Looks like Skynet draws nearer yet again – although on this occasion, it's a bit more literal.
UK researchers have taught a computer how to “see”, turning it into a machine that doesn’t just scan the night skies, but can also recognise galaxy types without any human help.
Astronomers and computer scientists at the University of Hertfordshire have developed an algorithm which taught an artificial intelligence machine to use image recognition to classify different galaxies. Just by studying an astronomical image, the machine can tell you which galaxies are elliptical and which are spiral
The really neat thing is they didn’t tell the machine what to look for, but rather taught it how to “see”. Researchers are calling this “unsupervised machine learning”.
The university’s Dr James Geach commented:
A human looking at these images can intuitively pick out and instinctively classify different types of object without being given any additional information. We have taught a machine to do the same thing.
So basically: it can do what we do when studying the same image. Only a lot faster.
What does that mean for the future of AI? In the future, we could develop robots that “see” better than we can, which could be useful to help doctors spot tumours, security personnel find suspicious items, or any other situations where no human could closely inspect every piece of data.