“CSI” teaches natural language to artificial intelligence. The well-known and very popular TV crime series has been chosen as the testbed to help artificial intelligence understand real-world natural language and its implications.
The researcher Lea Freermann, together with her team at the University of Edinburgh, has developed a study to understand whether, by entering a number of data and information into an artificial intelligence system, this is able to solve difficult problems for human beings. Like, for instance, the identification of who committed the crime in an intricate police case. The result? In 60% of cases, the system is able to reveal the perpetrator in the crime drama. A good result according to Professor Freermann, though still far from that achieved by human beings: 85%.
According to the study, collateral data, e.g. information on the cultural background, needs to be provided to the system to ensure more accurate guesses: to enter a mass of data into the system is not enough. The models should thus be provided with general information to form a body of knowledge that can allow this technology to (methodically) understand reality better.
2018 is set to be a very interesting year for the evolution of artificial intelligence, not only in terms of research but also of legislation. Indeed, a bill has been presented to the US Congress for the establishment of a committee dedicated to studying and understanding the development of artificial intelligence so that this innovation can benefit the economy and society in general.
2018 has only just started, yet it is already set to become the year of AI.