Surveillance and collaborating with Chinese authorities to identify "undesirable" people is the problem. Google is being allowed to trade ratting out people in exchange for money.
That's a pretty strong claim. Got anything to support it?
1: When the company sells the car to you, why should the data the car generates remain their property?
Do you want them to be liable if the self-driving system causes an accident? Since they need that data both to evaluate liability and to improve their system to reduce their liability, that seems like a reasonable trade. If you want to own the data, fine, but then they're not liable for any accidents caused by the system.
Did they really have enough logs to confirm there wasn't a leak?
Or are they saying that based on what logs they had available they could see no leak?
The second one. The first is not possible in practice: How can you know your logging code is bug free? It is in a system with a security hole, so claiming it is flawless is hubris.
It was not a "system with a security hole" in the normal sense of the phrase. It was a system with an API that provided access to too much user data.
Again, why would a rules based system even consider those variables?
Today's AI systems are not rules-based, they're sophisticated pattern matchers, and the training process gives you no control at all over what variables they choose to key on. Obviously they can't consider data you don't provide, but they can find subtle, non-obvious patterns in everything you do provide. That's actually the goal here: to find non-obvious patterns that humans can't easily see. Of course, what they want is the non-obvious patterns that indicate that this would be a good person to hire. But since you don't have a way to directly train it on that goal, you must instead train it on what you have: resumes of successful employees. But if nearly all of those are men, then it's perfectly possible that the pattern matcher concludes that non-obvious patterns associated with maleness should be highly rated. And this is true even when you remove every obvious indication of maleness.
Perhaps a solution is to explicitly gender-bias the input. Include all the high-performing female employees in the input data and only a small number of male employees so that your sample is 50/50 -- or even biased towards women. But my guess is that the people at Amazon are plenty smart enough to have tried that idea, and that for some (perhaps inscrutable) reason it didn't work.
Manufacturers can install their own app stores, and many do, but they need the Play Store or it's not Android and doesn't get updates from Google.
This isn't correct.
Any device that passes the Android Compliance Test Suite (CTS) and complies with the Compliance Definition Document (CDD) is "Android". Also, all Google updates go into AOSP, into both the master branch and backported into the relevant dessert release branches. This is true for all devices that use AOSP, whether CTS/CDD compliant or not.
In order to put the Play Store on their device, device makers have to pass an additional test suite (GTS) and comply with additional contractual agreements, which currently includes the pre-installation of Google apps. That's highly desirable because the Play Store is really important to consumers, but it is not necessary for a device to be "Android", nor to get updates.
none of them are remotely worth giving up having a single copy accessible on all devices and by all relevant people.
single copy editing is lame. If you work as a software developer at Google how can you not know about version control systems? They are far more powerful than simple collaborative editing of shared files. This has been a solved problem in software development for decades now. If collaborative editing was all you needed then nobody would be using version control systems like Perforce or Git.
Obviously I use VCSs all the time for software development, and I have been experimenting with an approach for using a DVCS for management of legal codes (legislative bills are just patches, and a branch-and-pull approach would provide a nice origin and edit history) and similar. However, typical document editing has a different sort of workflow, less formal, with different needs. The editing process needs to be lightweight and synchronization automatic; I think DVCS is an anti-pattern for this. I have used git for collaborative editing of LaTeX documents, and it's really not as smooth or nice as something like Google Docs.
Actually, DVCS-based management does sound ideal for ISO WG documents, though. Better than Google Docs / Office 365. Single copy editing might be good for rapid collaboration, but formal updates to the official working draft should ideally be commits.
Heh. I decided to create a OneDrive account (I have an Office license for the ISO committee work I'm doing) and give it a try. First doc I uploaded and tried to view/edit in a browser, I get an error "Sorry, this document can't be opened for editing." The view mode looks fine, though it took a long time to load. I suppose browser-based editing is probably a second-class citizen given that most people use a desktop Office install, so maybe that's understandable. Not a good first look, though. If it worked well, I'd like to delete the Office installation on my laptop and just use the browser-based view. That would let me use Office docs on my Linux desktop as well as on my MacBook... and on a Chromebook as well, without need for the Android app.
