Someone trained a computer to invent names for paint colors, and it went totally nuts
Are you one of those people who rolls your eyes at the names of paint colors? “Egg shell and cream ARE BOTH WHITE I literally CAN’T TELL THE DIFFERENCE!” we can hear you screaming now.
Well, we’re here to tell you that things can ALWAYS be worse.
Like, say, if you let artificial intelligence name paint colors.
And, boy, AI sure did help. Just not in the way anyone expected. Shane gave the neural network 7,700 Sherwin-Williams paint color names and their red, green, and blue color values, and the neural network spit back out some truly fantastic names.
It took a few checkpoints for the network to get the hang of this whole paint naming thing. This is how things looked at the first check point:
The training process takes some time. The neural network has to, you know, learn to spell words like “gray” and “green.” Though a few checkpoints later, it’s giving us some pretty silly paint names like “sane green” and “conk green.” (Which are both, for the record, clearly blue.)
Later in the training process, the network has figured out colors like white, red, and grey, and is giving us some pretty slamming color names. It’s so hard to choose a favorite. We love “Power Gray.” But we’re also obsessed with “Hurky White.” Real talk, we want to paint our bedroom pink, but how are we supposed to choose between “Ghasty Pink,” “Burf Pink,” and “Rose Hork”?
And finally we end up with these beauties:
We can’t even choose a favorite. How can we? Okay, maybe it’s Bank Butt. But there’s also Dorkwood, Stummy Beige, Caring Tan, Sindis Poop, and Stanky Bean! We kind of wish these were real Sherwin-William paint colors. Imagine how much more awesome your office would be if the walls were Bank Butt colored? One can only dream.