Software

The Next Killer App

Any successful information technology requires a killer app — an application so compelling that it can, all by itself, justify the purchase of a given device. When it comes to personal computers, there have been many killer apps, starting with the spreadsheet, but there haven’t been any new ones in a long time, which is a problem. This column is about a potential new killer app for the PC platform, perhaps its last one. What do you know about telepresence?

I have for the last several months been shooting for Maryland Public Television a new PBS documentary about how information technology has transformed our lives and businesses. It is an esoteric and very close look at a few technologies. Some, like the rise of office automation and the personal computer, are obvious: secretaries and telephone operators have disappeared while all the rest of us learned to type. Others, like RFID (radio-frequency identification) chips, are harder to see but just as transforming by creating a real-time distribution system in which we can know where everything is moving all the time, taking just-in-time inventory from a goal to a reality. One of the most striking of these technologies, which has yet to achieve wide use, is telepresence — high-definition video conferencing as a substitute for business travel. Telepresence is not far from being here on a wide scale and the effects — even beyond business — should be profound. I see telepresence shortly invading our homes.

The way vendors tend to implement telepresence today is fairly uniform if inconsistently interoperable. A special conference room is built that is actually half a conference room, half of a table set against a wall that is all video screens. Typically three big projection or LCD displays are side by side in landscape mode with a third landscape screen mounted higher on the wall above the middle lower screen. The three lower screens are used to show the remote participants in the meeting. Sometimes all three screens are devoted to participants at a single location, but up to four locations can be linked if needed. The top screen is used for meeting materials like PowerPoint presentations or videos intended to be seen by all the participants.

A couple weeks ago I used a system of this type (in this case it was from Hewlett-Packard but there are similar systems from Cisco and other vendors) to interview the people who designed it. I was in Palo Alto and they were in Corvallis, Oregon. In addition to the telepresence system (called Halo, whatever that means) I had a camera crew at both ends to record the action in each room.

These rooms are not cheap to build or run. The HP systems cost either $249,000 or $349,000 to build, depending on the model, and $18,000 per month to operate. This gets you a DS3 connection (45 megabits per second) to a low-latency global network and 24/7 support. Each of the lower screens uses six megabits per second, with the remaining 27 megabits per second for that fourth upper screen. If this seems like bandwidth overkill for PowerPoint, understand that this HP system was co-developed with the Dreamworks movie studio specifically to allow dispersed groups of executives to review 1080p HD footage from upcoming films and for dispersed editors to actually work together to edit films without having to be in the same city. With feature film production budgets now averaging $50 million, $18,000 per month for editing support is nothing. With just over 100 such rooms now in operation for HP, part of the high price is also simply investment recovery. If HP were selling 10,000 of these rooms per month the price would be substantially lower.

Video conferencing has been around for a couple decades, but telepresence is different from that. You can see the entire other side of the conference table, for example, and the people who are sitting across from you appear to be life sized. They can see you and you can see them. When another person speaks to you they can look you in the eye. Body language and emotions are easy to detect and the sound of each participant seems to come from his or her direction. You can watch the people who aren’t talking to see if they are even paying attention. It really is tele-PRESENCE and the fact that you are looking in a video screen is forgotten after a minute or two.

Here are some lessons I learned from the experience. For one thing, size really does matter. The big screens changed for me the entire experience, though I think a home telepresence system could do fine with a single big HD screen instead of three. Eye contact is important, too, and that is generally accomplished through two techniques — mounting the camera as far as possible from the subject then using some subtle video morphing software to make it seem as though the camera was actually mounted behind the screen. For a home system I believe most of this effect could be achieved by simply increasing the distance from subject to camera, reducing the angle at which the camera is seen above the screen. This is the major failing in video chat systems where the camera is mounted on the display.

