Because they generated the "before" shot of the training pair, not the "after" shot.
i.e.: They started with a painting by Van Gogh, then ran it through the "Smart Blur" filter of photoshop to remove the Van Gogh-esque-ness of the painting, leaving a textureless image. Then they ran the texture learning algorithm over that pair of images, and applied the learned texture to the target.
I have converted every single employee at my current company (far from an
IPO, but profitable nonetheless) to using the Google Toolbar because it saves so much time. (and
they in turn have converted most of their acquaintances...) If things continue this way, not only will
they be an attractive investment,
How does the fact that everyone uses Google mean that it is an attractive investment?
You're falling prey to exactly the same fallacy as everyone else in the mass hysteria of the 1999 IPO market did: that somehow eyeballs == good investment. It is exactly that kind of thinking that caused the bubble to hyperinflate, and then burst: "Everyone is looking at this company's web page, let's invest!" Once they realized that there wasn't actually any money there, the bubble burst.
Even, say, 100 billion hits a day isn't worth any investment at all, if there's no way to make money from them. Would you invest in Slashdot, for example? So what if lots of people look at it - it doesn't make any money. It loses money hand over fist.
I won't know if I want to invest in Google until I see an S/1 and know if they translate their technology into profits or losses. If it costs more to run their engine than they make, and that will be the case for the forseeable future, how is the investment attractive?
Automatic (or even semi-automatic) patching is the *dumbest* idea on Earth. [...] Imagine someone hacking the Patch Server, then inserting a 'patch' that contains malicious code. *BOOM* Every motherfucking machine that uses that server is then 0wned.
Why is this an argument against automatic patching? The same thing would happen to everyone who manually goes to their vendor's patch server and applies what's there. What criteria could a person applying a patch manually use to determine that the latest patch RPM is malicious, that an automatic patch script could not use?
Doesn't look like the
changes were too radical -- mostly just catching them up to current practice -- that's a lot of text that I haven't got through yet and there's surely some
gotchas in there. Does your mail client or server (or netnews client, since they
use the message format) comply?
Hello, the first paragraph of the RFC says
This document is a self-contained specification of the basic protocol
for the Internet electronic mail transport. It consolidates, updates
and clarifies, but
doesn't add new or change existing functionality
of the following:
- the original SMTP (Simple Mail Transfer Protocol) specification of
RFC 821 [30],
- domain name system requirements and implications for mail
transport from RFC 1035 [22] and RFC 974 [27],
- the clarifications and applicability statements in RFC 1123 [2],
and
- material drawn from the SMTP Extension mechanisms [19].
Re:Jeffrey B. Lotspiech at Stanford Tommorrow!
on
CPRM Voted Down
·
· Score: 1
I suspect your idea was rejected because it doesn't address the political or economic issues associated with running the Internet. It may be a good technical solution, but the Internet doesn't run purely on technology. It runs on money.
Perhaps the best example of this is the story of
Landmark Routing. Go read the paper. You probably need to read it 3 times before it completely sinks in, but it's a brilliant way of building a routing hierarchy that scales beautifully in virtually every dimension: number of end-stations, number of sub-networks, number of routers, number of everything.
The basic idea is that you route packets the same way you might ask for directions on a trip from New York to your friend's house in Los Angeles. You start out asking "which way to LA?" and someone might point you vaguely westward. Once in LA you can ask someone from LA, "where's Santa Monica?" and once in Santa Monica, ask "where's Ocean Blvd?" As you get closer and closer to your destination, you ask for progressively more and more specific directions -- and are more likely to get them because you're more likely to encounter people who know the local geography. Your route may not be optimal (i.e., because the person in New York who you first asked for directions didn't have global knowledge of LA's streets), but it works very well in a distributed way, and it scales. This is Landmark in a nutshell (but, really - don't take my word for it; go read the paper).
Anyway, as I heard it, back in the days when BGP was being standardized, Landmark was at one point suggested as an alternative. Now, my very 3rd-hand understanding is that it was ultimately rejected because it just didn't fit the model of what Internet routing algoritms need to do. It can't express policy. It can't support ISP X saying "I only want to send data to ISP Y, because I have a peering agreement with them. I don't want to advertise routes to ISP Z because they use me as transit without providing transit. I don't want to route these packets through Canadian ISPs because it violates local laws." The economic reality is that you don't always route through the shortest path, or the cheapest path, but the path through which you've negotiated peering agreements or are buying transit service. BGP lets you express these policies.
Landmark is a beautiful technical solution, and from the point of view of pure computer science, it works. Maybe your solution does, too. But that doesn't necessarily have any bearing on its usefulness in the real world. For your idea (or any idea) to be considered seriously, you have to address these issues.