However, since the view mode works, this does add immediate value for me: I can link ISO working group documents from my GSuite design docs without having to first generate PDFs. I'll also experiment with collaborative editing with other WG members. That could be a big time-saver if I can evangelize its use within the WG.
Yeah sort of like Office 365 except without all the usefulness of Office.
If Office 365 provides the cloud repository and collaborative editing features of Google Docs plus the richness and power of Office, it is indeed the best possible solution. I've never used Office 365 so I can't comment on whether it does.
Oh, one caveat: To be the "best possible solution", you also need solid cross-platform support, including on mobile devices. Not sure what Office 365 provides there.
Yeah sort of like Office 365 except without all the usefulness of Office.
If Office 365 provides the cloud repository and collaborative editing features of Google Docs plus the richness and power of Office, it is indeed the best possible solution. I've never used Office 365 so I can't comment on whether it does.
As much as I agree with your opinion on shared documents - they are a complement to the tools available on rich clients, not a replacement for them.
For me they're a replacement. I am happy to do without a few obscure features in order to have the benefits I described.
I am currently editing a grant proposal and will load the final version into OO in order to polish it. Google Docs is not able to even do proper hyphenation - this and much more will be left to OO.
OO doesn't do proper hyphenation either, though I'll grant it's a little better than Google Docs. If you really need polished output there's basically no alternative to LaTeX.
Not to mention that everyone can edit (if allowed), perhaps simultaneously, with all changes tracked, including who made them. And that it works on all major operating systems and device types.
My local university went full-on GSuite a few years ago. Maybe a little because that's what students are used to, but mostly because it's more productive for staff and easier for the university to manage. Oh, and I'm told it's much, much cheaper, too.
Collaborative editing has its uses but I wouldn't call that a killer feature. Which is probably why non cloudy variants haven't taken off yet, because it's not the architecture that is stopping it.
Collaborative editing is the killer feature, especially in the workplace. Actually, perhaps that's the second killer feature, right behind the ability to have a single copy of a doc that is accessible from all devices, by all interested people.
I was converted years ago (shortly after joining Google, actually, though the same events could have happened anywhere) during a design review meeting. I presented my design for the implementation of a new software feature to a group of other engineers. They shredded it, in a good way, providing many significant improvements and simplifications. Normally, this would have meant that I'd have left the room with a lot of work I needed to do, to document all of the changes. But during the meeting, eight people were simultaneously editing my design doc so by the time I left the only thing I had to do was to clean up some inconsistencies and polish the language a bit. What would have taken hours took less than 10 minutes.
In the years since, I've come to rely so heavily on collaborative editing that I cannot imagine going back. Even though the "collaborative" part is often sequential, having a single shared, cloud-based copy of the doc to pass around between people is fantastically better than emailing copies, tracking the most recent version and perhaps integrating changes from multiple copies. And not just at work, but at home as well. Whether it's kids wanting my thoughts on their school papers (I never edit directly, only add comments), my wife wanting me to edit the annual Christmas letter, a shared spreadsheet I built to track the distribution of my father in law's estate (my wife was the executor)... it's unbelievably better to have a shared document in the cloud. In every case. I can't think of a single time in my personal or professional life that I'd have preferred to keep separate versions.
I said I can't imagine what it's like to go back, but that isn't actually true. I don't have to imagine it. I recently joined a couple of international standards committees that still exchange documents the old way. Even with a shared document repository (iso.org web portal) it is still so painful to handle document sharing and versioning. We end up with dozens -- and I'm sure eventually hundreds -- of separate files that represent stages in the draft standard, not to mention an order of magnitude more documents containing comments and suggested changes from all of the participants. Also, because documents are too non-interactive for discussion, there are volumes of separate email threads about all of the above documents. It would be dramatically more efficient to have a single shared doc that allowed collaborative editing and in-doc comment and discussion threads. Google docs retains full version history so important "checkpoint" versions can be labeled for posterity, and of course all of the discussion on comments is retained.
