Enterprises face the challenge of staying afloat on an ocean of big, unstructured data. Individuals struggle to make sense of search results that return countless thousands of hits, including irrelevancies, dumb duplicates, and false positives.
Certainly, there are all-purpose analytics tools, and algorithms like those used by the major search engines, which make a start -- a good start, even -- on sifting the vast sands of information for the gems we really want. We still have a long way to go, though, before we can be confident that we're handling data in the most focused, agile, and customizable way.
Expertmaker, a Swedish company, is trying to close the gap with its artificial intelligence software solutions. I had a chance recently to speak with Lars Härd, founder and CTO of the company, about intelligent data mining and Web search, and about Expertmaker's bold claim that AI will "disrupt how the Internet works," making the online experience "much better" for enterprises and customers.
For example, Expertmaker contends that search technology is based on a "12-year-old algorithm" and leverages user history rather than responding to real-time personal interests. Understanding relationships, Härd told me, is as important as being able to capture and mine data, if you're interested in obtaining intelligent output.
Right now, Härd said, our Web experience throws up a sea of unstructured text and images, more or less sorted by search engines. In the future, if he has his way, it will be about "computing results." He promised that, "by using new tools, we will replace media content with knowledge."
Expertmaker offers a platform -- including server, toolkit, and API -- which leverages "machine learning" in the creation of customized, high precision, "human-like" output. Machine learning is the branch of AI devoted to helping algorithms learn to evolve based on data input. "I really need a white board to explain it," said Härd. Instead, he gave me a concrete example.
A major e-commerce enterprise, with millions of items in stock, was returning hundreds of thousands of search results but was simplistically ranked, leading to a confusing and frustrating experience for customers. By applying predictive analytics to user behavior data and advanced modeling, machine learning was able to teach the search engine to modify and personalize the customer's online experience in real time.
The company was able to add the AI functionality as a highly graphic and interactive layer within its existing Web environment. "It was a very nice package."
AI isn't just for the enterprise. Härd said he's been playing around, developing his own personal platform for analyzing and sorting incoming email. "It's our mission," he explained, "to lower the threshold for using data mining, analytics, and AI. Expertmaker claims that no AI knowledge or programming experience is required to use the platform."
"We also encourage individual developers to get involved." One vehicle for this was the recent hackathon sponsored by Expertmaker, together with Vodafone Xone, which invited contestants to use the AI platform to tackle "content and news overload on mobile devices, and create intelligent mobile app solutions for the challenges faced in a data-flooded world."
Among the winners was "BrainyCraig," a dynamic learning solution for discovering products on Craigslist using connected keywords to map concepts to products. In other words, it helps users find products without knowing the exact term that the seller is using.
This is a simple illustration of Härd's aim of driving our online experience in a less haphazard, more computational direction. In this context, he mentioned the "semantic Web," an analogous and long-standing project that seems ever less likely of coming to fruition. In other words, he and his Expertmaker partners, have picked a big battle.
But perhaps the prospect of an Internet that can teach itself to analyze and filter data with ever more precision and intelligence is worth the fight.
— Kim Davis , Community Editor, Internet Evolution