Tag Archives: astro teller

A Misguided Decision on Driverless Cars

On first glance, the California Department of Motor Vehicles’ recent proposal to ban the testing and deployment of driverless cars seems to err on the side of caution.

On closer inspection, however, the DMV’s draft rules on autonomous vehicles rest on flawed assumptions and threaten to slow innovation that might otherwise bring enormous, time-critical societal benefits.

At issue is the requirement that DMV-certified “autonomous vehicle operators” are “required to be present inside the vehicle and be capable of taking control in the event of a technology failure or other emergency.” In other words, driverless cars will not be allowed on California roads for the foreseeable future.

One problem with the human operator requirement is that it mandates a faulty design constraint. As Donald Norman, the technology usability design expert, has noted, decades of scientific research and experience demonstrate “people are incapable of monitoring something for long periods and then taking control when an emergency arises.”

This has been Google’s direct experience with its self-driving car prototypes, too. As Astro Teller, head of Google[x], told a SXSW audience in early 2015: “Even though people had sworn up and down, ‘I”m going to pay so much attention,’ people do really stupid stuff when they’re driving. The assumption that humans could be a reliable back up for the system was a total fallacy!”

The ramifications are more than just theoretical or technical. The lives and quality of life of millions hang in the balance.

Americans were in more than six million car crashes last year, injuring 2.3 million people and killing 32,675. Worldwide, more than 50 million people were injured, and more than one million were killed. Human error caused more than 90% of those crashes.

It remains unclear whether semi-autonomous or driverless cars would better reduce human error and lower this carnage. Thus, it is important to encourage multiple approaches toward safer cars — as quickly as possible. Instead, California has slammed the brakes on the driverless approach.

Another major problem with the human-operator mandate is that it slows testing and development of systems aimed at providing affordable transportation to the elderly, handicapped or economically disadvantaged. Millions of Americans either cannot drive or cannot afford a car. This hurts their quality of life and livelihood.

Driverless cars could enable Uber-like, door-to-door mobility-on-demand services at a fraction of today’s transportation cost. This will require, however, efficient, low-cost vehicles that do not need (nor need to accommodate) relatively expensive human drivers. It also requires empty driverless cars to shuttle between passengers. The California DMV rules, as proposed, would not allow the testing or deployment of such vehicles or fleet services.

The immediate victim of California’s proposed rules is Google. Google’s self-driving car program is the furthest along in the driverless design approach that the new rules would rein in, and its current efforts are located around its headquarters in Mountain View, CA. Google’s attempt to field a fleet of prototype driverless cars (without steering wheels) would certainly be dashed.

Other companies’ efforts might be affected, too. Will Tesla owners, for example, need to get separate DMV certification to use enhanced versions of Tesla’s autopilot feature? How about GM owners with Super Cruise-equipped cars? How will these rules affect Apple’s car aspirations?

The longer-term victim is California.

Silicon Valley is becoming the epicenter of autonomous vehicle research. Not only are native companies like Google, Tesla and, reportedly, Apple investing heavily in this arena, but the race to develop the technology has compelled numerous traditional automakers to build their own Silicon Valley research centers.

If California regulators limit on-road testing and deployment, companies stretching the boundaries of driverless technology will inevitably shift their investments to more innovation-friendly states (or countries).

The proposed rules must now go through several months of public comment and review before they are finalized. California needs to take that opportunity to reconsider its course on driverless cars.

Let’s Tone Down Hope for ‘Wearables’

There is an old line in Silicon Valley: “Never confuse a clear view with a short distance.” We should keep that in mind as we think about wearable devices such as the Apple watch that are designed, among other things, to help us monitor and improve our health. The view is crystal clear, but we’re still a long way from getting to the destination.

The vision is idyllic: Some day, a wearable device will monitor all our vital signs and relay the information second by second to a healthcare provider, where some combination of computers and doctors will monitor it. We’ll know two weeks ahead of time that we’re about to have a heart attack and will be able to head it off. Doctors, who currently spend only about seven minutes a year with the average patient, will mine the stream of information, spot chronic issues in more people and get them treatment for, say, high blood pressure. Our knowledge about health will increase exponentially because so many aspects of so many people will be tracked, and in real time.

