Believing the hype: Here comes the age of artificial intelligence

Has BI finally cracked its big data challenges? Is 2015 the year AI finally goes big? We’ve got the tools and there’s no doubt the tech is smart, so what’s the hold up?

Part of it's to do with funding.

The history of AI research is one of high hopes followed by crushing disappointments, and that’s reflected in the way AI projects get funded. AI, as a field of study, has experienced periods of unrestrained optimism followed by devastating setbacks, followed again by even headier periods of optimism. And those oscillations are acutely felt by those who control the purse strings.  

Those purse strings are about to get significantly loosened however, with companies – including if-anyone-can-do-it favourites Google ­– announcing big new initiatives and actual real-McCoy AI tech to back up the big talking.

Rise of the (smarter) machines

We’ve talked about the rise of the robots and health recently – it’s a rapidly expanding industry, especially for mobile – and that’s an opportunity start-up Your.MD plans to exploit with its new AI project. On the surface, it looks like a run-of-the-mill web diagnosis site, à la symptoms.webmd.com. What makes the product unique however, is that its website/appplatform uses machine learning to include variables not available to other, keyword-based diagnostic tools – such as reports of related symptoms in your area – meaning less potential for a hypochondria-inducing misdiagnoses.

Image: The Your.MD web interface

“The more people report the symptoms of the seasonal flu in a town, the higher the probability that the next user from the same place complaining about high fever and coughing will have picked up the same flu strain,” Matteo Berlucchi, Your.MD CEO told TechCrunch.

“In order to achieve this result without the support of human beings, you need to train an AI system to ask the same questions a doctor would ask and back the results up with reliable and clinically assured medical information,” he says.

“The solution we provide is scalable because of artificial intelligence, so that you can offer it to more and more people and make it better.”

If programming is the engine of AI, data is most certainly the fuel. More data produces better results, so if anyone’s looking closely at the possibilities offered by the big data revolution, it’s Google.

The company says it can improve on its email service’s 99.9 percent spam catch-rate by adding AI into the mix to detect and block spam emails.

“Now, we are bringing the same intelligence developed for Google Search and Google Now to make the spam filter smarter in a number of ways,” product manager Sri Harsha Somanchi said in a blog post last week.

He says that the spam filter used by Gmail now uses an ‘artificial neural network’ to detect and block the “especially sneaky spam—the kind that could actually pass for wanted mail”.

Munich-based company Neokami says it’s currently developing an AI-driven platform that will be able to predict the movement of stocks using traditional data, cross-referenced with outside ‘variable’ data such as Twitter sentiment and press coverage.

Ozel Christo, founder and CEO of the company says that the technology can “analyse millions of variables within seconds and create a customized predictive model for any stock”.

In Christo’s estimation, the program will have between 75 to 95 percent accuracy, a claim which coincides with the company’s announcement last Tuesday of a new $1.1M funding round.

This wouldn’t be the first time someone looking for funding says we’re at the tipping point of artificial intelligence going big. Moore’s law being what it is of course, things have got to happen sometime, but where are the results?

Taxonomy

Perhaps it’s a matter of definitions. While big data crunching and machine learning are both trademarks of what we’ve come to expect from AI, in the popular imagination, AI has an air of the magical about it, an idea that, somehow an intelligent machine is a sentient machine.

Well that’s a tricky thing to measure. In the past, the popular benchmark for machine 'intelligence' has been the Turing Test.

Simply put, the Turning Test is defined as “a machine's ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human”. The Turing Test seems to say: ‘If you can build a machine that acts enough like a living thing, it just might be one’.   

Image: Alan Turing

It’s certainly a strange concept, and, as a definition, one that seems somewhat incomplete, but that sense of incompleteness may say more about our limited understanding of what consciousness and intelligence actually is, rather than issues around the impressiveness of this feat.

So why is this important?

Because a computer passed the Turing Test last year.

60 years after Turing’s death, Eugene Goostman, a computer simulation of a 13-year-old Ukrainian boy pulled a real-life Pinocchio, convincing 33 percent of a panel of judges that it was a “real boy” over the course of a five-minute text-based conversation. (Take a look at the kind of conversations the Eugene Goostman program can participate in here).

And now even that’s old news.

Google have announced that its DeepMind artificial intelligence research is teaching computers how to have “context-sensitive” conversations about everything from philosophical issues to IT help-desk duties.

Google’s system doesn’t use the hand-coded base-information that forms the backbone of traditional chatbot technology, rather, it ‘learns’ language based on third-party documents that it’s supplied with. It’s that ability to learn and adapt that driving speculation that Google new DeepMind tech could soon power IT helpdesks without human interference.

Big data is big business

Big business gets interested in AI as a way to glean insight from data sets too large for un-aided human analysis. When big data and AI collide, startling insights can result – insights previously beyond human perception.

Computerworld offers an excellent description of this relationship:

“To date, humans had to come up with hypotheses, identify relevant variables and then write algorithms to test these theories against the information collected in big data sets.

“However, as data sets become larger, the ability for humans to make sense of it all becomes more difficult, and limits the insights that can be gained from all this information.”

Artificial intelligence may not yet be swimming in the mainstream, but New Zealand businesses are at least growing more aware of its potential. A recent survey conducted by the EMC Corporation found that two thirds of Kiwi IT decision-makers believe data analytics are important to their current business strategy.

Ed Hyde, CEO of Qrious, Spark’s big data business, put it this way: Senior leaders “intuitively understand the potential of data driven innovation but feel there is a gap between this potential and capability to deliver.”

That’s the challenge and the opportunity Auckland-based enterprise software company Touchpoint plans to exploit. The company has just received a $150k grant from Callaghan Innovation towards research and development of its own AI project, dubbed “the world's angriest artificial intelligence program”.

Business is becoming more complicated,” says Frank Van Der Velden, chief executive of the company. “With more and more regulation it’s harder for business to satisfy customers, and those customers have higher expectations.”

That’s leading to overburdened call centres, more angry customers and, ultimately, lost profits, he says.

The solution, says Van Der Velden, is ‘Radiant’, new software that can analyse hundreds of millions of angry customer interactions and then simulate those interactions to help companies better understand the behaviours and processes that trigger customer rage.

It does this by simulating the ‘worst’ kind of customers – grumpy, rude and impatient – and runs a range of ‘what if’ scenarios for call centre operators to train with. Armed with the Radiant software, businesses can identify their customers' pain points in an ‘off-line’ way, and redesign the way their call centres operate to produce the best outcomes for everyone.

“Our AI engine learns right across a range of different past interactions, some of which have and some of which haven’t worked out well.”

“We’re building an engine that can recommend solutions to companies around how they can improve the way they deal with certain issues that customers are facing. It provides companies with advanced warning of systemic issues. It provides people at the frontline with recommendations on what is the best resolution for a customer’s issue.”

“Think about trying to reproduce that sort of analysis manually. It’s just not possible.”

Image: Frank Van Der Velden, CEO and co-founder, Touchpoint Group

Van Der Velden says that we generate so much data, and that there are “secrets” in that data, but it takes the right technology to reveal it.

“We’re already capturing so much information,” he says, “but we often don’t know what to do with it. If we can actually learn from it on a continuing basis, that’s incredibly powerful.”

“It’s definitely complicated technology, but at the end of the day, it’s still about listening to what the customer is saying and using that as a way to improve performance.”