6 ms·

This might be unpopular here, but I kind of wish this and other similar systems were more open source, or even just shared between companies. If companies are g

This might be unpopular here, but I kind of wish this and other similar systems were more open source, or even just shared between companies. If companies are going to be putting a system out there on the streets anyway, and Tesla's is comparatively more quality and perhaps thus safer (I'm not making a claim, just positing a hypothetical), then it would only be to everyone's benefit if every company didn't have to (often shoddily) reinvent this wheel every time.

nwah1 8y ago
Open source tends to work great for re-implementing technology that is already understood. It tends not to work for exploratory research, since that requires risk and ingenuity, not a designed-by-committee approach. R&D of this sort is funded best with a strong profit motive, unless the government is willing to shell out for it.
- metildaa 8y ago
  I'd disagree, for R&D of novel technology there is a lot to be said for casual contributions. If ML was open source, the state of the industry would almost assuredly be better than the current state of things.
- Lorkki 8y ago
  I'm not sure how that follows. Most big open-source projects are in practice owned and controlled by a single company/entity, and the processes don't differ much from what you'd see in-house. What you're describing sounds more like how standardisation bodies work than the realities of functioning software projects.
  - nwah1 8y ago
    I guess the corporation, in this case, is acting as a benefactor just as the government would be in other cases. 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.
- ddoolin 8y ago
  I'm not disagreeing with you at all really (not sure I agree completely, either), which is why I see the need to approach cases like this (self-driving software sharing, not the specific theft of IP case at hand) a bit more specially. I don't have the solution, I just know that lives could potentially be at risk and if the only reason is profit motive, it kinda sucks. I know this happens in other places, too, but it's my opinion that the best safety should be the standard, not a premium. 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?
- cortesoft 8y ago
  > since that requires risk and ingenuity 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.
- jeltz 8y ago
  I would not describe many open source projects as designed-by-committee, and while open source might be less suited for exploratory research there have been quite many novel things first implemented in open source projects. Look at the various video codecs or the innovations in compiler technology gcc has contributed to, or the plenty of innovations made in various open source databases.
  - Xorlev 8y ago
    Much of the Apache foundation projects come to mind when thinking about the GP's assertion, though certainly it doesn't hold for all projects: many are fairly novel and exploratory.
    - nwah1 8y ago
      There are a number of them like git and bitcoin which start out with one guy building a novel algorithm and then releasing the source to the world afterwards. 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.
lefstathiou 8y ago
Yes, I wish everything was free too.
dessant 8y ago
There needs to be a shift in how proprietary technologies that have universal benefits for our society are viewed. 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.
- Quenty 8y ago
  U.S. specific, but this is what the patent system was originally designed to solve, and is a power given to the U.S. government. “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.“
- hfxtovan 8y ago
  Similar to GPS
- twblalock 8y ago
  Sharing technology would reduce death rates. But if technology was not shared, competitors would need to reduce death rates even further in order to compete, which might be better for society overall.
  - dessant 8y ago
    They'd also have an easier time further reducing death rates by getting access to tech from competitors sooner, and improve from there.
tfehring 8y ago
I completely agree, but it would be hard to incentivize investors to pay for all of the developer time (and, in the case of autonomous vehicles, driver/vehicle time) needed to get the systems up and running without the potential for huge returns to equity. 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.
linuxftw 8y ago
Simple enough to implement. The public owns the roads, the public sets access restrictions for using those roads. Just need a 'no code no road' law.
DeonPenny 8y ago
But that isn't what would happen. What would happen is tesla would go out of business because it's no better than it's competitors and because the industry is perfectly competitive no one would have money to reinvest and the tech would never advance.
basseq 8y ago
Counterpoint strawman: keeping these systems proprietary might also drive more innovation, as companies "re-inventing the wheel" will come up with different solutions or features that will continue to push the bleeding edge of the industry.
gnode 8y ago
The source code of this system may not be all you'd need, or even the most valuable or competitively differentiating element. As I understand it, the system is based on machine learning with a large training dataset gathered from customers (the "fleet learning" technology). 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-data-fleet-learning 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.
asdff 8y ago
The minute anyone sells a self driving car it will be completely dismantled and quickly reverse engineered by every auto manufacturer, like they already do for every product their competitors put out. The reverse engineering even happens before your product is out. Go to any auto show and you will find employees of other companies taking measurements or scanning paint pigments of concept cars that will never even hit production. You loose your proprietary advantage the minute you sell your first product.
- HeWhoLurksLate 8y ago
  Aaand? That doesn't make it legal to take internal data about how the car itself was made. Reverse engineering is a complicated process, and one that is costly, and thus makes it risky to do- especially when we're talking ginormous batteries one errant screwdriver away from discharging- and is why it's generally acceptable to reverse a vehicle and not acceptable to take internal documents. 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.