Researchers say the change in health will be like what has happened with cars. We used to wait until we saw steam coming out from under the hood, then fix whatever was wrong. Those cars lasted 60,000 or 70,000 miles. Now we have sensors all over the place in cars, learn about problems before they become acute and gather voluminous data on what works and what doesn’t, so cars can keep getting better. As a result, many cars last more than 200,000 miles. With people, once we can get those sensors “under the hood,” we should also see huge improvements in health and life expectancy – engine performance, too.

But three major things have to happen before we achieve that idyllic vision, and only one is even close to reality.

The one change that could at least plausibly happen soon is that people adopt wearables en masse. No more of this buy a Fitbit, wear it for a couple of months and then set it aside. At least millions of people, and maybe tens of millions, will have to buy wearable devices and keep them on 24/7 for basically forever, just to really get the movement started. That sort of adoption will require smaller and better-designed wearables and far better battery life – the early line on the Apple watch is that it won’t even go a full day on a charge. Makers of wearables will also have to agree on standards so that all health data can be integrated into any software and analyzed by any healthcare provider. At the moment, every wearable maker wants to own the standard, and standards fights can take years to sort out, but with Apple working its magic on consumers and with Microsoft introducing a well-regarded device, it’s at least possible to imagine mass adoption within a few years.

That’s the easiest problem.

The most severe problem is that wearables aren’t yet close to collecting the really useful information. Wearables can monitor your pulse and provide a reasonable estimate of how many steps you take, but that’s not the good stuff, as far as medicine is concerned.

I got a tutorial on this almost 15 years ago from Astro Teller, who cofounded Body Media, a pioneer in the wearables field. He said the data he really needed was blood pressure and information from blood tests. Astro is a seriously smart fellow – the grandson of the principal developer of the hydrogen bomb, Edward Teller, Astro has since 2010 been directing the Google X laboratory, meaning he has the Google Glass, driverless car and many other cutting-edge projects reporting to him – but Body Media never cracked the code before being acquired by Jawbone for $110 million, principally for its patents, in 2013. While there are glimmerings of progress all over, no breakthrough seems especially close.

Google, for one, has a project in the Google X lab that puts sensors in contact lenses that can measure blood sugar and send a constant, wireless signal to a wearable device, giving diabetics a noninvasive way to monitor themselves. But the technology must now be calibrated for different conditions. What if the wearer is crying? What if the weather is dry? What if it’s raining? It’s not clear how close to market the technology is.

Others talk about having people swallow sensors that would roam the bloodstream and report on all kinds of conditions, including watching out for cancer, but those are far enough out that they still read like science fiction.

A company has a prototype of a device that would measure blood pressure constantly, but, even if that proves workable, the device needs to go through multiple iterations and become tiny enough that it can fit into a general-use device – people may wear one health-related device on an arm, but they won’t wear two or three or four.

The final hurdle that has to be cleared is doctors and other practitioners. When I talk to doctors about the idyllic vision for the future of healthcare, they look at me like I have two heads. They’re feeling swamped just trying to keep up in a world where they see the average patient a few minutes a year, and their problems will only get worse if talk of a physician shortage proves true. Now we want them to go from seven minutes a year to 525,600 (the number of minutes in a year) for each patient? Yeah, right.

Even if doctors and other practitioners sign up for this new world of healthcare, every support system will have to change. Computer systems will have to be set up to do the vast majority of monitoring. Software will have to be written. A new class of data analysts will have to be developed. Health practices will have to reshape themselves around data streams. Insurers will have to adjust coverage. Courts will have to sort out where liability for mistakes falls – with a programmer, a doctor, someone else?

You could start the clock now on all these changes in medical practices, and they’d still take years to sort out.

The key issue to monitor in the progress of wearables is the sensors. Once someone can easily capture blood pressure information or conduct some important blood test without breaking the skin, well, then we’re talking. At that point, consumer adoption will be a solvable problem. So will adoption by medical professionals, though that will be a long slog.

In the meantime, we will soon be able to buy our Apple watches, and we’ll have fun with them. We might even get a little healthier if we keep wearing the things and somehow feel the need to walk a bit more. But that shiny vision of a world where care, insurance and everything else about health changes because of wearables? That’s still a long way out there.