Re:end of pay phones?!?
on
Paper Phones
·
· Score: 1
I hope it does mean the end of pay phones - and their replacement by phone vending machines:-)
Oops, correction! Looking at their financials and re-reading the article I see I was wrong - $232M was their revenue for the quarter, not their net income (after expenses) for the year.
Their net income for the year was only $70.8 million for the year ending 12/31/2000. That means, using my plan, they could actually increase their net income by 42%, to $100 million!
The Wall Street consensus was for first quarter earnings per share of 5
cents and revenues of $232.6 million, according to First Call/Thomson
Financial...
``Regardless of how the topline unfolds, we are committed to achieving breakeven for the full year,'' Decker said.
Koogle cited a balance sheet with about $2 billion
in cash and no debt.
Oh, come on, $2 billion is a lot of money! You're saying you can't make a measley $232 million when you have $2 billion in the bank? Your goal is just to break even?
Here's how you can stay profitable, Yahoo.
A) Discontinue all Internet operations. Turn off all routers, servers, etc. Fire everyone except your CFO. Get rid of all office space. Sell all furniture. Expenses should now be $0.
B) Send your CFO to Bank of America. For your convenience, I've even included a map that shows how to get from Yahoo corporate headquarters to the nearest branch. (Important: This is a link to Yahoo! Maps, so print this map before you do Step A.)
C) Tell the CFO to open a Prima Savings Account. $10,000 minimum balance, but that should be no problem. Deposit your $2 billion. (Note, you will also get free checking and free ATM card.)
D) Wait one year. You will be earning over 5% APR with your Prima Savings Account, just like I do. Pray that Greenspan starts raising interest rates again.
E) Celebrate! You've now made $100 million! Maybe not the $232 million that we wanted, but it's a lot better than breaking even, right?
Re:One GUI from the "Movie OS"...
on
MUD Shell
·
· Score: 3
You do know that the filesystem navigation in Jurassic Park was a real application, right? No special effect. It was a program called fsn (the FileSystem Navigator), which ran on SGIs at the time. In fact, I'm amazed to find (after 5 seconds at google) that you can still download it from SGI!
FSN is not fake, it actually looks just like what you saw in the movie. I think the Jurassic Park people added the sound effects, but the real FSN actually let you fly around a graphical representation of your filesystem, fly into subdirs by clicking on them, launch apps, etc.
AltaVista used to be pure, then went the portal route.
AltaVista did, at one point, fall from the purer faith. But then the clean, portal-less look of Google showed them the way, and they repented with Raging Search. Raging is exactly the same database and search engine as AltaVista, but with a Google-esque minimalist look. I can't imagine why anyone still uses AltaVista's front door when you actually your work done going through their back service entrance.
It seems that, 30 minutes ago (a little after midnight pacific time, early Friday morning), the Stage 3 emergency has been reinstated. At least, so it would appear according to the California ISO's up-to-the-minute Status Page.
I've seen many comments asking "I thought the U.S. Government encrypts the high-precision GPS signals so that normal civilians can't get the best precision?"
While this WAS true for a long time, the U.S. government turned off Selective Availability in May of 2000, making it possible for civilian users to get what used to be strictly military-grade positioning.
Of course, they might turn it back on in case of a "strategic conflict", as they'd say.
Re:A little quibble with the subtitle
on
Longitude
·
· Score: 1
I believe that what was solved was the greatest engineering problem of the time, not scientific. Everyone knew that you needed an
accurate timepiece, it was just that no one knew how to build one that worked on a ship.
I completely disagree. The requirement was that you develop a method of reliably determining longitude at sea. One of the various ways of accomplishing this is by determining your local time and comparing it with the time at a known longitude. And, one of the various ways of determining the time at a known longitude is by building an accurate timepiece that works on a ship.
Most people were convinced that a mechanical clock would never be accurate enough to keep Greenwich time. So the astronomers spent decades characterizing the moon's orbit, such that people would be able to determine the time by observing its current position and comparing that to a chart of time vs. predicted position. This turned out to be tough because the moon has a bizarre orbit (with something like a 17-year period), but the lunar method that was being developed alongside Harrison's clock ended up coming to fruition at the same time as Harrison's H4.
Now, put yourself in their position. Which method would you rather trust?
Harrison's clock, which while marvelous was relatively untested, and was hard to come by -- good luck trying to get Harrison to build one for you! And good luck finding someone who is nearly as skilled as he who might build it instead! Plus it might go overboard, break, be stolen, etc. And would be ungodly expensive.
