Wednesday, November 26, 2014

It's not a question of being too positive or negative but of being wrong in both directions

Given some recent discussions, I perhaps ought to go back and clarify my position on Google's driverless cars. The Google part is important. Lots of companies, particularly big auto makers like GM and Nissan, are seriously pursuing this research. However, when you read a news account about autonomous vehicles, most of the time it's a story about Google which is troublesome for at least two reasons: first because there are some big concerns that are particularly applicable to Google's approach; and second because Google has a way of playing to the worst tendencies in tech reporters. The result is a standard narrative that manages to get both the pros and the cons wrong.

The official account goes something like this: from a technological standpoint, the Google driverless car is virtually good-to-go. There is every reason to expect you will be able to buy one in a couple of years; the only clouds on this horizon are concerns with safety and, more importantly, regulation. This version is extremely popular. It is regularly reported in the New York Times and the Wall Street Journal. It also gets both the pros and cons wrong.

Though it's too early to say for certain, safety appears to be the one non-issue for this technology. All transportation carries an element of risk, but based on a pretty good sample of road tests, that risk appears to be considerably smaller for Google's self-driving cars than it is for other ways of getting around.

Regulation is only a slightly greater concern We have had well over a century to work out the problems associated with insuring and regulating a wide variety of high-speed and potentially very dangerous vehicles. The idea that a viable, highly anticipated, and, relatively speaking, very safe technology will be kept off the market due to legal issues not just in the United States but in Europe and Asia is simply not believable.

So, if the concerns are not safety and regulation, what are they?

As previously discussed many times, much of the more gee-whiz reporting has been based on the idea that autonomous cars will quickly reach 100% adoption. As unrealistic as that assumption is, it pales in comparison to another jump tech reporters seem to have made.

Based on this Slate article, it appears that Google's approach to fully autonomous cars requires a specialized and highly expensive data infrastructure, specifically a collection of incredibly detailed maps. In order to compile this level of data, dedicated vehicles with human drivers have to travel the roads in question multiple times. What's more, the process needs to be constantly repeated to keep the data up-to-date. Road construction, new houses, all sorts of things need to be taken into account by the system.

The primary advantage of automobiles over other forms of transportation is their flexibility. A car can go pretty much anywhere you want. You can even decide on a new destination while traveling. In order to be viable, new automotive technology needs to keep that flexibility. Apparently Google's current approach means it would take a prohibitive amount of time and money to map out more than a tiny fraction of the nation's roads. Put bluntly, if this is true, given these infrastructure costs and be wide array of transportation alternatives, the Google's autonomous cars will never be viable in its present form.

Caveats are important here. It is entirely possible that Google is on the verge of a breakthrough that will allow its cars to operate off of existing Google maps. That would come close to making the technology viable. For all I know, the company could be preparing a press release as I write this. If this is the case, I'll pen a sheepish retraction then call friends and family members who don't drive and share the good news.

For now, though, this appears to be a very difficult technical nut to crack and, rather than showing signs of progress, Google seems to be trying to divert attention from the problem. You'll notice that their latest highly touted 'advance' was actually a step down in this respect, going from actual road tests to far less demanding closed tracks.


No comments:

Post a Comment