A team of researchers at the Massachusetts Institute of Technology (MIT) infused artificial intelligence (AI) with the material describing death on the community’s website, creating the first AI that resembles a neurotic disease called Norman.
Agence France-Presse reported that the reason why the research team created Norman was to use popular language to explain how the algorithm works and to make people aware of the potential danger of AI.
According to the Massachusetts Institute of Technology (MIT), a renowned American university, the Norman representative is learning machine reading bias information of the algorithm, leading to dangerous cases of AI error.
Three members of the research team stated via e-mail: “When the machine is learning, it’s important that when instructing the machine to learn the algorithm, instilling machine information will have a significant impact on its behavior.” They said: “So When we talk about AI’s algorithm being unfairly biased, the culprit is usually not the algorithm itself, but the bias information fed into the algorithm.”
The inspiration for Norman’s name comes from Alfred Hitchcock’s 1960 movie Psycho, titled “Norman Bates” mental metamorphosis killer.
The research team only instilled essays describing human death on the Norman community website Reddit. The researchers then conducted a Rorschach ink psychological test on Norman in order to determine the vision in Norman’s eyes and compared Norman’s answer with the traditionally trained AI’s answer, only to produce deaf comparison results.
When the traditional AI sees “two people standing close to each other,” Norman sees “a man jumps out of the window”; while Norman sees “the man’s screaming wife shot,” the other AI sees “One person holds an umbrella.”