R&D of this sort is funded best with a strong profit motive, unless the government is willing to shell out for it.
In order to work on pie in the sky ideas, you need to hire brilliant people for long periods of time, with low certainty of success. And this is expensive, and corporations will not generally do this out of the kindness of their hearts... the stars align sometimes where the private incentives and public interests are the same.
Xerox PARC, Google X, Bell Labs, and so forth meet this description just as much as DARPA.
But the traditional direct profit motive is an important piece of the engine of progress, and relying only on this other incentive structure would cause stagnation.
The definition of open source might include such a thing, but a project that originates from a lone genius rather than a community is not what is usually meant when people speak of open source.
That is great for many things, but for something like self-driving tech, it is already being used in cars on the road. This is not just experimental, it already has people's lives on the line. That is a lot of trust in something that is not available to be audited.
EDIT: Lorkki's comment explains part of my thoughts on it well. Other companies subsidize open source project that exist at the core of their business. Is there a difference here other than scale? If so, what, and how would it change the situation?
It may be normal for a company to withold information to keep a competitive edge in a nascent industry, but it will also mean that their press releases lauding reduced death rates in their products will be double edged swords, unless they make the ethical decision to share their improvements with other players and reduce death rates across the industry.
“To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries.“
As I see it, there are two ways to avoid the wasteful duplication of effort that we're seeing today: Either a private monopoly pays for it, or the government does. Both options (and the status quo) have their pros and cons, but I'd personally prefer that the government fund it - ideally, borrow to fund R&D now, and then pay off the debt through taxes on autonomous vehicles once they go live. In practice, we're definitely trending toward the first option, one implication of which is that open source is a pipe dream.
For example: knowing that someone machined the 6061 aluminum in the engine is completely different from knowing that it took twelve tries and a five-axis mill to do so, and that the entire chassis takes twelve hours to assemble, twenty people, and fifty-three robots.
From a discussion about determining whether to interpret radar-detected objects as obstacles:
"Initially, the vehicle fleet will take no action except to note the position of road signs, bridges and other stationary objects, mapping the world according to radar. The car computer will then silently compare when it would have braked to the driver action and upload that to the Tesla database. If several cars drive safely past a given radar object, whether Autopilot is turned on or off, then that object is added to the geocoded whitelist."
https://www.recode.net/2016/9/12/12889358/tesla-autopilot-da...
It may be that much of what makes autopilot effective is just data, rather than intelligent algorithms.
Furthermore, giving away such data used to tune the system may create customer privacy concerns.
Just need a 'no code no road' law.