Using the system I quickly came to understand that the real power wasn’t in bringing together groups of big shots for huge powwows at which sweeping global decisions would be made. The quintessential telepresence meeting lasts 10 minutes and involves a group of people at any level who simply need to come to a concensus. Nobody flies 10,000 miles for a 10 minute meeting yet everybody walks down the hall for one. Being able to hold such a meeting is what can make a widely distributed group of workers function like people in the same building, which has been one of the nagging problems of global development and outsourcing. The fact that it already has eight rooms up and running in India may give HP some advantage in that respect.

We’re early in this process, but I think telepresece is going to be a big deal. It comes down to a big, high-resolution screen, good sound, mounting the camera far enough away to simulate eye contact, and of course throwing lots of bandwidth at the problem. Some of these components, like relatively cheap high-def big screens, are here today. The processing power required is here, too. The only significant obstacle to us having our own telepresence systems is bandwidth, and we can predict with some accuracy when we’ll have that.

This bandwidth calculation involves applying some variation of Moore’s Law in two dimensions. Available bandwidth at a reasonable price is always increasing over time. The second dimension involves changes in compression technology and increases in processor power that over time reduce the amount of bandwidth that will be required to carry a high-res video signal. With the passage of time, then, available bandwidth increases while, at the same time, bandwidth requirements decrease. This has the effect of amplifying Moore’s Law, accelerating that point at which telepresence at a reasonable cost is possible.

So when will it happen? When will we have telepresence capability in our homes? Some of us are there already and don’t even know it, the only remaining problem being one of integration.

The home embodiment of that HP Halo system would be a single big screen, which using even the current HP technology would require six megabits-per-second. Millions of Internet users in Asia and Europe already have that kind of upstream bandwidth and hundreds of thousands of U.S. residential customers (mainly Verizon FiOS users) do too.

But HP’s Halo system is old-tech, using MPEG-2 compression that is more than a decade old. A home system built around a more powerful codec like H.264 and using more powerful hardware could reduce the required bandwidth for home telepresence by at least half, making the likely barrier three megabits-per-second. That kind of bandwidth is nothing to users in Korea or Japan and it is nothing, too, for fiber-to-the-home users in the U.S. (mainly Verizon) and wouldn’t be that much of a stretch, either, for fiber-to-the-curb vendors like AT&T. Giving two megabits upstream to every cable modem user wouldn’t be trivial, but it is possible and could be — I think WILL BE — spurred by competitive pressures from DSL.

So the bandwidth is coming and millions of people will have it, even in America, by 2008. What’s missing is both consumer demand and painless satisfaction of that demand through easy-to-use high-volume products, which come down to big screens, cameras, and PC systems running the right software. The part of this that is both hardest and easiest is stimulating demand. People aren’t demanding telepresence because they have never experienced telepresence. If you show them they will come.

This is 100 percent analogous to the introduction of color TV in the 1950s. People didn’t know they wanted color TV until they saw color TV. But once they saw it, the lure of color TV was instant and obvious. What was difficult with color TV was that it required a large and very expensive video production and distribution infrastructure that cost tens of billions of dollars and required major financial commitments from vendors like RCA, which had to build transmitters, receivers, cameras, an entire TV network (NBC) and even subsidize the production of color programs like Bonanza to seed the system. Home telepresence requires almost none of that and, in fact, actually leverages the huge investment already made in HDTV, since that’s what those big telepresence screens will no doubt be used for when nobody wants to visit with Grandma or play video strip poker.

All that’s required to sell consumers on home or small business telepresence, then, is allowing them to experience it. And, of course, making it affordable.

I think Apple will be the first PC-only vendor to embrace the telepresence business. Steve Jobs would like another killer app. His last attempt at creating one — video editing — was only somewhat successful. The iPod of course qualifies for killer app status, but I don’t think it has actually sold many computers, though lots of iPods.

Apple had big screen TVs ready to introduce a year and a half ago but cancelled them at the last moment as too mundane. You could buy an HDTV from HP or Dell and Apple apparently didn’t have that much more to offer.