Accurate lunar orbit predictions that were gathered over the course of something like 40 years and, though they admittedly required complex computations and many observations that could be hard from the hull of a rocking ship, had no cumulative error. And were reproducible as easily as copying a chart full of numbers (no craftsmanship is required here!) I'm sure everyone reading slashdot appreciates how much easier it is to reproduce and disseminate information, rather than tangible objects.
Now, if it were me back in the 18th centry, I think I would have been amazed at the H4 clock, sorry that I didn't have one, and still put my money on the lunar method. I think the real unsung heroes in this story are folks who came after Harrison---like Arnold---who took Harrison's design and improved upon it in such a way that an accurate clock could be mass-produced at low cost. Later design of reliable clocks available to the masses is what made it viable.
Re:John Harrison: 1st Open Source Inventor?
on
Longitude
·
· Score: 3
The last thing that Harrison wanted to do was give away the farm, but the Longitude Board forced him to for a simple reason: the invention would have been useless without knowing how it worked.
Consider the problem that the Board was facing: they had a fleet of hundreds of ships that were in desperate need of reliable navigation. What good does one clock do you? As brilliant as a mechanical engineer as Harrison was, the burden was on him not to create just an instance of an accurate clock, but a method of reliably building accurate clocks in the quantity needed.
When I read Longitude I really thought the author gave the Longitude Board the short end of the stick. She painted a picture of an evil board that was trying to delay awarding the prize for no reason other than spite. While there was undoubtedly spite involved on the part of some people such as Maskelyne, I think the Board was exercising due diligence by asking Harrison to test the clock on numerous occasions, build replicas, teach other people how to build replicas, etc. A unique and unreproducible clock would have been almost as useless as no clock at all!
There's a live demo of in-flight access going on right now (or least, it was going on an hour ago) by Globalstar here. It's a sort of lame but mildly amusing in-flight webcam of 2 engineers from G* and a local news crew flying around Los Angeles, talking to people on the phone over the Internet, using AOL IM, getting email, etc.
I'm one of the people who found the same line in inetd.conf (above as an AC). I just checked my log files and, sure enough, found the exploit to rpc.statd. Thanks, AC. (The machine is about to be wiped and reinstalled anyway)
Shame on me for not running 'rhup' the moment I installed the machine.
The fine folks down at CAIDA do a nice job of collecting all sorts of statistics about the Internet, partly to answer questions like this one. It's a good place to look for more info.
For example, in their paper Measurements of Internet topology in the Asia-Pacific Region, they focus part of their study on which countries provide IP transit for other countries. In other words, they want to know how often certain countries carry traffic that is neither sourced nor destined for that country. They conclude, in part (see Sections 4 and 5):
U.S. networks do seem to dominate global Internet topology -- they provide transit for 71.4% of the total skitter paths that neither originate nor end in the U.S. U.S. networks appear to be especially significant for other countries in the Americas: all traffic to Mexico and 97.8% of traffic to Peru and Chile (SWA) crosses the U.S. on its way. Our sample also shows a large transit role played by U.S. networks for traffic to China-Hong Kong (90.3%), Taiwan (83.5%) and Oceania (77.8% of traffic to Australia and 79.6 of traffic to New Zealand).
[...]
The U.S. is the major Internet transit intermediary for the rest of the world: 71% of traces that neither start nor end in the U.S. still pass through it. In most connections between different countries, the U.S. is the only third party country that also appears in the path.
BTW, never pass up an opportunity to hear kc claffy speak, she's great.
So it appars he's not going *INTO* space [...] VERY smart, considering the trip back from beyond the atmosphere is *tricky*
I'm not sure I agree that it's actually so terribly smart of a thing to do. I mean, the guy is planning on going 60 miles into the atmosphere. NASA supposedly defines space as starting at 62 miles up. What happens if his already rough-sounding calculations are off by a couple of percent, and he ends up in space for real? Spend the next 30 seconds desperately trying to formulate an atmospheric re-entry plan? (Make sure Gary Sinese is on call...)
Personally, I don't use any mode of travel that comes close to escape velocity. But that's just me.
It's possible to not make money, you know
on
Iridium Saved?
·
· Score: 5
Also, they apparently will get Iridium for about two cents on the dollar. How could not you make money with a deal like that?
As impossible as it sounds to people who write.com business plans, it actually is possible to not make money if you're running a business wherein you're spending more money than people are giving you for your product or service.
Since they're spending close to $1m a month to keep it in the air, advanced quantum calculations reveal that it is possible for them to not make money if they don't generate at least that much revenue. For example, because most people spend most of their time in relatively urban areas, where the cellular infrastructure has been built up to ubiquitous proportions, and satellite phones don't work so well due to line-of-sight problems, multipath interference, etc.
