Archive for October, 2007
Google Earth Blog – Frank Taylor writes:
John Hanke, Director of Google Earth and Maps at Google, and the CEO of Keyhole before it was bought by Google, spoke at Berkeley where he got his business degree (video here – Real Player required). This is a fascinating lecture as it tells the story of his several business ventures and the challenging history of Keyhole before it was acquired. Also, at the end he shows some interesting slides about the success of Google Earth. I especially liked the [point] map showing all the placemarks on Earth in their search index on a map. He also shared a photo of the fleet of cars Google is deploying to take Streetview imagery which was taken by a blogger. Well worth watching if you want to know more about the history of Google Earth.
I have a few alternate recollections on the very early days, but I wasn’t as closely involved in the funding situation as John, Brian, and Michael. I recall we had a few leads on bubble-sized funding right before the April 2000 internet meltdown, the VCs then pulled back, and it took us until Jan/Feb of 2001 to close that first, smaller $5.5M in series A from other sources, which is coincidentally about the time I left.
I’m not sure about the first version of the technology John saw, but I thought we already had the software working with compressed imagery over the internet by the time he signed up to be CEO of the spin-off (when I got my formal offer to switch from Intrinsic to Keyhole full-time). The very next thing I remember working on in spring of 2000 was the giant quadtree scene management for data layers, etc.. and the 3D user interface (navigation, compass, placemarks, and the custom content system that preceded KML).
But as I’ve said before, the credit and reward for making Google Earth what it is today belongs to the people who stuck it out through the lean times and didn’t give up. Had I stayed on, I would have been happy to be paid in stock during the cash crunches (hint: unless you’re a founder, always try to maximize stock before salary — startups don’t have enough of an upside if you don’t take that risk). But I honestly don’t know how I would have dealt with In-Q-Tel being the financial savior of the company at that critical time. I was, however, very happy about the decision to merge into Google vs. go it alone and everything that’s happened since.
Following the Virtual Worlds fall conference last week, there’s some interesting fallout. First, a big WTF? to Microsoft, whom I’d given ample credit for having their heads on straight with respect to virtual worlds. Afterwards, I’ll cover the Linden/IBM "synergy." Read the rest of this entry »
Now here’s a camera I’d love to buy. It’s certainly not the first in the field of computational photography. But it’s nice to see companies start supporting this at the application level too.
Now, there are lots of ways to get 3D images via digital photography. You can shoot lasers or IR signals out from your camera to measure distance per pixel based on return trip time. You can take stereo pairs and infer depth. But this method goes much farther — it takes many similar pictures from slightly different angles and can compute the whole field of images from any point nearby.
In other words, it’s as close to real-time digital holography as we current can get. And the result is a photograph that not only knows the distance to each pixel, but can also tilt the image in 3D to see around objects, as well as modify the focal plane across the image.
And when companies start using cameras like these and computational photography techniques to compute novel viewing positions (i.e., positions that might sit between or outside of actual sample), we’ll have a truly 3D photographic world to explore.
The announcement is straightforward. Google Geo (Maps + Earth) has given Multiverse Networks an easy path to exporting 3D content, models, terrain, imagery (I imagine) into the Multiverse framework. So there’s now apparently a button to extract a small section of the Earth.
What was notably absent from the article, and we’ll see if the official announcement has more details, is any mention of licensing the GE rendering/streaming engine. That’s significant, because without that engine (or something equivalent), you can’t handle more than a small area of the Earth. Small areas are all the article talks about, so I’m not expecting any bigger news just yet.
Here’s the technical aside: level of detail management is critical to being able to handle more than a small swath of land. Without it, your graphics card would be quickly overwhelmed with the sheer amount of data it has to draw for no real benefit (see How Google Earth Works for more, at least w.r.t. imagery). The other technical note is that one of the reasons Google Earth would have trouble being a metaverse, allowing you to walk around at ground level, is that doing so requires some heavy duty "occlusion culling," meaning the software has to effectively remove everything that’s hidden from view (e.g., behind something else, e.g., behind a wall when you’re inside a room) or else your graphics card would be quickly overwhelmed with the sheer amount of data it has to draw for no real benefit.
Do you see a common thread there? Rendering fully-detailed planets with the ability to go anywhere and also go inside buildings is considered a hard problem in 3D engine design. Do not try this at home.
Anyway, the biggest beneficiaries of this announcement will not be what you expect. The biggest winners will be the 3D real estate companies who so far have had a hard time pushing their 3D walkthroughs into Google Earth due to some of the above constraints, not to mention that GE has no concept of walking on a floor, which makes a two story house kind of hard to navigate.
These companies will now flock to Multiverse’s free development system (they get their money on the back end, as do real estate companies — see the synergy?) to enable 3D walkthroughs of their real-estate with some added Google Earth content. It’s still not quite as good as doing a real walkthrough in GE, which has over 250 million downloads and can show you important things like school districts and supermarkets. But this will be better for at least one aspect of the real estate sell, plus it adds the ability to offer better interactive content via scripting, which GE lacks. The only downside is the state of the real estate market.
What I’m really waiting for is the announcement of the first 3D game or world that actually licenses GE’s engine, or at least key parts of it, for a new flavor of EarthViewer that’s not strictly limited to the real world. John Hanke has in the past said it’s on the table, though I have no idea what the licensing terms would be.
Virtual Earth, on the other hand, seems more geared towards embedding VE’s world in other people’s products, even going so far as to offer a free API. This welcome announcement may be a sign that Google Earth is finally opening up to more mashups than we can possibly imagine.
Virtual Worlds Conference and Expo – October 10-11, 2007 – San Jose Convention Center
I’ve been waiting for the big Microsoft Virtual World announcement — you know the one to compete with Sony’s Home and possibly Second Life — for a while now. I’m somewhat surprised by the venue, but it looks like it might just come next week at the Virtual Worlds conference.