To be sure, some people might poison it for personal gain. A person who embarrassed Google or another company developing autonomous robots could stand to gain by shorting their stock.
As to your list of reasons for criminal activity, have you heard of terrorism?
Finally, think of still other cases like the Iranian centrifuges.
If this research raises concern that outsourced training of AIs may include back doors, a committee of separately trained AIs that "vote" on identifying things ought to address this threat, unless somehow the same backdoor is inserted into all committee members' training, which could be guarded against.
This would also help to identify any such back doors, which could be found in an investigation whenever a particular vote is not unanimous.
Image recognition will be an important component of allowing autonomous robotic systems to function correctly. Robots will be more useful if they can recognize some thing by how they look, rather than requiring us to tag everything of interest in the real world with some secure system of correct identification. So anything that subverts image recognition raises a concern for safe and correct operation, rather than more typical computer security concerns, such as improper access control or authorization.
So why would anyone want to attack the safe and correct function of autonomous robots?
A malicious person might want to poison the training set to embarrass or hurt Google (or any other competitor making autonomous robots), more than in an effort to hurt a particular person, get around security, or damage a particular thing.
No, it's not just catching on. In New York City, for example, thousands of restaurants of all kinds have web sites as well as taking orders for delivery over the phone. That has been the norm for several years now. I doubt there is any other city that offers as much diverse food delivery at all hours as New York. Pizza, barbeque, chinese, mexican, sushi, thai, indian, italian, vegetarian/vegan, turkish, burgers, etc. can be ordered for delivery to your door in many neighborhoods. It's my sense that well over half of all the Manhattan restaurants deliver, too -- probably about 3/4 do. The average place in Manhattan that takes orders will deliver until about 10PM, and plenty deliver until midnight. Only a few (but some!) do 24 hours delivery.
In the last few years there have been an increasing number of web services that will take orders for lots of different restaurants, since most of them are not chain restaurants -- that's what the article is about. Even those have been around for a few years now. The NY Times is just catching on.
Do you find any role for Plan9 at Google? Does Linux (or Linux with whatever customizations, extensions, and metamorphoses Google imposes on it) do everything Google needs or wants out of an OS platform? Does your experience with operating systems research pay off directly in contributing to the shape of the Google platform, whether for individual machine OS's, or for co-operation and clustered operation on the network?
How does he get moderated to a "2" score for a factually incorrect "me, too" on my post, whereas I get a "1" score for a follow-on with the correction that references the article?
I totally agree that MS has every right to do this. I just don't think it's smart. That said, I suppose that the people running the MS copy shop and loading dock are probably not the most creative and original thinkers in the organization, and perhaps MS feels it's safer to have them run their operations by the book, even if occasionally the plan backfires. Again, that's totally their right, and I support that right. (FWIW, I don't think any NDA I've ever signed would have covered this situation, but who knows how they've done it?)
All that aside, I guess this issue comes down to personal boundaries. I find it moderately disheartening that a hugely successful business would either try to suppress one slightly irreverent crack, or feel so paranoid that they work hard to draw lines around what would be utterly harmless free expression in any other context. I think there's a real value to employees feeling free to express their opinions, even in public, even if the company might disagree with them, without fear of instant firing, as long as they keep truly confidential things confidential, don't go so far as to intentionally undermine or harm the business, and show basic respect. I don't think this employee failed this test, and if I were dealing with a borderline case, I hope I would do better in handling him or her by having a dialog.
First of all, NDA's often stipulate penalties. The penalty need not be firing. After all, as MS has certainly seen, firing the dude can just cause more hassle and cause his words and pictures to be more widely disseminated than they would have been had they *taken him up on his offer to take down his post*!
Now, the MS managers can console themselves with the thought "Well at least we honored the terms of our NDA". Businesses don't run on getting the letter of every rule and contract they sign honored -- sometimes it's advantageous to go for a better option. Blindly following rules doesn't make a successful business.
Secondly, how do you know there's such an NDA? You're hypothesizing it. Moreover, doesn't it strike you as remarkable that evidently this employee didn't know of such an agreement? If there is such an agreement and employees aren't aware of it, then MS has a far worse problem.
Now, I suppose one potential problem is the precedent set by "leaks". I wouldn't consider the location of the copy shop a "leak", nor would I consider the purchase of some computers a "leak", but what the hell, let's run with that idea. I still think they'd be better served suppressing the leak and giving a warning, but perhaps they feel the fear is a better motivator for others in the future.