Even for documents that I create on my own with no collaboration of any sort (though that's actually really rare) I prefer cloud-based docs, because then they're always available on all of my devices, or any other device I might use. I enable offline editing on all of my devices as well, so that's not a problem either -- though I'm really not often offline. Overseas flights and camping in the mountains are about the only times I don't have a network connection.
Yeah, there are a few features that office software packages have that their cloud-based versions lack, but none of them are remotely worth giving up having a single copy accessible on all devices and by all relevant people.
He repeatedly claimed in his testimony on Friday that the people Dr. Ford said were present had denied that such a party had ever occurred. This is a lie. What they said, as read into the record, was that they didn't recall such a party. In everyday speech the difference between absence of evidence and evidence of absence is often not called out, but a judge absolutely and thoroughly understands the difference, and Kavanaugh knew the stakes and that he was under oath... but he chose to mischaracterize their statements to his benefit. That constitutes a plain lie.
There are others, but that one is enough, IMO.
Several other aspects of his testimony disqualify him for SCOTUS as well.
How would you feel about losing the job of your lifetime to an accusation?
If he loses the job (which seems unlikely at this point), the reason will be less about the accusation and more about his testimony before the Senate. It was blatantly partisan, often belligerent, regularly evasive and misleading, and a few times outright false (e.g. his repeated claim that others Dr. Ford said were present said that no such event ever happened, when what they actually said was that they didn't recall it. A judge should understand the very important difference between absence of evidence and evidence of absence).
I get that if he's innocent he has every right to be angry about his treatment. But his response demonstrated a lack of self-control and ability to objectively and clearly weigh facts that we expect of judges, especially those in the highest court of the land. Well, either that or it demonstrated a sharp political mind who recognized that outrage would play well to the Republican base, and acted on that political insight to rouse the rabble in his support -- something we really don't want in the highest court.
Above all, I think any person who really cares about the integrity of the Supreme Court at this point has to realize that Kavanaugh is a poison pill. Perhaps through no doing of his own; maybe it's totally a Democrat smear campaign. But regardless of the origin, it's been so effective that a thoughtful and careful judge, interested above all in the integrity of the institution, would at this point realize that the best thing he could do for his country is to recuse himself from the entire process.
But after Kavanaugh's display on Friday, we all know very well that he is not such a judge -- which to me is the final proof that he is not a worthy nominee.
Had Kavanaugh said from the first that he was a heavy drinker and hellraiser in his youth, that he often drank to excess and had incomplete memories of some events of that time, that as a young man -- like many young men, especially 40 years ago -- he didn't adequately respect women and that although he had no recollection, he could not completely deny that in his drunken state he might have crossed a line with the young Dr. Ford, and if so he felt very sorry for it, I'd have respected him and felt differently about his nomination. If others felt the same, and I think many would, then the Senate could have gone back to discussing his judicial record and confirmed or denied him on that basis.
And here's the cold financial logic. Let's say that the cars kill over four times as many people human drivers - one every 20 million miles (Waymo has driven under half as many so far, a lot of that with heavy sandboxing). Let's say that the average wrongful death settlement is a well-above-average $2m. So $0,10/mi. The average rate for an Uber ride in the US is $2/mi, and the average car in the US costs (incl. depreciation) about $0,75/mi to operate.
Waymo can afford to kill people.
Financially, perhaps, but if Waymo's cars are more dangerous than human-driven cars -- or even just aren't significantly safer, since every single accident will be national news -- then it will be a PR disaster that will get them shut down quickly.
Slashdot Mirror
It's true. Google has very little experience with releasing search engines. We should totally believe him.
He's obviously not talking about technical challenges.
Surveillance and collaborating with Chinese authorities to identify "undesirable" people is the problem. Google is being allowed to trade ratting out people in exchange for money.
That's a pretty strong claim. Got anything to support it?
1: When the company sells the car to you, why should the data the car generates remain their property?
