As children, a lot of what we know of the world today—things we dub ‘common sense’—is learnt through visual cues. We observe the world around us and form judgements based on the information we receive. This may seem very rudimentary, but it forms the basis for most higher thought processes and is an essential part of our learning. Scientists at Carnegie Mellon have sought to determine whether this process—of gaining information through images—can be applied to computers. What emerged was the Never Ending Image Learner (NEIL) computer program. The NEIL program has viewed 3 million images since July and has been able to identify 1,500 objects, 1,200 scenes and 2,500 associations—all on its own. So, by just looking at images, the computer has been able to teach itself things like cars are found on roads, ducks can resemble geese and that zebras are found in the savannah. One of the more everyday uses of a computer with common sense is that online search results can become far more accurate and targeted. But the added levels of computational and analytical abilities could also have a direct impact on military technology as well. It’s no surprise, then, that the NEIL program is funded by the US Department of Defense and Google.
Full-fledged artificial intelligence of the science-fiction-kind is still a long way off. For one, the computational power needed to run even the NEIL program—200 processing cores—is far in excess of what is commonly available. Second, human intelligence is just too intricate and complex to be rendered in digital form any time soon. That said, AI is definitely on its way, and common sense at least may no longer be the sole province of humans.