In that case, I think MS is excessively paranoid, and doesn't appropriately draw lines around confidential vs. non-confidential information, which strikes me as dumb.
At none of the companies I worked for (all IT business related, except in two cases where I worked in IT for financial services industry firms) it would probably have been ok to take photos of a palette of computers on the loading dock. In no case would it have been a firing offense.
Candid photos of employees in the building are, of course, something different.
If I felt that something an employee posted might cause a negative perception, I'd discuss it -- I wouldn't just reach for the eject button and toss the employee out on his/her/its butt. What good does that do? It can easily generate more negativity and attention, and I can't call the words back.
I'm surprised so many people take Microsoft's side on this issue. I run a small business. I've worked for a few small businesses. At each of those places I and the people I've worked for would have had no problem with an employee putting information like this on the web. I hope that if my business ever grew to be huge, managers at the level of the copy shop would have similar perspective: it just doesn't matter. In fact, making this a firing offense probably made thing worse.
As far as I can tell, the facts disclosed are:
1. MS copy shop operations are in the same building as shipping and receiving.
2. There are trees, asphalt, and trucks somewhere on the MS campus. Sometimes there is sunshine.
3. MS bought a few Apple G5 machines.
As several people have pointed out, item #3 is no surprise, given that MS develops software for Apple computers.
What's the big deal?
Wouldn't a more enlightened company have requested that the employee go ahead with his offer to remove the text, or simply have asked that he note the preceding in a rebuttal? They can't call the information back, and does it really help MS to cultivate so much fear among employees about discussing even such innocuous details? If so, why?
A new category of spoiler
on
Decipher
·
· Score: 1
This review first piqued my interest and then destroyed it. The spoiler warning was not enough. This kind of spoiler needs a stronger warning, such as an "Every important detail of the plot revealed eliminating any possibility of suspense" warning.
} else { open LOCK,">/tmp/portald.pid" or die "Can't open/tmp/portald.pid: $!"; print LOCK "$$"; close LOCK; }
This is from the portald makeTmp subroutine. You have to use flock or fcntl to make sure of atomicity. You have a race condition between when you open the file and when you write.
Exactly. The 11 days in the interval come in this month (20 - 30 November), then the one month is December 1999, and then the 1111 years are the years 2000 - 3110. At the end of that period is 1/1/3111. Otherwise we could argue about how many days fall in a month and so on.
I just thought it was neat that there was one correct way to count the time that came out with so many ones. And it is correct.
Re:1111 years 1 month 11 days ... - Wrong!
on
Happy Odd Day!
·
· Score: 1
Sigh.
Today is Friday 19 November 1999.
There are 11 days left in November. Then there is December. Then the years 2000 - 3110 inclusive make up 1111 years. At the end of that interval is the first moment of 1/1/3111.
So there is an interval of 11 days, 1 month and 1111 years between now and 1/1/3111.
Get it yet?
1111 years 1 month 11 days ...
on
Happy Odd Day!
·
· Score: 4
That's the number of days in the interval between this odd day and the next one. Of course, it falls out from the specification of an odd day and arithmetic, (and a little luck with the number of days this month) but it's still kind of neat.
You could probably bootstrap authentication with a shared secret since you have to go to the trouble of agreeing to timing and so on. So if any two parties intending to communicate can somehow get some shared secret across to both endpoints, they can update the shared secret at the beginning of every later successfully secured connection. This new shared secret could be used to authenticate the next time. This protects against MITM to the extent you can trust both the secrecy of the original shared secret and the unpredictability of later ones. However, this shared secret can be a lot smaller than an agreed to pad, since it's only used to authenticate.
Having failed to control the availability of reasonable cryptography, the FBI wants to install a giant traffic analysis system. Some thoughts on this system:
o It represents a second best surveillance tactic after eavesdropping. If you can't tell what a person is saying, it's at least interesting to be able to tell to whom they're saying it (traffic analysis). If the Clipper chip was plan A, this sort of thing is plan B.
o This is definitely a lot bigger than what any private agency can do because they FBI can theoretically use the law to gain monitoring access at any network access point they desire, which a private entity could not, and likely would not, do.
o Federal law enforcement realizes that the public computer internetwork has become or is rapidly becoming the world's primary nexus of communication, and therefore they must be able to analyze it in order to snoop on the citizenry.
o Traffic analysis in order to hunt for "patterns of behavior that suggest illegal activity" might lead to a vague fishing expedition approach to law enforcement. Perhaps this is an attempt to do an end-run around troublesome fourth amendment protections, which are fairly well defined in the case of telephone wiretapping. Interestingly, this seems to me to require that the use of IP telephony would get far less protection from warrantless search than regular phone calls. For example, even if you encrypt your phone call, and even if you use anonymous forwarders, this type of system might theoretically allow the FBI to detect the end points of an IP telephony call, unless you handed the call off along the way to the PSTN (a normal phone company). The FBI could thus ensure that there is no reasonable expectation of privacy in telephone call end-points, which might then make such information admissable evidence in criminal prosecutions.
o The conflation of domestic and international concerns may be a new tactic in the constant pursuit of greater surveillance powers for law enforcement. I expect we'll see more of this. Theoretically these measures are for "national security", and defense against foreign attack, but that excuse is being used to justify snooping in the US. The internet, by being a global medium open to easy foreign access, may well represent the thin edge of the wedge for this sort of argument, where foreign threats are used to expand the powers of domestic law enforcement.
If a page is truly useful, likely someone is accessing it. A distributed program to harvest those pages could be quite useful. You could choose when to allow it to examine your browsing history, and when to pull back the curtain, as it were. Of course, you'd have to make privacy guarantees. You'd also want to make the source code visible to the world. If a page you were browsing was unknown to the system, then spidering from it would probably be quite productive, so the program could harvest your spare CPU cycles to spider from any pages that you visit that the search engine does not yet know about. Everyone would have an incentive to participate to make sure that the pages they want to see indexed are actually indexed.
To avoid the Netscape "What's Related?" fiasco, the authors should allow the end user editorial control, and provide for some discretion over and anonymizing of the results submission.
Slashdot Mirror
The old "sue our savviest customers" trick -- never a good sign.
Fortunately everyone loves airlines and air travel, so they've got good will to burn. Imagine if a widely detested business tried this!
To be sure, some people might poison it for personal gain.
A person who embarrassed Google or another company developing autonomous robots could stand to gain by shorting their stock.
As to your list of reasons for criminal activity, have you heard of terrorism?
Finally, think of still other cases like the Iranian centrifuges.
If this research raises concern that outsourced training of AIs may include back doors, a committee of separately trained AIs that "vote" on identifying things ought to address this threat, unless somehow the same backdoor is inserted into all committee members' training, which could be guarded against.
This would also help to identify any such back doors, which could be found in an investigation whenever a particular vote is not unanimous.
Image recognition will be an important component of allowing autonomous robotic systems to function correctly. Robots will be more useful if they can recognize some thing by how they look, rather than requiring us to tag everything of interest in the real world with some secure system of correct identification. So anything that subverts image recognition raises a concern for safe and correct operation, rather than more typical computer security concerns, such as improper access control or authorization.
So why would anyone want to attack the safe and correct function of autonomous robots?
A malicious person might want to poison the training set to embarrass or hurt Google (or any other competitor making autonomous robots), more than in an effort to hurt a particular person, get around security, or damage a particular thing.
No, it's not just catching on. In New York City, for example, thousands of restaurants of all kinds have web sites as well as taking orders for delivery over the phone. That has been the norm for several years now. I doubt there is any other city that offers as much diverse food delivery at all hours as New York. Pizza, barbeque, chinese, mexican, sushi, thai, indian, italian, vegetarian/vegan, turkish, burgers, etc. can be ordered for delivery to your door in many neighborhoods. It's my sense that well over half of all the Manhattan restaurants deliver, too -- probably about 3/4 do. The average place in Manhattan that takes orders will deliver until about 10PM, and plenty deliver until midnight. Only a few (but some!) do 24 hours delivery.
In the last few years there have been an increasing number of web services that will take orders for lots of different restaurants, since most of them are not chain restaurants -- that's what the article is about. Even those have been around for a few years now. The NY Times is just catching on.
Do you find any role for Plan9 at Google? Does Linux (or Linux with whatever customizations, extensions, and metamorphoses Google imposes on it) do everything Google needs or wants out of an OS platform? Does your experience with operating systems research pay off directly in contributing to the shape of the Google platform, whether for individual machine OS's, or for co-operation and clustered operation on the network?
Ah, interesting. No, I was not aware.
How does he get moderated to a "2" score for a factually incorrect "me, too" on my post, whereas I get a "1" score for a follow-on with the correction that references the article?
I totally agree that MS has every right to do this. I just don't think it's smart. That said, I suppose that the people running the MS copy shop and loading dock are probably not the most creative and original thinkers in the organization, and perhaps MS feels it's safer to have them run their operations by the book, even if occasionally the plan backfires. Again, that's totally their right, and I support that right. (FWIW, I don't think any NDA I've ever signed would have covered this situation, but who knows how they've done it?)
All that aside, I guess this issue comes down to personal boundaries. I find it moderately disheartening that a hugely successful business would either try to suppress one slightly irreverent crack, or feel so paranoid that they work hard to draw lines around what would be utterly harmless free expression in any other context. I think there's a real value to employees feeling free to express their opinions, even in public, even if the company might disagree with them, without fear of instant firing, as long as they keep truly confidential things confidential, don't go so far as to intentionally undermine or harm the business, and show basic respect. I don't think this employee failed this test, and if I were dealing with a borderline case, I hope I would do better in handling him or her by having a dialog.
But sure, MS has every right to fire the guy.
First of all, NDA's often stipulate penalties. The penalty need not be firing. After all, as MS has certainly seen, firing the dude can just cause more hassle and cause his words and pictures to be more widely disseminated than they would have been had they *taken him up on his offer to take down his post*!
Now, the MS managers can console themselves with the thought "Well at least we honored the terms of our NDA". Businesses don't run on getting the letter of every rule and contract they sign honored -- sometimes it's advantageous to go for a better option. Blindly following rules doesn't make a successful business.
Secondly, how do you know there's such an NDA? You're hypothesizing it. Moreover, doesn't it strike you as remarkable that evidently this employee didn't know of such an agreement? If there is such an agreement and employees aren't aware of it, then MS has a far worse problem.
Now, I suppose one potential problem is the precedent set by "leaks". I wouldn't consider the location of the copy shop a "leak", nor would I consider the purchase of some computers a "leak", but what the hell, let's run with that idea. I still think they'd be better served suppressing the leak and giving a warning, but perhaps they feel the fear is a better motivator for others in the future.
In that case, I think MS is excessively paranoid, and doesn't appropriately draw lines around confidential vs. non-confidential information, which strikes me as dumb.
At none of the companies I worked for (all IT business related, except in two cases where I worked in IT for financial services industry firms) it would probably have been ok to take photos of a palette of computers on the loading dock. In no case would it have been a firing offense.
Candid photos of employees in the building are, of course, something different.
If I felt that something an employee posted might cause a negative perception, I'd discuss it -- I wouldn't just reach for the eject button and toss the employee out on his/her/its butt. What good does that do? It can easily generate more negativity and attention, and I can't call the words back.
Uh, it says in the blog post " (MSCopy, the print shop I work in, is in the same building as MS's shipping and receiving)".
Did you RTFA?
I'm surprised so many people take Microsoft's side on this issue. I run a small business. I've worked for a few small businesses. At each of those places I and the people I've worked for would have had no problem with an employee putting information like this on the web. I hope that if my business ever grew to be huge, managers at the level of the copy shop would have similar perspective: it just doesn't matter. In fact, making this a firing offense probably made thing worse.
As far as I can tell, the facts disclosed are:
1. MS copy shop operations are in the same building as shipping and receiving.
2. There are trees, asphalt, and trucks somewhere on the MS campus. Sometimes there is sunshine.
3. MS bought a few Apple G5 machines.
As several people have pointed out, item #3 is no surprise, given that MS develops software for Apple computers.
What's the big deal?
Wouldn't a more enlightened company have requested that the employee go ahead with his offer to remove the text, or simply have asked that he note the preceding in a rebuttal? They can't call the information back, and does it really help MS to cultivate so much fear among employees about discussing even such innocuous details? If so, why?
This review first piqued my interest and then destroyed it. The spoiler warning was not enough. This kind of spoiler needs a stronger warning, such as an "Every important detail of the plot revealed eliminating any possibility of suspense" warning.
Where in there does it say that color of their skin is in any way considered over how well they perform their job? It just tallies the numbers.
By the way, there is a % white. It's very simple:
100% - % minorities.
Just like there's a % male, 100% - % female.
As far as I can tell, you're the only one crying racism.
> very bigoted, ego-centric view.
Yeah, good thing this paper doesn't say that. Pike's paper is about how systems software research is irrelevant.
Didn't they get enough complaints with prior releases?
Boy is this ever not the way to lock:
/tmp/portald.pid: $!";
} else {
open LOCK,">/tmp/portald.pid"
or die "Can't open
print LOCK "$$";
close LOCK;
}
This is from the portald makeTmp subroutine.
You have to use flock or fcntl to make sure of
atomicity. You have a race condition between
when you open the file and when you write.
Exactly. The 11 days in the interval come in
this month (20 - 30 November), then the one
month is December 1999, and then the 1111 years
are the years 2000 - 3110. At the end of that
period is 1/1/3111. Otherwise we could argue
about how many days fall in a month and so on.
I just thought it was neat that there was one
correct way to count the time that came out with
so many ones. And it is correct.
Sigh.
Today is Friday 19 November 1999.
There are 11 days left in November.
Then there is December.
Then the years 2000 - 3110 inclusive make up
1111 years.
At the end of that interval is the first moment
of 1/1/3111.
So there is an interval of 11 days, 1 month
and 1111 years between now and 1/1/3111.
Get it yet?
That's the number of days in the interval between
this odd day and the next one. Of course, it
falls out from the specification of an odd day
and arithmetic, (and a little luck with the
number of days this month) but it's still kind of
neat.
You could probably bootstrap authentication with
a shared secret since you have to go to the
trouble of agreeing to timing and so on. So if
any two parties intending to communicate can
somehow get some shared secret across to both
endpoints, they can update the shared secret at
the beginning of every later successfully secured
connection. This new shared secret could be used
to authenticate the next time. This protects
against MITM to the extent you can trust both the
secrecy of the original shared secret and the
unpredictability of later ones. However, this
shared secret can be a lot smaller than an agreed
to pad, since it's only used to authenticate.
Having failed to control the availability of reasonable cryptography,
the FBI wants to install a giant traffic analysis system. Some
thoughts on this system:
o It represents a second best surveillance tactic after
eavesdropping. If you can't tell what a person is saying, it's at
least interesting to be able to tell to whom they're saying it
(traffic analysis). If the Clipper chip was plan A, this sort of
thing is plan B.
o This is definitely a lot bigger than what any private agency can do
because they FBI can theoretically use the law to gain monitoring
access at any network access point they desire, which a private entity
could not, and likely would not, do.
o Federal law enforcement realizes that the public computer
internetwork has become or is rapidly becoming the world's primary
nexus of communication, and therefore they must be able to analyze it
in order to snoop on the citizenry.
o Traffic analysis in order to hunt for "patterns of behavior that
suggest illegal activity" might lead to a vague fishing expedition
approach to law enforcement. Perhaps this is an attempt to do an
end-run around troublesome fourth amendment protections, which are
fairly well defined in the case of telephone wiretapping.
Interestingly, this seems to me to require that the use of IP
telephony would get far less protection from warrantless search than
regular phone calls. For example, even if you encrypt your phone
call, and even if you use anonymous forwarders, this type of system
might theoretically allow the FBI to detect the end points of an IP
telephony call, unless you handed the call off along the way to the
PSTN (a normal phone company). The FBI could thus ensure that there is no
reasonable expectation of privacy in telephone call end-points, which
might then make such information admissable evidence in criminal
prosecutions.
o The conflation of domestic and international concerns may be a new
tactic in the constant pursuit of greater surveillance powers for law
enforcement. I expect we'll see more of this. Theoretically these
measures are for "national security", and defense against foreign
attack, but that excuse is being used to justify snooping in the US.
The internet, by being a global medium open to easy foreign access,
may well represent the thin edge of the wedge for this sort of
argument, where foreign threats are used to expand the powers of
domestic law enforcement.
If a page is truly useful, likely someone is accessing it. A distributed program to harvest those pages could be quite useful. You could choose when to allow it to examine your browsing history, and when to pull back the curtain, as it were. Of course, you'd have to make privacy guarantees. You'd also want to make the source code visible to the world. If a page you were browsing was unknown to the system, then spidering from it would probably be quite productive, so the program could harvest your spare CPU cycles to spider from any pages that you visit that the search engine does not yet know about. Everyone would have an incentive to participate to make sure that the pages they want to see indexed are actually indexed.
To avoid the Netscape "What's Related?" fiasco, the authors should allow the end user editorial control, and provide for some discretion over and anonymizing of the results submission.
I tried to submit this URL to slashdot, but
no one saw fit to post it:
www.realweasel.com