Do you want them to be liable if the self-driving system causes an accident? Since they need that data both to evaluate liability and to improve their system to reduce their liability, that seems like a reasonable trade. If you want to own the data, fine, but then they're not liable for any accidents caused by the system.
Did they really have enough logs to confirm there wasn't a leak?
Or are they saying that based on what logs they had available they could see no leak?
The second one. The first is not possible in practice: How can you know your logging code is bug free? It is in a system with a security hole, so claiming it is flawless is hubris.
It was not a "system with a security hole" in the normal sense of the phrase. It was a system with an API that provided access to too much user data.
Again, why would a rules based system even consider those variables?
Today's AI systems are not rules-based, they're sophisticated pattern matchers, and the training process gives you no control at all over what variables they choose to key on. Obviously they can't consider data you don't provide, but they can find subtle, non-obvious patterns in everything you do provide. That's actually the goal here: to find non-obvious patterns that humans can't easily see. Of course, what they want is the non-obvious patterns that indicate that this would be a good person to hire. But since you don't have a way to directly train it on that goal, you must instead train it on what you have: resumes of successful employees. But if nearly all of those are men, then it's perfectly possible that the pattern matcher concludes that non-obvious patterns associated with maleness should be highly rated. And this is true even when you remove every obvious indication of maleness.
Perhaps a solution is to explicitly gender-bias the input. Include all the high-performing female employees in the input data and only a small number of male employees so that your sample is 50/50 -- or even biased towards women. But my guess is that the people at Amazon are plenty smart enough to have tried that idea, and that for some (perhaps inscrutable) reason it didn't work.
Manufacturers can install their own app stores, and many do, but they need the Play Store or it's not Android and doesn't get updates from Google.
This isn't correct.
Any device that passes the Android Compliance Test Suite (CTS) and complies with the Compliance Definition Document (CDD) is "Android". Also, all Google updates go into AOSP, into both the master branch and backported into the relevant dessert release branches. This is true for all devices that use AOSP, whether CTS/CDD compliant or not.
In order to put the Play Store on their device, device makers have to pass an additional test suite (GTS) and comply with additional contractual agreements, which currently includes the pre-installation of Google apps. That's highly desirable because the Play Store is really important to consumers, but it is not necessary for a device to be "Android", nor to get updates.
What Tesla is $40k?
I bought one for $40k.
none of them are remotely worth giving up having a single copy accessible on all devices and by all relevant people.
single copy editing is lame. If you work as a software developer at Google how can you not know about version control systems? They are far more powerful than simple collaborative editing of shared files. This has been a solved problem in software development for decades now. If collaborative editing was all you needed then nobody would be using version control systems like Perforce or Git.
Obviously I use VCSs all the time for software development, and I have been experimenting with an approach for using a DVCS for management of legal codes (legislative bills are just patches, and a branch-and-pull approach would provide a nice origin and edit history) and similar. However, typical document editing has a different sort of workflow, less formal, with different needs. The editing process needs to be lightweight and synchronization automatic; I think DVCS is an anti-pattern for this. I have used git for collaborative editing of LaTeX documents, and it's really not as smooth or nice as something like Google Docs.
Actually, DVCS-based management does sound ideal for ISO WG documents, though. Better than Google Docs / Office 365. Single copy editing might be good for rapid collaboration, but formal updates to the official working draft should ideally be commits.
Heh. I decided to create a OneDrive account (I have an Office license for the ISO committee work I'm doing) and give it a try. First doc I uploaded and tried to view/edit in a browser, I get an error "Sorry, this document can't be opened for editing." The view mode looks fine, though it took a long time to load. I suppose browser-based editing is probably a second-class citizen given that most people use a desktop Office install, so maybe that's understandable. Not a good first look, though. If it worked well, I'd like to delete the Office installation on my laptop and just use the browser-based view. That would let me use Office docs on my Linux desktop as well as on my MacBook... and on a Chromebook as well, without need for the Android app.
However, since the view mode works, this does add immediate value for me: I can link ISO working group documents from my GSuite design docs without having to first generate PDFs. I'll also experiment with collaborative editing with other WG members. That could be a big time-saver if I can evangelize its use within the WG.
If itâ(TM)s not made by google it must suck amirite
Not at all. Competition is good.
Cool. Competition is good!
Yeah sort of like Office 365 except without all the usefulness of Office.
If Office 365 provides the cloud repository and collaborative editing features of Google Docs plus the richness and power of Office, it is indeed the best possible solution. I've never used Office 365 so I can't comment on whether it does.
Oh, one caveat: To be the "best possible solution", you also need solid cross-platform support, including on mobile devices. Not sure what Office 365 provides there.
Yeah sort of like Office 365 except without all the usefulness of Office.
If Office 365 provides the cloud repository and collaborative editing features of Google Docs plus the richness and power of Office, it is indeed the best possible solution. I've never used Office 365 so I can't comment on whether it does.
As much as I agree with your opinion on shared documents - they are a complement to the tools available on rich clients, not a replacement for them.
For me they're a replacement. I am happy to do without a few obscure features in order to have the benefits I described.
I am currently editing a grant proposal and will load the final version into OO in order to polish it. Google Docs is not able to even do proper hyphenation - this and much more will be left to OO.
OO doesn't do proper hyphenation either, though I'll grant it's a little better than Google Docs. If you really need polished output there's basically no alternative to LaTeX.
Not to mention that everyone can edit (if allowed), perhaps simultaneously, with all changes tracked, including who made them. And that it works on all major operating systems and device types.
My local university went full-on GSuite a few years ago. Maybe a little because that's what students are used to, but mostly because it's more productive for staff and easier for the university to manage. Oh, and I'm told it's much, much cheaper, too.
Isn't version control a solved problem?
Not in Microsoft Office. Or any of its competitors.
Collaborative editing has its uses but I wouldn't call that a killer feature. Which is probably why non cloudy variants haven't taken off yet, because it's not the architecture that is stopping it.
Collaborative editing is the killer feature, especially in the workplace. Actually, perhaps that's the second killer feature, right behind the ability to have a single copy of a doc that is accessible from all devices, by all interested people.
I was converted years ago (shortly after joining Google, actually, though the same events could have happened anywhere) during a design review meeting. I presented my design for the implementation of a new software feature to a group of other engineers. They shredded it, in a good way, providing many significant improvements and simplifications. Normally, this would have meant that I'd have left the room with a lot of work I needed to do, to document all of the changes. But during the meeting, eight people were simultaneously editing my design doc so by the time I left the only thing I had to do was to clean up some inconsistencies and polish the language a bit. What would have taken hours took less than 10 minutes.
In the years since, I've come to rely so heavily on collaborative editing that I cannot imagine going back. Even though the "collaborative" part is often sequential, having a single shared, cloud-based copy of the doc to pass around between people is fantastically better than emailing copies, tracking the most recent version and perhaps integrating changes from multiple copies. And not just at work, but at home as well. Whether it's kids wanting my thoughts on their school papers (I never edit directly, only add comments), my wife wanting me to edit the annual Christmas letter, a shared spreadsheet I built to track the distribution of my father in law's estate (my wife was the executor)... it's unbelievably better to have a shared document in the cloud. In every case. I can't think of a single time in my personal or professional life that I'd have preferred to keep separate versions.
I said I can't imagine what it's like to go back, but that isn't actually true. I don't have to imagine it. I recently joined a couple of international standards committees that still exchange documents the old way. Even with a shared document repository (iso.org web portal) it is still so painful to handle document sharing and versioning. We end up with dozens -- and I'm sure eventually hundreds -- of separate files that represent stages in the draft standard, not to mention an order of magnitude more documents containing comments and suggested changes from all of the participants. Also, because documents are too non-interactive for discussion, there are volumes of separate email threads about all of the above documents. It would be dramatically more efficient to have a single shared doc that allowed collaborative editing and in-doc comment and discussion threads. Google docs retains full version history so important "checkpoint" versions can be labeled for posterity, and of course all of the discussion on comments is retained.
Even for documents that I create on my own with no collaboration of any sort (though that's actually really rare) I prefer cloud-based docs, because then they're always available on all of my devices, or any other device I might use. I enable offline editing on all of my devices as well, so that's not a problem either -- though I'm really not often offline. Overseas flights and camping in the mountains are about the only times I don't have a network connection.
Yeah, there are a few features that office software packages have that their cloud-based versions lack, but none of them are remotely worth giving up having a single copy accessible on all devices and by all relevant people.
You forgot a third case, which I believe is the truth: He does not know if he did the thing he is accused of.
where exactly has he been proven to have lied?
He repeatedly claimed in his testimony on Friday that the people Dr. Ford said were present had denied that such a party had ever occurred. This is a lie. What they said, as read into the record, was that they didn't recall such a party. In everyday speech the difference between absence of evidence and evidence of absence is often not called out, but a judge absolutely and thoroughly understands the difference, and Kavanaugh knew the stakes and that he was under oath... but he chose to mischaracterize their statements to his benefit. That constitutes a plain lie.
There are others, but that one is enough, IMO.
Several other aspects of his testimony disqualify him for SCOTUS as well.
How would you feel about losing the job of your lifetime to an accusation?
If he loses the job (which seems unlikely at this point), the reason will be less about the accusation and more about his testimony before the Senate. It was blatantly partisan, often belligerent, regularly evasive and misleading, and a few times outright false (e.g. his repeated claim that others Dr. Ford said were present said that no such event ever happened, when what they actually said was that they didn't recall it. A judge should understand the very important difference between absence of evidence and evidence of absence).
I get that if he's innocent he has every right to be angry about his treatment. But his response demonstrated a lack of self-control and ability to objectively and clearly weigh facts that we expect of judges, especially those in the highest court of the land. Well, either that or it demonstrated a sharp political mind who recognized that outrage would play well to the Republican base, and acted on that political insight to rouse the rabble in his support -- something we really don't want in the highest court.
Above all, I think any person who really cares about the integrity of the Supreme Court at this point has to realize that Kavanaugh is a poison pill. Perhaps through no doing of his own; maybe it's totally a Democrat smear campaign. But regardless of the origin, it's been so effective that a thoughtful and careful judge, interested above all in the integrity of the institution, would at this point realize that the best thing he could do for his country is to recuse himself from the entire process.
But after Kavanaugh's display on Friday, we all know very well that he is not such a judge -- which to me is the final proof that he is not a worthy nominee.
Had Kavanaugh said from the first that he was a heavy drinker and hellraiser in his youth, that he often drank to excess and had incomplete memories of some events of that time, that as a young man -- like many young men, especially 40 years ago -- he didn't adequately respect women and that although he had no recollection, he could not completely deny that in his drunken state he might have crossed a line with the young Dr. Ford, and if so he felt very sorry for it, I'd have respected him and felt differently about his nomination. If others felt the same, and I think many would, then the Senate could have gone back to discussing his judicial record and confirmed or denied him on that basis.
black box need to be safe from manufacturers deleting data so they can't cover up when there software fails.
Yes, I said that.
Your conspiracy theories intrigue me and I'd like to subscribe to your newsletter.
Or do you think "courts" follows your reasonable ?
Yes, courts are generally very reasonable. There are exceptions, plenty of them, but not usually around liability assignment.
And here's the cold financial logic. Let's say that the cars kill over four times as many people human drivers - one every 20 million miles (Waymo has driven under half as many so far, a lot of that with heavy sandboxing). Let's say that the average wrongful death settlement is a well-above-average $2m. So $0,10/mi. The average rate for an Uber ride in the US is $2/mi, and the average car in the US costs (incl. depreciation) about $0,75/mi to operate.
Waymo can afford to kill people.
Financially, perhaps, but if Waymo's cars are more dangerous than human-driven cars -- or even just aren't significantly safer, since every single accident will be national news -- then it will be a PR disaster that will get them shut down quickly.