It's impossible to say exactly how much money Uber makes from drunk people, but if the number of bleary-eyed people wandering around on Friday and Saturday nights trying to find their summoned cars is anything to go by, it's probably quite a lot. The company clearly knows its audience: this week, it applied for a patent for an AI that could spot drunk or high passengers simply by the way they walked, typed or held their phone. According to the patent, the AI could measure a user's walking speed, watch for unusual typos or sense whether a phone is swaying or being held at an unusual angle. This, it suggests, could "predict user state using machine learning" and recognise "uncharacteristic user states". The company almost certainly believes that this information would be used for good, and it's undeniable that the option to avoid intoxicated passengers would come as a blessed relief to many drivers.

After revealing a first look at its flying car prototype, slated to launch in LA by 2020, Uber looks to further improve its service, as the ridesharing giant has explored identifying drunk passengers with artificial intelligence, according to a patent application. CNN first reported the patent, which describes a method of measuring the user's behavior on their phone against their usual behavior. The system relies on an algorithm to weigh a variety of factors, including typos, how precise a user clicks on links and buttons, the speed at which the user is walking, and how long it takes to request a ride. The patent also suggests that the service Uber provides to the user could change, as a result. Drivers may be warned of their passenger's state, and riders in an unusual/drunken state may only be matched with drivers with relevant experience or training, the patent application says.

Uber is working on technology that could determine just how drunk passengers are when ordering a ride. According to CNN, Uber Technology Inc. filed a patent for a machine learning application that could accurately predict a user's state of sobriety and alert the driver with this information. The patent application describes artificial intelligence that learns how you typically use the Uber app, so it can better spot unusual behavior. Its algorithms weigh a number of factors, including typos, walking speed, how precisely you press in-app buttons, and the amount of time it takes, standing on the curb, to order a ride. Someone botching most words, swaying side-to-side on the curb, and taking 15 minutes to order a ride late on a Saturday, for example, is probably fairly intoxicated.

Uber has drawn a plan for artificial intelligence that could detect a user's drunken behavior in order to decide whether or not they want the passenger …