But this time it will be different. Imagine one of the new aluminum and glass iMacs only instead of a 24-inch screen make it 42 inches. The familiar iSight camera will be there in the bezel. but this time the camera will have HD resolution. This hang-it-on-the-wall iMac would establish yet another category of computers, which is what Apple loves to do. They’ll sell a million units to the faithful and all it will take is putting an active telepresence system in every Apple store connected to every other Apple store for prospective users to play with. This gets Apple into the big screen TV business with a system that has higher margins simply because it isn’t just a TV but is also a Mac. Look for all this after Christmas along with refreshed Macs featuring the H.264 encoder chip I pre-announced a number of months ago. Look for Apple to also facilitate telepresence by turning it into a service as it has more and more wanted to do. Then imagine that system connected to a 3G iPhone.

For Apple the point is to create a platform to allow more natural implementation of “lean back” content. Apple TV was the first push in this direction, but this telepresence system will be both easier to use and more expensive, two attributes near to Steve Jobs’s heart.

Swimming with the Fishes

The way change happens varies over time as technologies and markets mature. This column is about two examples of new directions of change for the Internet. First there is gillznfinz.com, a fishing web site that teaches us (and the Pentagon too) new lessons about niche markets and the future of television. Then there is Adobe's upcoming Flash 9 and its competition with Java.

There are two key differences between watching television over the Internet and watching Nightline in your bedroom: 1) Nightline usually looks a lot better than Internet TV because of higher production values and greater available bandwidth, and; 2) an Internet TV show can serve a truly global audience while Nightline is limited to the audience of ABC affiliates in the U.S. These two factors combine to define the opportunity for Internet TV, which generally comes down to content for thinly dispersed but rabidly enthusiastic audiences who won't care if the picture is a little shaky or the jokes are bad.

The example of this effect that I always like to use is the Nazi Memorabilia Channel, which would have a hard time finding carriage on any U.S. cable system because the number of potential viewers is so small. But if the viewers could be drawn from a global Internet population that includes such ex-Nazi hotspots as Paraguay and Brazil, maybe the channel would have a better chance.

Which bring us to our latest poster child for Internet TV -- Gillznfinz.com.

Gillznfinz.com is the brainchild of ESPN2 alumnus Capt. Adam Paul (he's captain of a fishing boat, hence the title -- no saluting required but you buy the first round) and is intended to take all those cable TV bass fishing shows to a whole new level. Paul, 28, who seems to have devoted his life to offshore fishing, wanted to find a way to let the 30 million fishermen and women in the U.S. and their untold millions of counterparts overseas continue to experience their sport whenever they were in front of a computer screen. Gillznfinz is a saltwater-logged combination of MySpace and YouTube where for $15.95 per year normal people without webbed feet (yes, Capt. Paul's feet ARE webbed, I am not making this up) can follow along on the video and text adventures of a growing number of professional offshore fishermen. The site covers travel, fishing, local entertainment, and where to get bail money if needed.

"It's half the price of most fishing magazines yet the content is changing 24/7," said Capt. Paul at the site's July launch. From a business standpoint, the combination of user-generated content (blogs, pictures, video, e-mail) plus professional content from contributors, who are mainly other captains trolling for new fishing customers, makes a lot of sense.

But what sets gillznfinz.com apart from other fishing sites is its use of technology. The site relies heavily on video and will shortly launch its first live fishing shows linked by satellite from almost anywhere in the world. "We'll cover tournament fishing with a live satellite uplink from one boat and wireless cameras on other boats in the fleet sailing up to a mile from the uplink."

Not even Nightline can do that.

Actually, NOBODY can do that, which is what makes Gillznfinz interesting to, of all outfits, the U.S. Department of Defense.

What Gillznfinz has done with its fishing tournament coverage is to create a unique picture of almost everything that is happening in a fishing fleet covering about three square miles, making nearly all events and participants individually addressable through a combination of video, audio, and text messaging. Everyone in the tournament is addressable from outside the tournament, as are up to dozens of video cameras. As a viewer pretending to work at your PC on the 14th floor of some office building, you can watch (and communicate with) your favorite fisherman or figure out from the chat traffic where the real action is, then beam over there to see some other fisherman land that big fish.