Slashdot Mirror
Training a computer to act like Van Gogh reminds me of the funniest piece of computer humor I have ever read.
i.e.: They started with a painting by Van Gogh, then ran it through the "Smart Blur" filter of photoshop to remove the Van Gogh-esque-ness of the painting, leaving a textureless image. Then they ran the texture learning algorithm over that pair of images, and applied the learned texture to the target.
You're falling prey to exactly the same fallacy as everyone else in the mass hysteria of the 1999 IPO market did: that somehow eyeballs == good investment. It is exactly that kind of thinking that caused the bubble to hyperinflate, and then burst: "Everyone is looking at this company's web page, let's invest!" Once they realized that there wasn't actually any money there, the bubble burst.
Even, say, 100 billion hits a day isn't worth any investment at all, if there's no way to make money from them. Would you invest in Slashdot, for example? So what if lots of people look at it - it doesn't make any money. It loses money hand over fist.
I won't know if I want to invest in Google until I see an S/1 and know if they translate their technology into profits or losses. If it costs more to run their engine than they make, and that will be the case for the forseeable future, how is the investment attractive?
Why is this an argument against automatic patching? The same thing would happen to everyone who manually goes to their vendor's patch server and applies what's there. What criteria could a person applying a patch manually use to determine that the latest patch RPM is malicious, that an automatic patch script could not use?
What is the syllable count?
Are links poetry?
Do Haiku Masters
Consider metadata
Like roots of the tree?
Attached, integral?
And thus counted, considered
Inseparable?
Or, like leaves and fruit
Are they transient, and thus
uncounted? Cast away?
I love your work, but:
Has "coverage" three syllables?
I think not as you.
Hello, the first paragraph of the RFC says
You can see streaming video of this talk here.
Perhaps the best example of this is the story of Landmark Routing. Go read the paper. You probably need to read it 3 times before it completely sinks in, but it's a brilliant way of building a routing hierarchy that scales beautifully in virtually every dimension: number of end-stations, number of sub-networks, number of routers, number of everything.
The basic idea is that you route packets the same way you might ask for directions on a trip from New York to your friend's house in Los Angeles. You start out asking "which way to LA?" and someone might point you vaguely westward. Once in LA you can ask someone from LA, "where's Santa Monica?" and once in Santa Monica, ask "where's Ocean Blvd?" As you get closer and closer to your destination, you ask for progressively more and more specific directions -- and are more likely to get them because you're more likely to encounter people who know the local geography. Your route may not be optimal (i.e., because the person in New York who you first asked for directions didn't have global knowledge of LA's streets), but it works very well in a distributed way, and it scales. This is Landmark in a nutshell (but, really - don't take my word for it; go read the paper).
Anyway, as I heard it, back in the days when BGP was being standardized, Landmark was at one point suggested as an alternative. Now, my very 3rd-hand understanding is that it was ultimately rejected because it just didn't fit the model of what Internet routing algoritms need to do. It can't express policy. It can't support ISP X saying "I only want to send data to ISP Y, because I have a peering agreement with them. I don't want to advertise routes to ISP Z because they use me as transit without providing transit. I don't want to route these packets through Canadian ISPs because it violates local laws." The economic reality is that you don't always route through the shortest path, or the cheapest path, but the path through which you've negotiated peering agreements or are buying transit service. BGP lets you express these policies.
Landmark is a beautiful technical solution, and from the point of view of pure computer science, it works. Maybe your solution does, too. But that doesn't necessarily have any bearing on its usefulness in the real world. For your idea (or any idea) to be considered seriously, you have to address these issues.
I hope it does mean the end of pay phones - and their replacement by phone vending machines :-)
Their net income for the year was only $70.8 million for the year ending 12/31/2000. That means, using my plan, they could actually increase their net income by 42%, to $100 million!
Sheesh.
Oh, come on, $2 billion is a lot of money! You're saying you can't make a measley $232 million when you have $2 billion in the bank? Your goal is just to break even?
Here's how you can stay profitable, Yahoo.
A) Discontinue all Internet operations. Turn off all routers, servers, etc. Fire everyone except your CFO. Get rid of all office space. Sell all furniture. Expenses should now be $0.
B) Send your CFO to Bank of America. For your convenience, I've even included a map that shows how to get from Yahoo corporate headquarters to the nearest branch. (Important: This is a link to Yahoo! Maps, so print this map before you do Step A.)
