Much of the public debate about AI has focused on the often bizarre errors and biases in AI-compiled facts and figures. But there is another, perhaps more sinister danger: we are now entering a new era in which photographic evidence can no longer be trusted.

Here is an example:

Of course, the “1963 Mercedes 600 Pullman” never existed. Obvious mistakes (such as the nonsensical air vent and the missing hinge gap on the door’s lower section) make it easy to tell that this is just a rendering. And yet, most social media commenters were tricked into believing that this was a classic vehicle.

Given the rapid pace of development, obvious mistakes in AI-generated photographs will soon disappear. Soon it will be impossible to tell that a photograph or video was made up in a computer.

People have been conditioned to believe what they “see with their own eyes”. It is time to stop doing that.