Interview With Powerset’s Barney Pell

A couple of weeks ago I wrote an article about Powerset and their recent restructuring. Today, I talked with Barney Pell one of the founders and former CEO of the startup. As I have said before many times, Barney is an extraordinary person with a brilliant vision and Powerset has been one of my favorite topics of discussion. As far as my previous post, I had some concerns over the underlying causes for Steve Newcomb's absence and also Barney's shift away from CEO. So, we had a great talk as always and I thought our readers would be interested in where Powerset is and also in what Barney had to say.

The Gold Ring

Because of the tone of the last article I was compelled to ask Barney about Steve Newcomb's departure. Barney said that he would rather concentrate on where Powerset is and where it is going if it was alright. In thinking of this, I actually understood his point of view because he obviously cannot speak for or about Steve, and as a point of focus I am sure he is tired of fielding these questions. I really don't blame Barney for wanting to move forward and so with that I should tell you all about our chat. Barney and Powerset are quite obviously focused now on the gold ring at the end of their vision.

Powermouse Evolution

I was curious about how the Powerset team had come up with Powermouse, and Barney explained that as the engineers worked on the vision – the need for developing an interface to set the queries and access the database became paramount. In creating this interface the team discovered how interesting it was to actually see how Powerset functioned across it and decided to make the interactive tool available for everyone in Powerlabs to see. I must say, that discovering natural language search via the little tool is fascinating. One can easily see how complex even the simplest queries are, and also follow the tail of essentially how Powerset's engine thinks.

Expanded excerpt from Powermouse – the tail of the mouse?

Natural Language Expanded

I have had many conversations with my friend Dr. Riza Berkan of hakia about the extended uses of semantic search, and today I asked Barney about future applications of the Powerset technology. It is important for the reader to understand that when we speak about these search engines, we are not simply talking about returning links from a keyword or two typed into a search bar, but a search capability far beyond the way we view it now.

With the advent of AI and true natural language search the search interface may seem very simple, but the degree of interaction possible is not really easily imagined. Barney said that: “In the end, a person might simply ask for as list of something, and the query result will be just that – a list or timeline reflecting the user's rich human intention.” This is the power of true semantics and AI, to effectively return exactly what a search is intended to find – not simply a row of popular or ranked results.

So, a site like The New York Times or Reuters, where finding and exact story or report is often a nightmare, and engine like Powerset can and will be applied to not only find results with more relevance and faster, but also initiate a process for much greater usability. This is pretty nebulous I know, but hakia and Powerset are already revealing a much deeper “understanding” of the language behind the website. These technologies can be used for a far more broad spectrum of uses and collaborations than are readily visible.

From Powerset Quotes – At least parity with conventional engine apparent

Cool Collaborations

Powerset is now making active use of Freebase, a fascinating accessible database I wrote about back in August. The database has gown geometrically since then and these “narrowed” queries provide a very powerful tool for Powerset and others to integrate with natural language. Barney and company are also making heavy use or WordNet, Princeton University's cognitive science lab, to enhance their lexical English database and conceptual semantic relationships.

Tentative Results

From what I saw doing some simple comparative searches at Powerlabs, I estimated that Powerset already has the capability to exceed most conventional search results by a margin of 25%. I asked Barney if this was a valid assumption and he assured me that testers have rated this much higher – or about 60% more effective or desirable than other engines. I really do not doubt this, especially given the narrow test medium of Wikipedia (which all test queries are run on). Given the short span of time Powerset has had with WordNet and really in developing some of the most advanced backend furniture on the planet, the Powerlabs is actually quite impressive from an aesthetic and a functional point of view.

Conclusion

Am I still concerned about Powerset's management shift – yes frankly, but I know that Barney would be as candid with me as possible and his enthusiasm and obvious vision is still fully in tact. The thing we talked about is that all these startups begin in rather archaic fashion but with their goals clearly set. However, Powerset has become a much larger organization than the “just moved in to the office” group I wrote about initially, so some change was evident as seen in other large startups.

I think the innovation and the drive to achieve true natural language relevance is still strong (maybe stronger), but the curious part of me wonders what went on with Newcomb and etc. In the end, it is the art and the effectiveness that will tell the tale I know. I still think hakia is ahead and has the better architecture personally, but I am not discounting Powerset in saying this.