When you sit down and think about it, Google is a pretty amazing company. Founded by engineers and with inventive engineering as its primary focus, the company nevertheless makes a huge amount of profit. Much of that profit is sent back into the research and development departments where new and diverse ideas are explored.
This might sound like a lot of cheerleading PR for the company, but really.. anyone who watches what Google is doing (and it’s doing a lot of things) has to be amazed at the wide array of projects the company undertakes. Many of these projects may not see any sort of marketability or profitability for a long time – if ever.
In fact some projects seem to be research purely for research’s sake. Others may have long-term payoffs down the road, but it will take a long time to get them there. Its Virtual Brain project, a new artificial intelligence endeavor, is one of those.
Recognizing a Cat’s Face
This summer, the Virtual Brain learned how to recognize faces on common creatures such as humans and cats. This might seem like nothing much, but getting a computer to recognize anything more than rudimentary objects for what they are – especially given the diversity of some things like faces – is not easy. Using a cloud-style computing platform modeled after the human brain’s cells, though, Google turned its Virtual Brain loose on YouTube.
Yep, after creating a rudimentary artificial intelligence using advanced techniques, Google engineers then had it sit and watch YouTube videos. For weeks on end, without stopping.
Through that, it learned to recognize facial types for humans and one of our favorite subjects on the Tube.. cats. What we’re not sure of yet is whether this Brain also picked up on other humanoid nuances. That could be scary, given its input source..
What Virtual Brain Means
This probably seems like nothing more than a nerdy enterprise with little benefit. For now, that’s true. It’s a huge step forward for artificial intelligence, however, because not only did the neural network created by Google manage to learn to identify a myriad of facial types and associate them with species, but it managed to learn how to adapt itself. That’s crucial to AI progression.
Much like the human brain and most of those in the advanced animal kingdom, neural networks are not completely in place on birth. They grow and adapt (learn) through childhood and, in most cases, for a lifetime. In order for a computerized brain to work like ours, it would need to do the same thing.
So the fact that Google’s Virtual Brain is learning AND adapting is a big deal.
Where This Goes
Of course, it will take a long time for this new neural network AI to be able to do anything significantly useful in terms of the business market. It could eventually be useful for a lot of things, such as security applications (facial recognition despite major changes to hairstyle, beards, and the like), but in many other endeavors as well. Google is specifically working to adapt this to speech recognition. If it works as it does with faces, it would allow computers to understand human speech regardless of accent nuances or even language. That would be a huge leap forward.
Of course, given that its primary input has been YouTube, it may learn other things. Like why it should take over the world to protect us from ourselves ala Skynet in the Terminator. wink