C) Tell the CFO to open a Prima Savings Account. $10,000 minimum balance, but that should be no problem. Deposit your $2 billion. (Note, you will also get free checking and free ATM card.)
D) Wait one year. You will be earning over 5% APR with your Prima Savings Account, just like I do. Pray that Greenspan starts raising interest rates again.
E) Celebrate! You've now made $100 million! Maybe not the $232 million that we wanted, but it's a lot better than breaking even, right?
Just a suggestion.
Actually, there are.
FSN is not fake, it actually looks just like what you saw in the movie. I think the Jurassic Park people added the sound effects, but the real FSN actually let you fly around a graphical representation of your filesystem, fly into subdirs by clicking on them, launch apps, etc.
AltaVista did, at one point, fall from the purer faith. But then the clean, portal-less look of Google showed them the way, and they repented with Raging Search. Raging is exactly the same database and search engine as AltaVista, but with a Google-esque minimalist look. I can't imagine why anyone still uses AltaVista's front door when you actually your work done going through their back service entrance.
It seems that, 30 minutes ago (a little after midnight pacific time, early Friday morning), the Stage 3 emergency has been reinstated. At least, so it would appear according to the California ISO's up-to-the-minute Status Page.
While this WAS true for a long time, the U.S. government turned off Selective Availability in May of 2000, making it possible for civilian users to get what used to be strictly military-grade positioning.
Of course, they might turn it back on in case of a "strategic conflict", as they'd say.
Most people were convinced that a mechanical clock would never be accurate enough to keep Greenwich time. So the astronomers spent decades characterizing the moon's orbit, such that people would be able to determine the time by observing its current position and comparing that to a chart of time vs. predicted position. This turned out to be tough because the moon has a bizarre orbit (with something like a 17-year period), but the lunar method that was being developed alongside Harrison's clock ended up coming to fruition at the same time as Harrison's H4.
Now, put yourself in their position. Which method would you rather trust?
Now, if it were me back in the 18th centry, I think I would have been amazed at the H4 clock, sorry that I didn't have one, and still put my money on the lunar method. I think the real unsung heroes in this story are folks who came after Harrison---like Arnold---who took Harrison's design and improved upon it in such a way that an accurate clock could be mass-produced at low cost. Later design of reliable clocks available to the masses is what made it viable.
Consider the problem that the Board was facing: they had a fleet of hundreds of ships that were in desperate need of reliable navigation. What good does one clock do you? As brilliant as a mechanical engineer as Harrison was, the burden was on him not to create just an instance of an accurate clock, but a method of reliably building accurate clocks in the quantity needed.
When I read Longitude I really thought the author gave the Longitude Board the short end of the stick. She painted a picture of an evil board that was trying to delay awarding the prize for no reason other than spite. While there was undoubtedly spite involved on the part of some people such as Maskelyne, I think the Board was exercising due diligence by asking Harrison to test the clock on numerous occasions, build replicas, teach other people how to build replicas, etc. A unique and unreproducible clock would have been almost as useless as no clock at all!
There's a live demo of in-flight access going on right now (or least, it was going on an hour ago) by Globalstar here. It's a sort of lame but mildly amusing in-flight webcam of 2 engineers from G* and a local news crew flying around Los Angeles, talking to people on the phone over the Internet, using AOL IM, getting email, etc.
Shame on me for not running 'rhup' the moment I installed the machine.
For example, in their paper Measurements of Internet topology in the Asia-Pacific Region, they focus part of their study on which countries provide IP transit for other countries. In other words, they want to know how often certain countries carry traffic that is neither sourced nor destined for that country. They conclude, in part (see Sections 4 and 5):
BTW, never pass up an opportunity to hear kc claffy speak, she's great.I'm not sure I agree that it's actually so terribly smart of a thing to do. I mean, the guy is planning on going 60 miles into the atmosphere. NASA supposedly defines space as starting at 62 miles up. What happens if his already rough-sounding calculations are off by a couple of percent, and he ends up in space for real? Spend the next 30 seconds desperately trying to formulate an atmospheric re-entry plan? (Make sure Gary Sinese is on call...)
Personally, I don't use any mode of travel that comes close to escape velocity. But that's just me.
As impossible as it sounds to people who write .com business plans, it actually is possible to not make money if you're running a business wherein you're spending more money than people are giving you for your product or service.
Since they're spending close to $1m a month to keep it in the air, advanced quantum calculations reveal that it is possible for them to not make money if they don't generate at least that much revenue. For example, because most people spend most of their time in relatively urban areas, where the cellular infrastructure has been built up to ubiquitous proportions, and satellite phones don't work so well due to line-of-sight problems, multipath interference, etc.