CES is back yet again and bringing all the heat again down here in Las Vegas. The Guru8 team has already pitched its tent and dedicated to bringing you all as it happens live.
Today, we were moving around checking out the Best Tech CES 2016 has to offer and boy Oh Boy we saw some jaw dropping stuff. Focus was put mostly on smart home devices, Fitness and Health Devices and the Smart Cars.
All this will be brought to you in a bit but first we need to blow you mind by the NVIDIA Drive PX 2. It sports 12 CPU cores and has 8 teraflops worth of processing power — similar to about 6 Titan X video cards. This is a supercomputer with the equivalent of 150 Macbook Pros and wait for it…. It’s gonna be used to run a self-driving car, Damn. I know right?
Well, as you must have guessed wrong, this is not a huge device aimed for a “Magulu Kumi” No. It’s a very summarized contraption that fits in a “Lunch Box”. You must be telling me to “Gerr outta here” in your mind but bear with me, am not making any of this up.
So, these guy are blaming all the fatalities of a car Humans. This has been rounded off to nearly a Million fatalities across the world each. These guy have concluded with making a self-driving car and eliminating the Human.
Nvidia research recently made the Pascal GPU which you guys already know about. To make this self-driving car, the machine needed to learn or perceive what is around it. This is where the Pascal GPU came in handy. It was able to boost the self-learning training exercise up to 30 – 40 times. Am entirely basing on the fact that you guys know about Artificial Intelligence in Robots and stuff, keep up guys… Keep.
This whole learning process has been dubbed Deep Learning which is able to achieve super human perception capability, not the Avengers Super Human caps but am guessing heightened sense of Hearing, Sight et cetera
All in a nut shell, this juggernaut has the ability to recognize every single object be it simple or complex. So in lay man language, it can detect a car, pedestrian or stray dog crossing the road. It was able to achieve this Image Net Data Set in just a month and without the Pascal GPU acceleration, that would have taken years, trust me.
During this training, one feature was coded by hand. Which is basically holding out Millions of Flash cards for the computer to recognize them. And also, correcting it when it gets them wrong. Sounds like a pile of work.
Apart from objects, these guys were put to the test of making this computer aware of circumstances as well. Things like types of buses, for example to stop when it’s a school bus and not to when it’s just a regular passenger bus, pulling a side when there’s an ambulance and things of the sort.
There you have it guys, if you think you have that paper, begin warming up. Hope it hits the market soon. Keep it here at GURU8 for more updates on CES 2016