Assigning an age rating to a film can be a surprisingly arduous task: from graphic violence to 10 hours of paint drying, it’s hard work to watch every film for objectionable content. And that’s before the rise of streaming video potentially increases that workload many times over.
So it’s no surprise that the British Board of Film Classification (BBFC), which classifies films in the UK, is turning to artificial intelligence to try to lighten the load. A new partnership with Amazon’s cloud computing division seeks to teach an AI model to identify and tag “content issues” such as bad language, dangerous behaviour, sex and violence, to save time when classifying a film or other video content.
The board insists that the AI system isn’t intended to do away with the work of its professional compliance officers, with another four hired recently. Instead, it saves time, cutting the amount of work required by as much as 60%.
“With the exponential growth of online content over the last few years, we’re investing in these new products and the development of scalable solutions to improve our service by making the guidance we provide even more useful to families. Although in its infancy, we’re confident that this project will bring added value to the wider industry by bringing down the cost of classification in the future,” said David Austin, the board’s chief executive.
Some aspects of the classification system, such as bad language, are fairly easy to automate. Others, like nudity, have become more possible in recent years with the progress of machine vision technology. But some categories the BBFC needs to highlight, such as dangerous behaviour or sexual violence, are harder to teach an AI to look out for.
The next phase of the project will see AI systems trained to determine and assign international age ratings, in conjunction with the tagging tool. Ultimately, the idea is that streaming services will be able to get age ratings for their content for multiple territories at once, with the goal of driving down the cost of classification in the future.
The BBFC’s work has been controversial in the past. Financially, the board relies on distributors for its funding, and they are paid for each minute of screen time they classify. That led to “Paint Drying”, a 10-hour film of a freshly painted wall shot by film-maker Charlie Shackleton in 2016. The project was funded by Kickstarter, with Shackleton promising to make the film as long as he could afford to do so. (In the end, £5,936 was raised, and all the money after Kickstarter’s cut was sent straight to the BBFC.)
Due to its length, Paint Drying, which was submitted as one 310GB video file, was assessed by the board’s compliance officers over two consecutive days. Counterintuitively, though, the AI system would be unlikely to have helped speed things up: with no content issues to note at all, the film was passed on its first viewing, and given a U rating indicating that it had “no material likely to offend or harm”.