We’re not quite at The Matrix stage yet, but as the amounts of data stored in databases continues to increase and be used in new ways, machine learning experts say artificial intelligence will become more and more common in the enterprise.
Looking to capitalize on that coming trend,
Edmonton’s CodeBaby Corp. has signed a memorandum of understanding with the Alberta Ingenuity Centre for Machine Learning (AICML) at the University of Alberta to share information about machine learning.
A software development company, CodeBaby is commercializing an avatar-based personal virtual assistant that sits on a company’s Web site to guide a visitor to what they’re looking for, or help them with forms or tasks. They’re on the commercial end, using machine learning to improve the functionality of their products, while the AICML is a government-funded centre of university researchers investigating machine-learning techniques.
“”We’re hoping the relationship will help us generate ideas for implementing artificial intelligence into our products,”” said CodeBaby CTO Neil Lamoureux. “”We see the trend more and more in technology moving to artificial intelligence. The U of A has some world class people in AI and we want to tap into that.””
Lamoureux said CodeBaby wants to get an idea of where the leading edge of research in machine learning is right now, what could result, and see if it might be applicable for their own products.
“”We’re starting to see some machine learning, but I think it’s going to become more and more complex,”” said Lamoureux. “”I think we’re on the cusp of artificial intelligence making it into business and consumer applications.””
AICML manager Michael DeMarco said the centre was opened in September 2002, around an already strong cluster of AI talent at the U of A, to begin looking at the possibilities machine learning could bring.
DeMarco said the agreement with CodeBaby allows both groups to protect their intellectual property and share information about their research around machine learning, with the possibilities for sponsored research projects down the road.
“”We’ll have detailed discussions with them about what our researchers are doing and how it might benefit them,”” says DeMarco. “”They’ll make available to us information about what their experience has been on the commercial side of the equation.””
DeMarco said people are beginning to realize that at a certain depth of data, you get the sort of computation problems for which you need machine learning.
Geoffrey Hinton agreed, adding that boiled down, machine learning is really just fancy statistics. A professor in the department of computer science at the University of Toronto, Hinton also specializes in machine learning.
“”There’s no real sharp line between machine learning and statistics,”” Hinton said. “”It’s used all over the place, anywhere people want to go looking for things in big databases.””
Supermarket loyalty cards are a good example, allowing supermarkets to track what their customers are buying. Stores can use that data to organize products in the store or generate special offers.
Another example Hinton uses is credit card companies, which monitor the types of transactions a customer makes looking for patterns, with sharp variations from those patterns being a red flag for possible credit card fraud.
“”Many of the big companies that have the resources are researching this,”” Hinton said. “”But the little companies are all trying to survive until tomorrow, typically they wouldn’t have machine learning researchers, although they’ll often make use of the methods.””
However, for the most part, the U of A’s DeMarco said companies are just beginning to look at machine leaning, either by building their own machine learning groups or making connections with research institutes.
“”I would say it’s really at the beginning of the market opportunity for this sort of technology, my view is it will be far more prominent in all aspects of life just as the sheer volume of data increases,”” said DeMarco. “”There are a lot of companies that first need to be educated about what the possibilities are before they can really begin to implement it in their own technologies.””