Linux screensaver for window

Think you'd like to break in as an intern at Microsoft's research lab? Fine, but you'd better have a Ph.D.

Like many other Silicon Valley companies, Microsoft Corp. has begun to open its doors to showcase some of the next-generation technologies the company is working on. This past week, the software giant invited press, analysts, and even the general public to get a glimpse of what might be in Microsoft's future products.

The event was organized in a two-part session: first, Rick Rashid, senior vice-president of Microsoft Research, talked about the organization's mission within Microsoft. More details of that talk can be viewed here in a report by PC Magazine. My key takeaway? Microsoft has over 150 Ph.D. interns -- interns, mind you – in Redmond alone.

After Rashid's presentation, Microsoft invited its guests into a nearby room where attendees had a chance to meet with the researchers and talk to them about their pet projects. Again, some of this has been covered in the PC Magazine news story, but I also had a chance to ask a few questions myself of some of the scientists.

I took a look at five different technologies. You can jump to each one using the table of contents below, or just click the next key to read about each of the sequentially.

Lets jump right into MyLifeBits.

One of the most interesting initiatives Microsoft is working on is a plan (I hesitate to call it a project, due to its massive scope) is MyLifeBits, which is in effect a superblog, or your own personal Truman Show. It is, in short, a plan to store your entire life online.

Granted, this is something which works much better in concept than reality. Gordon Bell, who became famous for designing a host of Digital Equipment computing projects, has made it his mission to begin documenting his life as much as possible, and storing it online. Today, this isn't much more than a few digital photos and digital clips organized into a timeline. But Bell anticipates the not-too-far-off day when today's 250-Gbyte hard drives will morph into a full terabyte, more than enough storage to start archiving the banalities of everyday existence.

"Every conversation you'll ever have could be in that terabyte of storage," Rashid said. "Depending on what part of the country you're from, you could have a lot of disk space left over."

Certain reports have already begun to raise the fears of privacy (shame on you, David Coursey!), but this isn't exactly Big Brother implanting a camera in your brain. Microsoft bears watching, no doubt about it—but for right now this is just a glorified image archive.

What is interesting, however, is the file system Microsoft will need to implement to make this work. An image file today is just that—an image. Windows doesn't know what it "is", unless you label it as a picture of a picnic or of your brother Stephen's graduation. Future versions of Windows may contain intelligence to "tag" each data file with important information for archiving. Of course, that information is the valuable bits, and those tags are the ones you'll most likely want to keep secret, unless you're the exhibitionist type.

Next up: Windows Gestures

I always rib one of my best friends about his Clapper, which turns lights off and on when he, well, claps. To me, it's just lazy. On the other hand, it does save him the small motion necessary to turn his lights on and off. Over time, those motions add up.

Have you ever used one of those "nipple mice" on IBM notebooks? As part of a notebook keyboard layout, the concept makes sense. But using it over and over again is just asking for an attack of repetitive stress disorder.

Windows Gestures uses a pair of binocular cameras (webcams work just fine) mounted on your monitor. The software looks for motion, generally homing in on your extremities. Move a hand, and your mouse pointer moves. Leave it still, and the cursor stops, and then flicks back and forth between various options.

I don't know what sort of a PC the technology demo was running on, but it consumed about 48 percent of the CPU cycles when the system was actively interpreting gestures, and about 16 percent in the "standby" state.

It appears Microsoft hasn't quite nailed down the concept of adapting a two-button mouse to this system, though. The software isn't smart enough to pick up finger movements, either—no virtual Nintendo Power Glove here. One smart idea the designers have come up is using the two cameras to define a 20-cm depth of field, creating a "hot zone" in which gestures move the cursor.

Still, it's hard to believe that the technology is practical without speech recognition, for the simple reason that keyboards are often placed close to the computer screen. Microsoft researchers, meanwhile, are struggling to visualize what other forms of user input might be used instead.

"Ten years into the future, will we still use mice and keyboards?" asked Andy Wilson, the designer of the technology. "I'm not sure we will. But I'm not sure we'll use speech recognition, either."

Next, never lose a document, with StuffI'veSeen.

Tied to the MyLifeBits concept is a new tweak on a familiar menu item: "My Recent Documents". Working on a single project every day may be relatively easy to manage, but I constantly multitask, wandering through hundreds of web pages, discussion groups, Word files, PowerPoint presentations, emails, and other documents in search of facts and news. Thank goodness it's all on my PC—I once had a fire marshal threaten to prevent me from working unless I cleaned up my office.

The problem, of course, is that I know I've seen the information somewhere, but when it's time to write a story, that digital scrap of paper is nowhere to be found. Enter "StuffI'veSeen".

Like the Windows Gestures technology, StuffI'veSeen depends either on eye tracking via some sort of a camera, or a metadatabase (sort of a combination of the "history" window of Internet Explorer, and a "My Recent Documents" folder). That's it, really. The sophistication comes from the context in which you viewed it: maybe you know it was the first file you opened after your 10 o'clock meeting, or that you know a colleague sent it to you via a peer-to-peer program. Perhaps it was the file you physically looked at most frequently last Tuesday.

If there's a hidden subtext here, it can be summed up in one word: databases. What I took away from this event is that Microsoft's corporate strategy is to find new ways to make it easy to store data digitally, then provide the front end and management tools to access, control, and manage that data. Remember, this is the company that forgot about the Internet until two years after it went mainstream, and now controls the browser market.

Next, new ways to fight spam.

We've written quite a lot about spam, on new proposed technology to defeat it, and even legislation to try and halt the spread of spammers and their ubiquitous emails.

Microsoft researchers Mike Burrows and Cynthia Dwork have developed a method of "computational spam fighting". The idea is a combination of the "it's me" system MailBlocks uses, and the background number-crunching used by SETI@home screensaver.

The idea is that users should be required to "pay" to send email, but not money. Instead, they should be willing to prove their existence by solving some sort of puzzle, automatically, via their computer. Burrows and Dwork posit that the best way to do this is by proving there's a PC on the other end of the line.

Authenticating the source of the email would require a puzzle to be solved. The complexity of the puzzle would require 23.8 seconds of computation by a 366-MHz Pentium II-based laptop, 11 seconds for a 1.2 GHz Pentium III laptop, and Pentium II-based laptop, and 14.3 seconds for a 2.0-GHz Pentium 4 desktop. (Yes, those figures were accurately recorded—apparently the Pentium III fares better at those tasks.) Note that the crypto algorithms are CPU bound; the amount of memory in a given system shouldn't affect the computing times all that much, Dwork said. Of course, this would seem to put a hitch in the plans of Linux users who like to run distros on outdated hardware.

Once authenticated, the user wouldn't have to recrunch the crypto algorithms. The idea is to make sending email moderately painful for new users, who have to send out one or two at a time. Spammers, however, would see their PC come to a grinding halt.

Next on my list, Microsoft's new methods for improving search engines, where I answer the question, "what's a ShinglePrint?"