The quotes following that are what annoy me about journalism today. No one is saying "never test around humans!". It's that wouldn't it be nice to put, say, 50 autonomous vehicles in a controlled city-scape and see how they behave? Run them 24/7 and see how long it is until an accident occurs. Now add in 10, 20, 30, etc human driven cars and repeat the experiment.
If you're going to develop technologies for humans living in cities, testing said technologies absent of human interaction would likely be misleading.
It's like YC advocating not talking to your users while you develop and test your product.
An example they gave was self-driving cars. Are there sufficient challenges in the actions that don't involve human? It seems to me that the challenges that do involve humans are much greater in scope and importance.
Why do we need a new corporation-town?
Am I wrong?
http://www.theverge.com/2015/8/31/9233435/google-self-drivin...
What percentage of possible scenarios involving getting on and off a highway do you think autonomous vehicles have encountered so far? I would guess the number is quite low.
Now extrapolate that across a city. And the point isn't that it is not possible to test these scenarios in the real world, but that it would be far easier to do so when you can control scenarios to a greater extent.
The bigger point: When you create a new version of software, do you re-run your existing tests on it, or just play around with it in production and call it good enough? The real world is production. This would be your QA environment.
My point is that these things have already been tested on a much larger scale than this QA (I actually think it was billed as Research) facility would seem to provide.
It's being used by several different manufacturers apparently: