Facebook will utilize clips from police body cameras to skill its automatic detection machines to remove and spot videos of mass shootings. The firm declared this week that it will join hands with law enforcement in the UK and the US to get clips from their firearms training projects. Data from those clips might assist systems from Facebook to detect first-person, real-world clips of violent crimes.
As per media, Facebook will offer the Metropolitan Police of the UK with body cameras for free. In return, Facebook will have authorization to that info. The firm is supposedly in discussions with US law enforcement about a same association.
Facebook was condemned this fall when, after the New Zealand Christchurch shooting footage from the attack was accessible on Facebook for a long time after the event. Part of the issue, Facebook claimed, is that it did not have sufficient content portraying first-person video of violent crimes to skill its ML tech. Facebook expects video from the cop training initiatives will allow it to enhance its systems.
On a related note, the most enhanced AIs nowadays can do specific errands with superhuman-akin efficiency, overtaking the best people have to provide at intricate titles such as Go and Starcraft II. But ask the same AIs to do something outside of their expertise, and they will fail. Making an AI that is a generalist is one of the upcoming major hurdles of AI study. And it is something a group of scientists at Facebook is trying to do by educating an AI how to master a game.
On the other hand, this time around the game in question is Minecraft. As per Facebook, the sandbox nature and simplicity of the popular game make it the ideal training ground for an AI assistant.