Every once in a while, a company comes up with something that is more amazing than the firm actually realizes. I think that was the case with the initial iPod. It was an almost hobby-like Apple product but eventually the market, and Apple, caught on that it could be so much more. Not only was the iPod critical to the firm's turnaround, but it spawned the iPad and the iPhone (a product that nearly ensures Apple's entire valuation today).
I think that is also the case with the NVIDIA DRIVE PX 2 computer. Yes, it is being applied to self-driving cars initially. But it is a computer that can see, hear, evaluate, learn and respond instantly to a massive number of sensory inputs. It is really the first commercially available brain-like product.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=iThis suggests that the applications that this very unique computer could be applied to are far broader than just self-driving cars. It could be applied to general robotics, large-scale smart buildings and even smart cities.
Let me explain.
DRIVE PX 2 Specs
The DRIVE PX 2 is a relatively small computer designed to handle a combined 2.3 Teraflops of data, blending 12 cameras, radar, lidar, and a variety of sensors into a data stream that can be applied to a decision matrix. This matrix is the result of deep learning methods that allow the computer to recognize different objects and respond appropriately to them. These objects cover virtually all vehicles, signs, lane markers, people, animals, other forms of transportation, weather, lighting and road anomalies (like potholes).
This product is built to use NVIDA DIGITS, a heavily researched set of algorithms designed to make a computer capable of making decisions in real time without human interaction. It is developed on a system that can handle 7 Teraflops of information and form the basis for training the DRIVE PX 2.
The result is a digital brain that can look 360 degrees at once across a variety of sensors and create a real-time emulation of the world around it, about which it can then make real-time decisions. And the 360-degree limit is only because that’s all a car really needs. In theory, it could also look over and under the car at the same time, as well as provide a global view of the vehicle.
So, basically, this is an all-seeing brain trained to be able to view and respond to anything the sensor can see around it, potentially in a full globe.
So what else could it be used for?
One of the most obvious uses would be military defense systems. The older Phalanx system in use for U.S. military ships has a small fraction of the processing power available to DRIVE PX 2 and can only defend against a comparatively small arc of potential attack vectors. This has more recently been enhanced with a SeaRAM-blended system of guns and missiles, providing a more comprehensive defense solution.
Updating these systems with another purpose-built computer would be prohibitively expensive but that’s what makes a general-purpose product like the DRIVE PX 2 attractive: It can be adapted to almost anything. One computer, or two if you wanted full redundancy, could cover a ship both above and below the waterline, with reaction times that a human couldn’t match. It could even include taking emergency control of the helm to safely avoid the threat. Since DRIVE PX 2 is designed to network with other vehicles, it‘s already set up to coordinate with a battle group of ships, coordinating a response between a carrier and its destroyer escort far better and far more cheaply than deployed systems.
You’d end up with a result both better and far cheaper than is in use today, and it could be retrofitted relatively easily given its small size and teachable nature.
Smart Cities, Smart Buildings
The idea of integrating actual cameras into smart buildings and smart cities is hardly new, but doing it inexpensively has been very difficult. When you talk about being able to respond in real time to problems and threats having to do with people, facial recognition systems tend to be too slow and so unique that a human needs to be in the data chain to make decisions.
However, the DRIVE PX 2 is designed to handle visual information expertly and react to it. Think of traffic lights that could recognize police or fire vehicles; systems that would automatically block traffic or shut down power in the face of related problems; security systems that could recognize someone that is authorized vs. an employee who isn’t supposed to be in an area or is doing something inappropriate and more effectively track movement across a city or through a building. These would be capabilities this system could adapt to easily.
Integrating security with building or city management systems has normally been problematic. But a system like DRIVE PX 2 could be an ideal way of making this all real and would be far less expensive than the typical highly customized approach.
Wrapping Up: And Robots
The DRIVE PX 2 can do a lot more than just make cars autonomous. Given that an autonomous car is basically a rolling robot, it seems like the next step is for DRIVE PX 2 to be applied to construction equipment, public transportation, factory floors and, as noted, defense systems and integrated security smart city/building systems. For now, however, DRIVE PX 2 is only focused on cars, which is why I think NVIDIA hasn’t yet realized just how powerful a digital brain tied to deep learning could actually be.
Rob Enderle is President and Principal Analyst of the Enderle Group, a forward-looking emerging technology advisory firm. With over 30 years’ experience in emerging technologies, he has provided regional and global companies with guidance in how to better target customer needs; create new business opportunities; anticipate technology changes; select vendors and products; and present their products in the best possible light. Rob covers the technology industry broadly. Before founding the Enderle Group, Rob was the Senior Research Fellow for Forrester Research and the Giga Information Group, and held senior positions at IBM and ROLM. Follow Rob on Twitter @enderle, on Facebook and on Google+.