Milestone Group Quarterly: October 2007
Articles
Face to Face:
Nova Spivack, CEO of Radar Networks
Milestone: Tell us about Radar Networks.
Spivack: Radar Networks is a company that's working on end-user applications of the Semantic Web. In particular, we're interested in knowledge management, collaboration, on-line community and productivity tools. We just released our first product, Twine.com. Twine, like the name implies, ties everything together. It's a new service that helps individuals, groups and teams manage their knowledge more effectively and it makes use of a number of next generation technologies based on the Semantic Web.
Milestone: You say Twine is not about who you know (as in social networking applications) but what you know (the semantic and social webs combined). Is this Web 3.0?
Spivack: Web 3.0, the way I define it, is really the coming third decade of the web, so 2010 to 2020 and during that decade there will be a number of shifts in how software works and in the way the Web functions. These will be based on a number of technologies coming together, one of which is the Semantic Web. The overall theme is that we're transforming the Web from a bunch of documents, basically, to something much more like a database.
In the case of Twine, it is a Web 3.0 application, because it's already making use of these technologies. It's a little bit ahead of its time, I think mainstream Web 3.0 will really be this coming decade but Twine is really starting here and positioning ourselves to be one of the central services of that decade.
When we say it's about what you know and not just who you know, what we mean is that we're combining social networks with knowledge networks to create something that's better than either on its own. So, it's not merely just a lot of information, it's not merely a social network, it's the combination of the two so you can start to see, manage and share knowledge along social relationships in a smarter way.
Milestone: You've said that Web 3.0 will result in Web sites and software that will grow smarter, how are these technologies going to be able to acquire this new intelligence?
Spivack:Well, the key innovation of the Semantic Web is that it can add knowledge that software needs into the data, rather than into the applications. Think of metadata, where you're adding some additional data about that describes it into documents, for example. The Semantic Web takes that that to its ultimate conclusion, you can add very rich metadata to any piece of information whether it's a data record or unstructured text documents and other kinds of things, you can add this semantic metadata so that it's machine understandable.
So it says what is this data or document about? What's it for? Who's it by? What are the concepts in it? What's it related to? And all of that is in machine-understandable form. What then happens is it that it connects and refers to other data in other places that further explains these concepts or other things that are mentioned in the document. So when software sees that it can not only understand the document and its context but it can also find related information to have even a deeper understanding of what that document refers to. All of this happens outside of the applications themselves, so a fairly generic Semantic Web application can have access to an endless amount of knowledge about domains that it might not have specifically been programmed to understand.
In a sense we're moving some of the intelligence that an application would require out of the application and into the data just like the Web did with content in general. It used to be that you had to program an application to render each kind of different document specifically. Now with the Web and HTML, the browser is our universal renderer for any kind of content and it can render many kinds of things. You can see videos, pictures, text documents, you can see all kinds of things in a browser and, similarly, the next generation of browser will be able to understand all kinds of data, even data that the browser was not specifically programmed to understand.
Milestone: You're trained as a cognitive scientist and Twine reflects this knowledge with its focus on semantics. What other cognitive processes are present in the technology?
Spivack: Twine models a number of different processes. For one thing, it's like an associative memory so as you add information into it Twine works behind the scenes to make associative connections, just like we do when we remember things or when we think. Twine can then search along those associative connections, similar to the way that humans think. So a lot of that was modeled on those ideas.
In addition, Twine learns from feedback in various ways and so as we put more information in and as users do things in Twine, the service will start to self-organize and prove in an emergent, bottom-up way, which is also similar to the way we learn and self organize our own concepts and belief systems over time.
I would say that Twine, long term, has the potential to become intelligent, but there's rudimentary kinds of intelligence in it today. In the long-term, it will support an agent model for intelligence, where you might have certain intelligence processes cooperating or even competing within Twine and then presenting their suggestions to a user. In some ways, that's how the brain works. We have different processes taking place and some of them rise up to the realm of attention and we act on them while others continue subconsciously until they are ready to rise up.
Milestone: Your grandfather is arguably the most influential management theorist of all time. Peter Drucker is synonymous in part with the concept of the knowledge worker. Does Twine in some way continue the legacy of his work?
Spivack:My grandfather was more on the theory side and my work is much more on the applied side. I am not really coming up with any new theory or inventing any new kind of worker or broad new idea of what an enterprise is. I am making tools that knowledge workers can use and what my interest has always been is how we make knowledge workers smarter?
I spent a lot of time thinking about collective intelligence, that was really one of the key factors behind Twine was to make a better tool for collective intelligence. One thing I think of, and work on, a lot is this notion of what I call collective IQ, or how intelligent is a group of people? And strangely enough, more often than not, we see that today with existing tools, collective IQ decreases inversely with group size. Basically the larger the group gets, the more stupid it gets and the less productive it becomes. What we really need to do is just flip that ratio because actually the potential intelligence of the group is exponentially greater than the number of people. However, we're really not getting that in practice today.
Twine and other work that I'm doing is really focusing around enabling groups to distill and extract more of the potential collective IQ than they are able to get today so that ultimately, in the best case, a group becomes exponentially more intelligent as you add more members. That's really one of the things that Twine is going to help with. Unlike other information tools, with Twine, the more information you add, the better it gets, which is quite different from things like e-mail or file servers where the more information you add, the harder it is to use. Twine actually is the opposite; as you add more information it actually gets better.
Milestone: Starting with Earthweb, you've seen first hand the evolution of myriad distributed technologies. How do you describe these on a continuum and how do you describe the point at which we have now arrived today?
