| Filling in the Gaps Amazon used the net and massive computing power — resources available to all — to build an outstanding customer interface and the algorithms it uses to figure what products to push to consumers (creative user). But Amazon's real triumph (even before AWS itself a brilliant example of listening to the technology) was that it built the best warehouses in history — and the robots to man them. Google's search software triumphed across the internet. But it succeeded only because of unique — and contrarian — decisions about how to build its data centers, including early and expensive use of flash memory. Ping An transformed itself from a dowdy insurance company into a zero to one provider of medical services, smart city traffic control, and financial platforms because it solved one fundamental technological problem after another when off the shelf solutions were not available. Now it might seem pretty easy to figure out when you need to be a supplier and when you can be a creative user. Companies get it wrong because they go tone-deaf to the technology, which causes disastrous results. In artificial intelligence(AI), Google, Tesla, and maybe a dozen more wannabes perceived the promise of autonomous cars. Get enough data to the AI fast enough and it will drive the car better than any human. Were these contenders listening to the technology? Absolutely. AI can enable autonomous cars and there really is no way to achieve autonomy without it. Only with massive parallel processing enabled by digital circuit densities and processing speeds practically impossible just a few years ago could we even imagine a safe self-driving car. Moore's Law, which predicts the periodic doubling of computer power on a chip, could almost be said to have put a date on the plausibility of autonomous cars. Yet so far each of these contenders has failed in the autonomous project and I believe will continue to fail unless they radically change their strategy. Curiously, they are failing because they overestimated the technology — or one part of it. Lesson Learned: Don't Overestimate Tech Enthralled by the power of digital processing (the AI) — all these pioneers slighted the one ability the DMV requires of every driver, the ability to see. Vision starts in the analog, sensory world. All these early systems used available, off the shelf sensors and then relied on the AI to make up for the lost data. The sensors, called LIDARs, for light detection and ranging (a sort of optical radar) were not and still are not good enough. To compensate, most contenders use multiple LIDARs and process their combined input at the AI. The result is a near-sighted kluge that's slow and dim. What was needed was a new, far more powerful LIDAR that was integrated right onto the same chip as the processor. That way better data to start and low latency because all the processing was done right on the chip. It did not exist. It took years to create. The company that did it, called Luminar (unfortunately it's still private) will dominate the field. Because it listened to the technology. It did not try to force a fully digital solution — the AI — onto an analog problem. Listen to the technology. Sound hard? That's OK. We can listen together. We may hear some beautiful things. Regards, 
George Gilder
Editor, Gilder's Daily Prophecy | 
Dear Daily Prophecy Reader,
To end your Gilder's Daily Prophecy e-mail subscription and associated external offers sent from Gilder's Daily Prophecy, 
No comments:
Post a Comment