nwah1 8y ago

Open source tends to work great for re-implementing technology that is already understood. It tends not to work for exploratory research, since that requires risk and ingenuity, not a designed-by-committee approach. R&D of this sort is funded best with a strong profit motive, unless the government is willing to shell out for it.

metildaa 8y ago
I'd disagree, for R&D of novel technology there is a lot to be said for casual contributions. If ML was open source, the state of the industry would almost assuredly be better than the current state of things.
Lorkki 8y ago
I'm not sure how that follows. Most big open-source projects are in practice owned and controlled by a single company/entity, and the processes don't differ much from what you'd see in-house. What you're describing sounds more like how standardisation bodies work than the realities of functioning software projects.
- nwah1 8y ago
  I guess the corporation, in this case, is acting as a benefactor just as the government would be in other cases. 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.
ddoolin 8y ago
I'm not disagreeing with you at all really (not sure I agree completely, either), which is why I see the need to approach cases like this (self-driving software sharing, not the specific theft of IP case at hand) a bit more specially. I don't have the solution, I just know that lives could potentially be at risk and if the only reason is profit motive, it kinda sucks. I know this happens in other places, too, but it's my opinion that the best safety should be the standard, not a premium. 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?
cortesoft 8y ago
> since that requires risk and ingenuity 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.
jeltz 8y ago
I would not describe many open source projects as designed-by-committee, and while open source might be less suited for exploratory research there have been quite many novel things first implemented in open source projects. Look at the various video codecs or the innovations in compiler technology gcc has contributed to, or the plenty of innovations made in various open source databases.
- Xorlev 8y ago
  Much of the Apache foundation projects come to mind when thinking about the GP's assertion, though certainly it doesn't hold for all projects: many are fairly novel and exploratory.
  - nwah1 8y ago
    There are a number of them like git and bitcoin which start out with one guy building a novel algorithm and then releasing the source to the world afterwards. 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.

lefstathiou 8y ago

Yes, I wish everything was free too.

dessant 8y ago

There needs to be a shift in how proprietary technologies that have universal benefits for our society are viewed. 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.

Quenty 8y ago
U.S. specific, but this is what the patent system was originally designed to solve, and is a power given to the U.S. government. “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.“
hfxtovan 8y ago
Similar to GPS
twblalock 8y ago
Sharing technology would reduce death rates. But if technology was not shared, competitors would need to reduce death rates even further in order to compete, which might be better for society overall.
- dessant 8y ago
  They'd also have an easier time further reducing death rates by getting access to tech from competitors sooner, and improve from there.

tfehring 8y ago

I completely agree, but it would be hard to incentivize investors to pay for all of the developer time (and, in the case of autonomous vehicles, driver/vehicle time) needed to get the systems up and running without the potential for huge returns to equity. 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.

linuxftw 8y ago

Simple enough to implement. The public owns the roads, the public sets access restrictions for using those roads. Just need a 'no code no road' law.

DeonPenny 8y ago

But that isn't what would happen. What would happen is tesla would go out of business because it's no better than it's competitors and because the industry is perfectly competitive no one would have money to reinvest and the tech would never advance.

basseq 8y ago

Counterpoint strawman: keeping these systems proprietary might also drive more innovation, as companies "re-inventing the wheel" will come up with different solutions or features that will continue to push the bleeding edge of the industry.

gnode 8y ago

The source code of this system may not be all you'd need, or even the most valuable or competitively differentiating element. As I understand it, the system is based on machine learning with a large training dataset gathered from customers (the "fleet learning" technology). 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-data-fleet-learning 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.

asdff 8y ago

The minute anyone sells a self driving car it will be completely dismantled and quickly reverse engineered by every auto manufacturer, like they already do for every product their competitors put out. The reverse engineering even happens before your product is out. Go to any auto show and you will find employees of other companies taking measurements or scanning paint pigments of concept cars that will never even hit production. You loose your proprietary advantage the minute you sell your first product.

HeWhoLurksLate 8y ago
Aaand? That doesn't make it legal to take internal data about how the car itself was made. Reverse engineering is a complicated process, and one that is costly, and thus makes it risky to do- especially when we're talking ginormous batteries one errant screwdriver away from discharging- and is why it's generally acceptable to reverse a vehicle and not acceptable to take internal documents. 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.