A new experimental website from Wolfram Research will attempt to identify any pictures you throw at it.
The developer of the Wolfram Alpha knowledge engine has launched a new website that uses powerful artificial intelligence algorithms to identify the subject of images.
Simply drag and drop an image onto the Wolfram Language Image Identification Project home page and, after a couple of seconds of deliberation, the underlying ImageIdentify software will suggest what it thinks it is.
The results are, by Wolfram Research’s own admission, a little hit and miss. It’s capable of both some stunningly specific results and some glaring misses.
“It won’t always get it right, but most of the time I think it does remarkably well,” writes Stephen Wolfram on his blog. “And to me what’s particularly fascinating is that when it does get something wrong, the mistakes it makes mostly seem remarkably human.”
He’s not wrong. In our own brief experimentation with the website it managed to identify an Olympus Stylus SH-2 as a “point-and-shoot camera,” while it identified a head-on picture of the new Chromebook Pixel laptop as a “television.”
Most amusingly, it identified a photoshopped picture of a lizard in a hat as a “tricorn”, correctly identifying the specific type of hat while completely missing the fact that it was being worn by a lizard.
Read more: Microsoft Build 2015: Just the highlights
However, the tool has been built to get better with time and practice – every picture submitted (with simple feedback) helps it learn. Its code can also be put to use in a variety of other applications.
“If one had lots of photographs,” speculates Wolfram, “one could immediately write a Wolfram Language program that, for example, gave statistics on the different kinds of animals, or planes, or devices, or whatever, that appear in the photographs.”
It certainly beats trying – and completely failing – to guess a person’s age, doesn’t it?