Vortex DNA - Monetising web relevance via auction
On Friday at the Tim Harford lunch, I had the pleasure of finally meeting in person Raf Manji from Vortex DNA (Raf’s commented on this blog occasionally). We met up later in the afternoon for a coffee, so I could learn a little more about what Raf and his colleagues are up to.
Vortex DNA in its most public form offers a downloadable enhancement to the Firefox web browser which markedly improves the relevance of your web browsing and suggests other results which should be of interest to you. The video clip below explains it far better than I can. The key is a seven digit code (your DNA) which links to information which matches your profile. The system is self-learning - it updates itself as you change your interests. Privacy is a strong feature, since only the code is exchanged.
However, useful as that sounds, it’s not the primary reason for my interest. Vortex DNA is auctioning the rights to use its algorithms in selected vertical markets, such as insurance. This is a novel way to monetise intellectual property, and I’ll watch the outcome with interest. If it works, I know several other clever people who might go that route.
Trackback uri
March 4th, 2008 at 7:56 pm
Hi Jim,
Thanks for the coffee and the heads up on the lunch with Tim. The first chapter of his new book is a real barnstormer.
I’ll let you know how the auction process goes.
Regards
Raf
March 4th, 2008 at 9:22 pm
Jim said…
The system is self-learning - it updates itself as you change your interests. Privacy is a strong feature, since only the code is exchanged.
Umm! But that is hardly new at all. Self-learning & adaptive algorithms are the heart & soul of machine learning domain. I have been using the VortexDNA plugins for about 5 months now, and I do question its claim based on my experience on using it.
The fact is, there is no such thing as proprietary algorithm that is going to remain secret for ever. There are millions of researchers around the world, and the chance that somebody, somewhere might have independently co-invent some algorithm with another person from another country is quite high. As it turned out, this often happens. The 2 independent inventors only aware of each other’s work, once their inventions are published which are then being made publicly available .This happens all the time not only in the computing domain but in other disciplines too.
I would like to see VortexDNA release its test-data (not their algorithm) on their website, so others could use the exact same data to bench-mark VortexDNA algorithm against other recent algorithms. There are lots of test data available on the internet for benchmarking algorithms in the domain of search engine, recommendation engine, image classification (pattern recognition), breast-cancer detection, etc, etc, such as the popular UCI Machine Learning Repository.
Perhaps, VortexDNA could download those datasets (one that is relevant to their algorithm), and run their algorithm on. Those datasets at UCI are usually based on some published algorithm by researchers, where they decided that if any one out there wants to reproduce their claim in their publication, then they can download the data from free repository such as UCI. Usually , those datasets accompany the test & benchmarks of the algorithms of those authors in comparison to previous or existing algorithms.
March 4th, 2008 at 9:29 pm
The UCI full dataset repository is found here :