In his recent book on AI, Cory Doctorow called attention^[1] to that case^[2] where a man seeking information about bereavement fares from an airline chat bot was given incorrect information, acted on it, and ended up legally fighting against the later human interpretation of the rules that were different from what he was initially given by said chat bot. 

Hey, I get how easy it might be for a giant corporation to shrug and suggest that mistakes were made, and that maybe common sense could prevail for us all to see that if you’re gonna put your logo on something you had better stand behind what it says.

But look. This whole story got me thinking about the offloading of accountability that was going on here, and that probably occurs more and more each and every day as AI permeates our world: blame the AI. 

It was hallucinating. It was trained on bad data. It was out of date.

Shrug.

Nope. We don’t get to offload this one.

There is no dark pit where we can point and say it wasn’t our fault. There is no aiming blame at an algorithm. There is no handwaving away accountability.

At the root of this someone decided that a machine could do the job of a human… and they were wrong. At the core of the issue is that a call was made, with or without enough understanding of the risks, and something broke. The folks at the top get paid the big bucks to make calls like this so they can’t shirk the responsibility when the technology falls on its face.