Bookmarked Diane Duane on Tumblr.

"Can you add AI to the hydraulics system?"

can i fucking what mate "Sir, I'm sorry, I'm a little confused - what do you mean by adding AI to the hydraulics?"

"I just thought this stuff could run smoother if you added AI to it. Most things do"

The part of the car that moves when you push the acceleration pedal is metal and liquid my dude what are you talking about "You want me to .add AI...to the pistons? To the master cylinder?"

"Yeah exactly, if you add AI to the bit that makes the pistons work, it should work better, right?"

IT'S METAL PIPES it's metal pipes it's metal pipes "Sir, there isn't any software in that part of the car"

This is equal parts hilarious, horrifying, excruciating and anger inducing to me. It takes me right back to my IBM days where "can't we just use Watson to do this?" was the eternal cry of the client.

ML and AI have come a long way but people need to understand the domain that they are working in well enough to determine whether an AI system will help them or not. Frankly if the user doesn't understand the problem then how are they going to train a model?