I trained a neural net from scratch on lists of colors I could find online, and with no prior training on English or any language (and therefore no idea what paint colors were) it managed to reproduce some of the letter combinations in the originals. And these neural networks have not only seen English before, they have also read all the popular paint color sites. As the January 2022 Futurist in Residence at the Smithsonian AIB, I've spent the month thinking about how much difference our starting points make, how we have to take care what foundations we build on when working with AI. So as my starting point I used that original neural net's garbled paint colors. A computer screen generates its colors by illuminating tiny red, green, and blue dots that blend together when you view them from a distance. But it had seen enough Red, Green, Blue color listings that it was able to follow the format from just a couple of examples.

One of my first experiments with neural network text generation was to generate and name paint colors. I trained a neural net from scratch on lists of colors I could find online, and with no prior training on English or any language (and therefore no idea what paint colors were) it managed to reproduce some of the letter combinations in the originals. In the nearly five years since, a lot has changed in AI. Now we have giant internet-trained neural networks. And these neural networks have not only seen English before, they have also read all the popular paint color sites.

So if you've ever picked out paint, you know that every infinitesimally different shade of blue, beige, and gray has its own descriptive, attractive name. Tuscan sunrise, blushing pear, Tradewind, etc… There are in fact people who invent these names for a living. But given that the human eye can see millions of distinct colors, sooner or later we're going to run out of good names. For this experiment, I gave the neural network a list of about 7,700 Sherwin-Williams paint colors along with their RGB values. One way I have of checking on the neural network's progress during training is to ask it to produce some output using the lowest-creativity setting.

I've encountered a lot of artificial intelligences, both the ones I've trained for my blog AI Weirdness, and the ones I've written about for my book on artificial intelligence, You Look Like a Thing and I Love You: How AI Works and Why it's Making the World a Weirder Place. I focus on the machine learning algorithms that exist today, the ones that sort spam, tag photos, and drive cars. We call them AI, but they're as different from the AI of science fiction as a toaster is from a person. In the book, I spend a lot of time explaining why today's AIs, with their tiny worm brains, don't understand their tasks or the human world. They won't be taking over from people, but they also won't be saving us by questioning bad orders.

Artificial intelligence (AI), like time travel, is a perennial subject for writers of science fiction. And, like time travel, AI is subject to a number of misunderstandings which can make writing a story in that setting, on that subject, using it as a McGuffin, or as a character, problematic. With movies like Blade Runner 2049, Ex Machina, and Her keeping AI in our geeky gestalt, writers will continue to tackle the topic. A little research might spark a new, or more realistic, take on the AI tale, though. The first thing to understand is that true AI, a machine that is self-aware and thinks independently, is the stuff of science fiction.

A scientist fed a database of 7,700 paint colors into a neural network. Then, based on her training data, she asked the computer to create a few paint colors of its own. And names that didn't correspond at all to the color: If you get value out of these emails, please consider supporting our nonprofit.

A new AI experiment recommends colors like "Sink" or "Bank Butt." SEE ALSO: We're not saying you're obsolete, but Google is making music humans can't Researcher Janelle Shane trained a neural network to come up with paint names, and it resulted in gems like "Stoner Blue" and "Turdly," she wrote in a blog post. She conducted an experiment by using a training data set of about 7,700 paint names from Sherwin-Williams along with the RGB (red, green, and blue color values). She then had the network generate new colors and names for those colors. The results were hugely varied, including colors with Gray or Blue in the name that had nothing to do with those colors, or completely abstract and awesome names like "Stanky Bean."

Humans aren't nearly as creative as we think. Craft brewers, for example, have run out of fun names and are sending each other cease and desist letters for coming up with the same ideas. So, what if we let computers come up with new names for us? That's the problem optics industry research scientist Janelle Shane has been trying to solve using neural networks, but with paint colors. The initial results are downright ridiculous.

At some point, we've all wondered about the incredibly strange names for paint colors. Research scientist and neural network goofball Janelle Shane took the wondering a step further. Shane decided to train a neural network to generate new paint colors, complete with appropriate names. The results are possibly the greatest work of artificial intelligence I've seen to date. Writes Shane on her Tumblr, "For this experiment, I gave the neural network a list of about 7,700 Sherwin-Williams paint colors along with their RGB values. Could the neural network learn to invent new paint colors and give them attractive names?"