| Jay Valentine Sets Things Straight He points out: “Software engineers once delivered purpose-built apps based on knowledge of how a CPU works. They wrote applications that pipelined data in its most efficient form.” Luminar’s systems for advanced autonomy for automobiles move crucial software, once relegated to remote data centers, into the vehicle itself, integrated with the chip. Such advances portend serious problems for old paradigm software in the age of big data and its gigantic towers of hierarchical programming layers. These structures survive only because of the biannual doublings of hardware cost effectiveness wrought by Moore’s Law. The fabulous advances in chip technology enabled the FANG companies (Facebook, Amazon, Netflix, Google) to bail out their increasingly counterproductive software towers with ever faster and more parallel computer architectures. As Valentine points out, a typical software stack in these companies begins from the hardware up with general purpose layers, comprising “a data management layer, middleware, virtualization, security, app code, a user interface.” Because the layers are general purpose, they must accommodate every feature, “95% of which no one customer needs. But they remain and must be supported.” “Each layer introduces I/O [input-output] wait states. Every I/O wait state means the CPU is wasting time not doing anything productive; the application is flailing. A flailing app eats up energy and compute resources, with no productive business outcome.” Now we are reaching chip geometries in the low nanometers where quantum tunneling begins to occur. TSMC (Taiwan Semiconductor Manufacturing Company) and Samsung in Korea are aiming to move to 3 nanometers and below. But this amazing parade of miniaturization marks the end of the line for the Google strategy. They will no longer be able to compensate for ever more bloated software stacks with ever denser and higher speed systems on a chip. Valentine points to the emergence of “System Oriented Programming” and microservices. “A microservice makes code reusable, cuts programming time, takes advantage of containers, yet runs with pretty much the same efficiency as the technology it replaced.” Furthermore, “System Oriented Programming delivers a micro APP… with the full tech stack it needs to operate built in. The data management, middleware, security, and even GUI are purpose-built for that micro-app. If the micro app needs to manage cell-call billing, its data management understands how cell phones generate transactions. If it needs to manage an electric meter with unique types of data feeds, that comes as part of the data management layer. The result is efficient use of the machine processor. Reducing I/O wait states makes a micro app run 1,000 to a million times faster.” Such claims of thousand or millionfold acceleration undermine his argument. But we can agree that “the disruption coming to software is on par with the development of the cell phone, the graphical user interface, and the personal computer.” Valentine cites a cell phone billing system as an example. “Right now, it costs a billion dollars a year to run. It took 20 years to build. It takes 25 days to process 50 million bills in a data center so large you need a golf cart to traverse it. It consumes the energy equivalent of 100,000 homes. Using new software paradigms, they rewrote it in 45 days. It produces bills for 100 million customers in minutes, not days. The data center: a small group of inexpensive computers on a dining room table. Cost a billion dollars? Nope. Ten thousand dollars for the computers, excluding the dining room table. Energy usage? Less than charging a Tesla.” Valentine says this development ultimately dooms Facebook, Google, and Amazon to become be “the highest-cost producer.” Top U.S. Economist George Gilder Makes Big Prediction Today’s Prophecy To move to a new software paradigm, they would need to rewrite their entire systems. “They just cannot get there from here,” he says. Enabled will be new Facebooks and other social networks. “Anyone can now build a community if he has essentially free, or almost free technology and a user base.” “The hurdle of raising tens of millions of dollars for massive data centers and armies of programmers is over. Any group of competent developers can now deliver a Facebook-like community, running 100 million users in months.” In Life After Google, I anticipated something like this, telling the story of Lambda Labs and Stephen Balaban funded by the same 1517 ventures who endowed Austin Russell and Luminar. Lambda is pioneering AI data centers on the edge of the network and offer a further example of the kind of disruption described by Valentine. This is a new era in information technology and it continues to accelerate. Regards, 
George Gilder
Editor, Gilder's Daily Prophecy | 
Dear Daily Prophecy Reader,
No comments:
Post a Comment