Spivack:I look at decades. Maybe one thing I have in common with my grandfather is that I tend to look at the big picture. If you go back to pre-Web, say the last half of the 80s, the focus was very much on the personal computer and making a better user interface. That gave rise to Windows and the Mac; the focus was on making personal computers more accessible. Also, a lot was going on with desktop publishing and the de-centralization of documents and early groupware.
When we transitioned to the first decade of the Web in, say from 1990 to 2000, what I call Web 1.0, the focus was very much on the fundamental back end tools of the Web --- Web servers, Web protocols and standards, browsers, the basic underlying technologies that enabled the web as well as some of the first big services. We moved from a desktop distributed model to a server-based distributed model. In fact you could sort of think of the Web as a universal client/server architecture, where the client is the browser and the server is the Web server and those two things can pretty much do everything that old client/server applications used to do.
So we had moved to a more general, more universal kind of client/server architecture to a more generic client/server architecture and moved off the desktop for the most part.
The Web has been a big shift towards the center of gravity being in the cloud rather than on the desktop. Now with Web 2.0(2000 to 2010), the focus there has been on improving the user interface on the front end of the Web. So we see this decade-long cyclical shift, like a pendulum swinging back and forth. First it swung to the front end of the PC, then to the back end of the Web, then it swung to the front end of the Web in Web 2.0 and we'll see it swing in Web 3.0 to the back end of the Web again.
In Web 3.0, the focus is going to be on upgrading the back end of the Web again, fundamental upgrades by adding semantics, natural language processing, open APIs that enable you to connect and match up data, there's a lot going on that will start to make the Web function in a more unified way. Today the Web is still quite fragmented and different services are effectively still separate silos. It's still quite difficult to connect data between them. As we move into Web 3.0, I think data will become more and more portable across applications, it will also become more and more connected and a lot of this will be due to semantics. You can think of Web 3.0 as establishing a better foundation for decentralization and distributed applications, where an application in one place can draw on data and functionality from many other places around the Web, the desktop and the enterprise.
We'll see the pendulum swing towards a better underlying infrastructure which will then, in Web 4.0 (2020 to 2030), focus on a better front-end focus on next generation interactions and interfaces to this data.
Milestone: Besides Twine, what technologies or companies do you see as well poised to take advantage of this current environment?
Spivack:I think there are many companies working this space. One company is Metaweb; they're working on a Wikipedia type project based on the Semantic Web. There's also an Opensource project similar to that, which is called DBpedia, as in database, dbpedia.org. Powerset is a company working on natural language processing, or natural language searching of the Web. We also see companies like Adaptive Blue working on annotation tools that help mark up things that you are looking at with related links.
We also see companies in the enterprise space. For example, Top Quadrant has a platform for Semantic Web applications for use in the enterprise. Franz was once a list company but now they have a powerful new database for the Semantic Web that enables storage of this metadata. Oracle is adding support for RDF and OWL directly for its database products now and actually many other companies are working to start to add support for these capabilities. Adobe and others are adding support for RDF. I think we're seeing this really start to happen. It's still going to take a while before it becomes mainstream because there really aren't a mature set of development tools yet. But we're getting there; and I think that within five years it will be quite mainstream.
Milestone: What would you tell other entrepreneurs is the single most valuable lesson you've learned over the years?
Spivack: I guess it's what Winston Churchill said: "You never, never, never give up." It's also important to really listen, particularly in the early stages of your project and to be willing to make changes that you need to make.
In our case, we started with a vision for the future of the Web and we started almost seven years ahead of that curve so we were early in research and development at the beginning. When we started, it was hard to get any kind of funding or even any serious commercial interest because it seemed like a research project. After many years of R&D, the market has caught up with us and we have focused on educating the market. It required a tremendous amount of faith, "stick-to-it-iveness," and sheer willpower to get through that initial period of time when we were way ahead of the curve. I would say if you are working on a project like that, you really have to be prepared for the long haul. Make sure you're really committed, otherwise don't do it.
I think there is a terrific benefit when you start getting feedback from users and our project is so technology intensive, it was quite some time before we could do that. Formal user studies are one way of getting that feedback. It's really helpful to do those, even if you're not ready to release. That's one thing that's really helped us a lot.
Milestone: What's the one piece of advice you wish you'd listen to in this process with Radar Networks?
Spivack:I think we shouldn't have tried to raise venture capital in the early years; that was a big waste of time. I think, until your product is essentially ready to be used, it's just not worth raising venture capital. You just spend too much time trying to educate investors. We were able to raise capital but it was very hard in those days and we were lucky that we found Paul Allen, who was a visionary but not your typical VC.
A typical venture investor is not looking that far ahead; they're looking out 6-18 months. If your product is not ready to generate revenue in that timeframe, trying to raise venture capital is a waste of time. Instead of going after VC's, we should have just gone after angels or other sources of funding. That said, we did raise venture capital, so we were successful at that, but it was very hard to do in those days before we had a product.
Nova Spivack is one of the leading voices of the emerging Semantic Web, often referred to as Web3.0. Nova founded Radar Networks to develop semantic social software.
In 1994, Nova co-founded EarthWeb (IPO 1998). Nova has worked at Individual, Xerox/Kurzweil, Thinking Machines, and also with SRI International on the DARPA CALO program and nVention. Nova founded Lucid Ventures, and co-founded the San Francisco Web Innovators Network.
As a grandson of management guru Peter F. Drucker, Nova shares his grandfather's interests in the evolution of knowledge work. In 1999 Nova flew to the edge of space in Russia with Space Adventures.
|
