Whether you think this is a good thing or not depends on whether you like the idea of art being created by machines rather than people. Meanwhile, retailers are using cognitive computing to fight back against the onslaught of online competition.
“With customer engagement and MobileFirst for iOS, shoppers get a unique blend of virtual and physical service. Macy’s On-Call, developed with SatisfiLabs, is an in-store, Watson-powered shopping assistant that lets smartphone users find real-time information about product availability, store layouts and items on sale,” said Burns.
“Boots will be launching Sales Assist, another MobileFirst for iOS app, across its 2,500 British outlets in an effort to make it easier and simpler for customers to find the products they need.”
Trickle down
To say that artificial intelligence, machine learning and cognitive computing have all made significant developmental inroads over the last 10 years is an understatement. However, it is only recently that trickle-down functionality is starting to appear.
“AI and machine learning in particular have gone through big jumps in the last ten-ish years in terms of what they can do, particularly in key areas like image recognition. Machine learning techniques and what’s called deep learning have made a huge difference to how well we can build things like image recognition systems,” said Brian McNamee of the Centre for Applied Data Analytics Research (CeADAR) in UCD.
“Similarly areas like speech transcription—the fact that you can talk to your phone and it can transcribe pretty well what you’re saying—comes from deep learning approaches. This has made things massively more accurate.”
McNamee is a computer science lecturer in UCD and one of the principal investigators in CeDAR, where he carries out applied data analytics research. His research focuses on artificial intelligence and machine learning.
Talking up
“These kind of core jobs are getting better and better, and have made technology appear to leap forward and become more intuitive. Take, for example chat interface, and bots that can appear to hold conversations. You can talk to a bot online and have a reasonably natural feeling of interaction to get a particular job done,” he said.
“It’s not yet at the point where you can have a real free-ranging conversation that can go anywhere and you wouldn’t guess that you’re talking to a computer. But if you want to do a narrow job, you can.”
Cognitive computing and artificial intelligence are gradually developing to the point when they will pass the so-called Star Trek test—where a person can converse using natural language with a computer and have 100% comprehension.
“A lot of what has been achieved has been driven by data, such as the Facebook personal assistant system M. It’s largely person-driven, so basically when you use it you’re having typed interactions with your personal assistant and there’s a person somewhere having typed interactions back with you,” said McNamee.
“But part of the goal is to collect a massive data set to use to train up an AI system that will be fully automated. Facebook will collect, say, 10,000 examples of someone having a conversation about ordering flowers for delivery. Using that, there’s a decent chance a system can be trained up to learn how to have that interaction without needing a person involved at all.”
Data sets
According to McNamee, a big part of the success of cognitive computing is that it is being built on a foundation of enormous data sets.
“Machine language translation is the poster child for that—people have been plugging away at machine translation for years, usually trying to make a logical representation of what an English language sentence says in another language. Typically in the past it didn’t work very well, but increasingly they’re getting better and better as they’re being used more and more,” he said.
While examples of cognitive computing such as these are no doubt fascinating insights into the future of how people will interact with technology, the question remains—how can such technologies be used to improve decision making? This is a point that McNamee feels strongly about.
“The whole point is to make people make better decisions—if we’re not doing that we shouldn’t be messing around with the data in the first place. But there are lots of examples where it does, and that’s the drive behind what we do,” he said.
Ltd
