Here's the thing that scares me about AI. A couple years back, I was a stay-at-home dad and was looking for ways to earn money from home. I came across some apps which allowed me to take surveys, and do various tasks to earn gift cards and cash. One of the tasks gave me a picture of a rain soaked road and I had to "draw" on the screen where I saw the lane lines. The pictures were often very grainy and trying to draw them on a small iPhone screen was rather difficult. I didn't always get it right, but the computer took my responses anyway and paid me like 10 cents per picture. In a half hour, I could do a couple hundred pictures and earn some money, even sitting on the crapper. I really didn't care why I was doing it, but it was really easy to do and I made money. I got to thinking, "any idiot can do this". And that's pretty much who it's targeted to. Any idiot with a smart phone can download this app and make money from their couch filling in lines on a picture.
Later on, I got an email from the company that was paying for the "research" I was contributing to thanking me for my time helping them develop the software that will eventually run autonomous vehicles. So, the moral of this story is that AI developers are using the cheapest possible research, from the least reliable sources for what will most likely be the most dangerous leap of technology.
