The push for automation and lower costs has many companies around the world rapidly adopting robots, neural networks, and similar technologies. Putting aside the sociopolitical conflict tied to that trend, the truth is that sometimes artificial intelligences perform horribly. The latest example comes from a bot on Amazon called "My Handy Design," which is pulling random photos from the internet to design cases for iPhone and Samsung Galaxy S5. From adults in diapers to heroin injections, anything goes.
The Unsupervised Bot
One of the many lessons we learned from Microsoft's experiment with its Tay bot on Twitter is that some AIs need supervision. Of course, that goes against the goals of any AI, but without enough care, you fall into the classic trap of "garbage in-garbage out." What happens when we let them work on their own without restrictions? Well, Tay became a racist, Holocaust-denying psychopath in less than 24 hours due to user influence. Now imagine a bot that has spent almost a year choosing images to design cases for iPhone and Samsung Galaxy...
The Bizarre Designs
"My Handy Design" was detected on Amazon by data artist Russell Jurney. If we check the sales profile, we'll see that not many products remain (after the coverage, Amazon swept it and decided to censor results), but nothing is ever fully erased from the web. And the images... pure delirium. An elderly woman using an inhaler. A spoon with heroin and a syringe. A printer. A woman in the middle of a waxing session. A muscular man using another's abs as a table while cutting cheese. An adult in diapers and crutches. Basically, a mix of health, medicine, drugs, beauty treatments, personal hygiene, bathroom and kitchen accessories, "adult toys," and the occasional piece of technology.
The prices of the cases range between $5 and $25, and if we go by the reviews from some daring users and audacious ones who decided to buy one, the quality (both of the print and the case itself) is dreadful. I think product personalization is an aspect that AIs will have to keep optimizing for many years.
Source: IFLScience
