So… I’m working on editing the first draft of my flow-of-consciousness cyberpunk novel, as you do, and my ‘day-job’ is working in A.I. (I say ‘day-job’ as being a scientist is very much a part of who I am, not just something that I do until I make it as an author).
And sometimes, you just see something so crazy.
Essentially, to try and raise awareness of problems of training data for A.I., they trained an algorithm with data from the darkest corners of the net, and then compared its ability to caption images with an A.I. trained with more normal data, the result is that they have made an A.I. that would be diagnosed as a psychopath. And they did this to warn us all of the danger of making a psychopathic A.I. by accident (luckily, they did it by design, then!) The link is below, and probably about a 5 minute read…