If this sounds to you like watching paint dry, then Gillznfinz might not be for you, except, of course, for the beer and vodka ads. But to some fishermen who work for the DoD, that tournament looks a lot like a battle, those captains and fishermen like officers and enlisted men, and the chance to jump interactively into such action is a first for the Pentagon.

That is not to say Iraq and Afghanistan aren't already appearing on a screen near you. The ability of the U.S. theater command to be based in Florida, rather than closer to the battlefield, has been attributable to video for some time. Real-time battlefield video presently makes its way to Florida by satellite using proprietary video compression technology from a Scottish company called Essential Viewing. Having earlier tried to slug it out with better-backed codecs like QuickTime and Windows Media in the commercial video market, Essential Viewing stumbled upon the defense market and now thrives in that world of $900 hammers. The real-time capability of its codec, originally developed at the University of Strathclyde, is paramount in an application where command and control are dependent on knowing what's happening on the battlefield RIGHT NOW.

Capt. Paul, fighting that big striper back in some million dollar bass tournament, doesn't have Essential Viewing technology with its highly efficient low-bandwidth connections, but he has something better -- the bidirectional integration of text and audio from a virtually unlimited number of sources like e-mail and chat along with a lot more cameras. Having more data sources means more information from the tournament (or battle) and viewers who can get much quicker to where the action is.

A generation ago this sort of application would originate on the battlefield then slowly make its way into consumer applications. But today the consumer electronics market is so large and the rate of product development so fast that it isn't surprising at all to see consumer technology leading the military.

Another example of commercial competition leading to really significant advances in technology can be seen in the next versions of both Java and Flash. As I have written before, Java and even Microsoft's .NET suffer in comparison to Flash, which is more widely deployed than either development environment and features smaller programs with higher performance. Java and .NET, too, have suffered from the popularity of AJAX applications, which are also lighter and faster.

Sun is fighting back with a new feature called Java Kernel in Java 7. Java Kernel is a version of the Java virtual machine that is greatly reduced in size, loading only those classes that are required for an application and downloading even those only as needed, the goal being to make Java applications and, especially, applets lighter and more nimble. And who can argue with an 80 percent reduction in code size? Java Kernel is a programming triumph and about time, too.

Of course it won't make much difference in the long run, but that doesn't make Java Kernel any less impressive.

Though JVM downloads drop dramatically in size, it isn't at all clear that this will lead to significantly faster applications or even applications that at least start faster. It's waiting all that time for applets to start running that has hurt Java's adoption rate and to a certain extent .NET's, too. Flash apps are still likely to load and start quicker.

But wait, there's more! According to Adobe, the next version of Flash -- Flash 9 -- will ship with dramatically expanded codec options allowing significantly better and faster video.

There was a time not long ago when expanded Flash codec choices wouldn't have mattered. Streaming video had already failed and the market was moving to downloaded video where data rates and total file size were less immediately relevant. Then came YouTube and its commitment to Flash video. Now video streaming is again a hot idea and Flash 8's choices of the H.263 or VP6 codecs just aren't enough. Most implementations of H.263 are limited to 320-by-240, and VP6, though very efficient, just isn't a mainstream technology. In that respect it is like Essential Viewing's codec.

But the next version of Flash video will support H.264, AAC audio, most HD frame sizes, and -- here's the most important part of all -- will work with your graphics card to make it all run faster and with less CPU load.

This is a huge kick in the head to both QuickTime and Windows Media, though of course QuickTime has an important role in video production in most editing systems and in parts of the H.264 codec, itself. Windows Media and its VC-1 codec also have an enduring role in the production of professional content. But when it comes to video client software that is high performance, cross-platform, and available already in 97 percent of all computers, well Adobe wins this round easily.

