it might result in an emergency event if it categories a rifle as a turtle (or other harmless objects).