Walking is harder than playing chess. That’s strange coming from an adult, especially one who seems to be perfectly ambulatory. But it was perfect in the context of the recently concluded Sensors Expo, in the heart of Silicon Valley, as the subject turned toward the difficulty of implementing artificial intelligence (AI) in a practical way.
It all starts with sensors, those tiny electronic devices that can be the eyes, ears and touch of a machine. In fact, given the abilities of sensors to sense almost everything, it’s a disservice to the sensor industry to compare them to human senses, those being limited to five. Sensors are employed to detect magnetic fields, vibration, the smallest strains, radar, infrared and microwave radiation human eyes cannot, and sound out of earshot or frequency range. If they are all put into a machine, such as a fully sensored car, it makes a super-aware being.
There are so many sensors in self-driving cars that an entire day of the Sensor Expo is devoted to them. Light detection and ranging (LiDAR) vendors battled it out with each other and radar, and visible and invisible light cameras. The move to driverless cars may well be a boon for sensor vendors. New technologies and patents abound. Big auto companies, which have always built their cars for drivers, have to face a future where drivers may be optional. They are testing their own fleets and developing their own technologies.
AI – A Lot to Learn
As technophiles swoon for AI, investment money pours into AI technology without any sort of promise. Startups are inventing ways to add AI to their products, or at least dropping “AI” into their pitches, anything to get venture capitalists to open their door.
Drunk with the success of a few high-publicity wins (beating humans at chess, go and every other game) as well as business processes (credit card checks and tax returns), the flush-with-investment AI industry now faces the reality of everyday tasks, (walking and gripping, telling a dog from a cat, putting one mechanical foot in front of the other) and finds itself wanting. This is not easy, AI would say, if it could talk. Dumb, sinceit thought it was so smart.
At the Sensors Expo, attendees heard from a PhD from MIT, Stanford grads and other geniuses solving problems that we weren’t aware were problems until we attempted to program machines to solve them. They are having a devil of a time with deep learning neural networks and algorithms that try to approximate the structure of human brains but have yet to duplicate the brain of a baby.
The Uber Accident
With impatient Silicon Valley investors saying “faster, faster,” a rush toward smart, smooth driverless cars was on—until one couldn’t identify a woman walking her bike and ran her over.
The Uber accident was mentioned, but the victim’s name was not, like a footnote as the page turns to the next chapter in the compelling story of technology. “We are going to have self-driving cars, dammit,” seems to be spirit of the show. Level 5 autonomy is not a dream; it’s just a matter of when. Many thought it would be here already, but now it maybe as early as2020, or 2025.
AI in automobiles may, in fact, be the most immediately realizable, significant implementation of AI, Alexa/Google Home notwithstanding—they are only silly little devices that help us shop or tell the weather. In America, it’s all about cars. Happiness is geared (pun intended) to vehicles. Peopleriot when the price of fuel goes up.
Technologists are resolute in their belief that, overall, they are working for the good. A few pedestrians or drivers may be lost along the way. The public, press and politicians will issue a hue and cry, but look at the lives that will be saved. Stats show that a Tesla operating in self-driving mode has less fatalities per mile than human-driven cars.
The person on the street may have thought the next revolution in automobiles was going to be internal combustion to electric drive, but at Sensors, that is passé and not important. Any kind of drive train can be driverless. Car interiors are shown without steering wheels. One wonders why headlights are even needed. Can’t radar and LiDAR see through darkness?
Velodyne, the most visible of all sensors, as it is the LiDAR unit on top of every self-driving car at the moment, except Tesla, was the target of every company in the LiDAR face off. Getting the most attention was Ouster, which has managed to reduce the size of the LiDAR unit from the size of a coffee can to the size of a coffee cup, and then even smaller.
Motor City Tries to Keep Up
A representative from Big Auto showed how the automobile can emulate what is on a cell phone. Plug a phone into the car, and almost all the apps can be accessed on the dashboard. Like video? Text messages? Are they thinking of people who have stepped into the road while looking at their cell phones? The most distracting apps are taken off the display, according to the representative.
Automakers, with annual upgrades of their products, struggle to stay abreast of hand-held computers, i.e. cell phones, in terms of UI, usability and, lately, desire. When a person reaches driving age, what is more likely to impress their friends and be more useful to them, the latest cell phone or a car? When the car can be summoned from the cell phone, like Uber, car ownership becomes even less desirable. Automakers seek to make their vehicles like the smartphone the younger demographic buys. Since most of the functions on smartphones involve paying attention to it, success can only be limited.
Taming the Big Rig
People love disruption, according to those who benefit from it—the purveyors and financiers of technology. Tesla taking on big auto is only one example of Silicon Valley bent on disruption, eager to launch waves of services that have disrupted normal, established activities and method. Take Amazon for shopping, Uber for taxis, Airbnb for hotels and Netflix for television and movies—all are fueled, at least initially, on venture capital.
Ripe for disruption, even more than passenger vehicles, are commercial vehicles, and not just robo-taxis. What may be the scariest of all road risks is the fully loaded tractor-trailer: 80,000 lbs barreling down the highway, getting bigger and bigger in the rear view mirror. This may be relegated to nightmares, thanks to mandatory collision detection and avoidance systems, according to one company that outfits semis with such systems. Attendees also heard how truck drivers, annoyed at being constantly warned of crossing lane markers, have taken to clipping the wires to silence the warnings.