By Robert X. Cringely

Surviving Immortality

I've been thinking about the Technological Singularity, which to proper geeks is that point where computers become smarter than humans and supposedly all bets are off as technological development races forward faster than we can catch it and you and I are either left eating bonbons or are put to death by computers no longer amused by serving us. Life post-Singularity will, of course, be somewhere in between those two eventualities. Zits may be abolished but youth will still be anguished. Computers may be designing warp drives but I'll still be paying my mortgage. Rather than a technological Hell or Utopia, the Singularity is likely to leave us still in our sitcom just with different props. What's fascinating about the Singularity is not so much guessing what life will be like then as looking at our very approach to the concept and some likely side effects we'll bump into along the way.

Some very smart people are getting really worked up about the Singularity. Artificial Intelligence pioneer Ray Kurzweil, who makes his living from explaining and describing the Singularity, thinks it is generally good, that the Singularity will transform our culture in mostly positive ways and allow us to become effectively immortal. Bill Joy has a darker view, seeing really smart machines as a threat that might enslave us and certainly expose us as a culture to unexpected risks. To me the Singularity feels a lot like Y2K and the anticipation people had as the millennium approached, testing the capability of our computers to keep streetlights running and not make airliners crash. In Y2K the people most upset about the concept, those folks who sold their homes and moved to the mountains, were also the ones most excited by it. As much as they were scared by the concept of Y2K, I saw that they really wanted it to happen in the same sense that members of an apocalyptic cult might find that very apocalypse to be reaffirming, even if it means we'll all be vaporized or shipped directly to Hell.

There are many unanswered questions about the Singularity. Will it take place? How will it take place? When will it take place? What will be the effects of it taking place? If the effects of the Singularity are negative is there anything we can do about them? Can we shape the Singularity to our advantage? But what I find most fascinating is wondering what will happen between now and the Singularity as we anticipate and prepare for what I am expecting to actually be an anticlimactic event.

While I don't want to dwell too much on Ray Kurzweil, I think he is also worried more about what's between here and there. Ray believes the Singularity will bring immortality, so it has become very important to him to take care of himself in anticipation of that eventuality. Where some of us might think that a coming cure for obesity, for example, would be a great excuse for eating more chocolate, Ray is taking a much more pragmatic approach. He is trying very hard to live long enough to live forever.

This reminds me of an evening several years ago when I sat at a technology awards dinner next to a guy who happened to have a Nobel Prize in medicine. Just a little drunk, he explained to my wife and me that his work in genetics would allow our children to live forever.

"What about us?" I asked.

"You're screwed," he replied.

Someone always has to be the last person to die of a disease that is being conquered. And if you take a very positive view of the Singularity that means someone has to be the last person to die, period. Ray is just determined that person isn't going to be him, and I wish him all the best in that quest. Now pass me the chocolates, please.

It occurs to me that I haven't yet made the point that I believe the Singularity WILL occur. Oh it's coming all right, despite the spectacular under-performance of artificial intelligence over the years. But like every other rite of passage, this one will be both more and less than we expect it to be. Our troubles won't go away, they'll just become different troubles.

For example, the Singularity is phenomenon with both technological and economic components. Moore's Law works the same way. The underlying concept of both is the level of technological development we can reach AT A CERTAIN PRICE. The most powerful supercomputers can cost tens of millions of dollars and it is logical to assume that something on the order of a supercomputer would be the first machine to reach Singularity status. That's fine and would undoubtedly result in the creation of knowledge that would have an impact on all of us simply through the existence of that knowledge and its subsequent use by people and machines who may not have yet grown to Singularity, but what will really change everything is when the price of Singularity drops low enough to apply to the computer on our desks or on our wrists.

When will $1000 buy a machine that is 10,000 times smarter than the average human? Ray pins that at 2029, 22 years from today. I'll be 76 years old if I make it. My youngest child will be 23 and just embarking on an adult life that can't help but be transformed by technology. My planning for the Singularity will be a lot easier than his. And whether we've broken the speed of light, created immortality or not, I imagine I'll still be hoping he joins a boy band and supports his old man.

