My first PC was a powerhouse in its day: a Pentium 200MHz MMX processor, an insane 32MB of RAM, and a massive 2.5GB hard drive complemented by a blisteringly fast 16 speed CD-ROM. That was back in 1997. However, for about 5% of the cost of that (and probably less than 5% if we talk in real terms), there is now something so much more powerful available. I mean, that’s probably stating the obvious given Moore’s Law, but it’s how wide the gulf is that amazes me.
The Pentium 200MHz MMX processor came with a single core, and clocked in at a top speed of about 0.2 GFLOPS. Now, a little over two decades later, something new has come along that excites me quite some bit. The Jetson Nano from nVidia clocks in at an incredible 472 GFLOPS. That’s 2360 times faster than my old Pentium, which (more stats coming) is more than 100 times faster for every year that has passed since those exciting early days in the Autumn of 1997.
Deep learning is becoming increasingly easy to get started with now, and with devices like the Jetson Nano being purpose built for the kinds of data analysis at scale that training a machine learning model requires, the bar to entry is forever falling lower. I have one of these on order, and when it ships, look forward to some guides on how to set it up, common pitfalls to avoid, and some examples on how to build your first machine learning model.