But back to the topic for today. Take a guess at which stock chart the blue line represents in the graphic below:
The green line is the S&P 500 index for the year 2016 – it was up 11%. Not bad, but our mystery company was up 247% in 2016, the best performing stock of the S&P 500. I’ll give you a hint: up until very recently, this company was primarily a component supplier to the gaming industry, but now it is very much an integral part of the artificial intelligence trend. OK, I’ll end the suspense: the answer is Nvidia, which makes graphical processing units (GPUs). GPUs have been a key enabler for life-like, immersive video game experiences, but something happened in 2012. Call it, the Big Bang for Artificial Intelligence, and Deep Learning in particular. In 2012, Alex Krizhevsky of the University of Toronto won the ImageNet computer image recognition competition. According to Nvidia, Krizhevsky beat handcrafted software written by computer vision experts by a wide margin. But Krizhevsky and his team did not write any computer vision code. Rather, using Deep Learning, their computer learned to recognize images by itself. They designed a neural network called AlexNet and trained it with a million example images that required trillions of math operations on Nvidia GPUs.
It turns out that the parallel processing architectures of GPUs are really good at the heavy calculations required for Deep Learning algorithms. Deep Learning engines are similar conceptually to the neural networks that were first created back in the 1950s, just more complex, with more layers (ie, “deeper”). So we’ve had neural networks for 60 years, but here’s what's new:
We now have the massive amounts of data needed to feed into and train these Deep Learning systems.