Between now and the Singularity a lot of things will change and one of the most frustrating aspects of that change for many people is how hard it is to be specific about it. One of those monthly science magazines, for example -- the kind that likes to have cover pictures of flying cars -- asked me not long ago to describe the personal computer of 10 years from now. It's simple to do a few numbers and calculate CPU performance, memory capacity, storage and such, but harder to say that we'll even still be using PCs, though I'm guessing we will. And the magazine was especially frustrated that I could say we'll have umpteen terabytes of storage, for example, but couldn't say exactly what storage technology would allow that. Holographic? Quantum? Some technology from behind Door Number Three? Nobody can say for sure. Or at least I can't. But I feel confident of the number, if not the way that number will be achieved.

What we have working here is the Law of Accelerating Returns, the best example of which is the Human Genome Project, which set out on a 15-year quest to map the human genome and, nine years into the project had only mapped about one percent. To linear thinkers this appeared to be a failure. But completing one percent of the map wasn't the same as completing one percent of the TASK, which included developing the technology for efficient genome mapping. The project was actually completed ahead of schedule and under budget. We're going to see a lot of that kind of thing in the near future with massive effects long before the Singularity.

I spoke recently with folks from BitTorrent, for example. This peer-to-peer data distribution technology has dominant market share and is poised to do some very interesting things in coming months, but at its core P2P is really just a coping strategy, a way of making today's bandwidth enough for tomorrow's applications. But with Internet backbone bandwidth doubling every year for the foreseeable future, there will come a time when we'll have enough bandwidth to not need something like BitTorrent. For the backbones that time looks to be around 2013 and for our home connections to catch up will take until 2016. How is this going to happen? Beats me, but I know it will, barring nuclear war or alien invasion. So the market opportunity for P2P technologies like BitTorrent AS WE ARE CURRENTLY USING THEM will really last for only another six to nine years.

That may not seem very long but it is enough time to build and sell a good business, so this is not in any sense a negative statement about BitTorrent or P2P. It's just the way the world works. And we might well find that technology advances eliminate the need for many servers putting us right back to using P2P, though for somewhat different purposes and in somewhat different ways.

There are half a dozen technology families that have been in parallel development for the last 30 years. Processing power, memory, storage, displays, communication, and manufacturing technologies have been on parallel paths leading to a synchronistic convergence. Remember how in Back to the Future Doc Brown's time-traveling De Lorean went from requiring lightning strikes or stolen fissionable material as a power source to running on recycled garbage? THAT's an accelerated return and exactly the kind of effect we can look forward to.

The real peril in all this is that our social, cultural, and political technologies probably won't keep pace, meaning we'll have whole new ways to hurt ourselves and others along with the same old ways to keep ourselves from doing so. The Singularity may well bring with it the end of death, but I am 100 percent certain that taxes will survive.

Posting from within Google Docs

I just discovered this today, and I don't remember hearing about his before.  You can post documents directly to your blog from within Google Docs.  This is very similar to the functionality that Microsoft Word has.

You do this from the Publish tab for each document.  You can configure this for a hosted provider, or any blogging software that supports the standard xmlrpc APIs.

Publish settings:

Blog Site settings:

Restore from Apple Backup backups

I have not been able to restore files from the backups that I have made with Apple Backup.  When I attempt to restore, using Backup, the progress dialog appears to show the restore working correctly, but the resulting folder doesn't contain any files.

I found this comment on this post that describes a solution.  The backup files created by Apple Backup contain disk images with the files being backuped.  It is easy to get access to the files that you want to recover.

First use Backup to find the backup file that contains the file that you want to restore (if you are using incremental backups).  Then run the following command to mount the disk image:

hdiutil mount -readonly BackupFile.IncrementalBackup/Contents/Contents/Backup.dmg