Managing a vast amount of content is likely to be a growing problem for Amazon in the coming years. Its streaming service has 18,032 films in the UK, according to Statista data, compared to 3,802 films on Netflix.
Amazon’s AI classification service could be added to its Rekognition service, a facial recognition technology that the company has licensed to government agencies and businesses.
AI experts have warned that even advanced systems can still backfire, however, meaning any reliance by Amazon on this may cause problems.
“It’s super important that we machine decision making with human decision making,” said Daniel Hulme, the chief executive of artificial intelligence business Satalia. “If you just leave it to the machine, there are going to be things that the machine will miss.”
“The machine also can’t pick out innuendo. What it is terrible at doing is understanding context which is what humans do,” he added.
Amazon has previously struggled with AI systems, with the company scrapping an AI hiring process in 2017 after it found that the technology was biased against women.
The data fed into the algorithm, which included 10 years of CVs, meant that Amazon’s AI hiring system ended up favouring male job applications, penalising CVs that included words such as “women’s”.
Amazon’s patent filing is not the first time that experts have attempted to use AI to predict a film’s rating.
Last year, researchers from the University of Southern California’s Viterbi School of Engineering published an academic paper revealing that they had developed an AI system trained on 992 scripts that predicted a film’s classification based on its script, without needing to examine the finished film.
Despite Amazon already possessing advanced AI systems that could handle film classification, the filing of a patent for the system doesn’t guarantee that it will eventually be built.
Technology businesses often file patents as a form of marketing or to block rivals from developing features. Amazon in particular is well-known for filing zany patents such as drone waterskiing handles and whipping satellites into space that have little chance of becoming reality.