Talk is cheap, but security is king in the AI future of conversational commerce.
Researchers at TSSG in Waterford are on the case to make artificially intelligent (AI) voice-based commerce tangible and safe.
The speed of the introduction of services such as Apple’s Siri, Microsoft’s Cortana and Amazon Echo’s Alexa makes it easy to forget that the AI future we all dreamed of is actually here.
Yesterday, Gartner fellow and vice-president Tom Austin pointed this paradox out when he said that most of the AI on our smartphones – many of which have been with us for years – are taken for granted.
‘We are seeing cool apps coming out but we have to make sure that we underpin the technology by addressing core security and privacy concerns as well’
– ERIC ROBSON
And yet, the hottest areas of tech today – AI and machine learning – hold the key to a whole generation of services that will transform our lives forever.
For some, this is already happening. The combination of APIs, the cloud and AI will enable users to simply talk to their phone or Amazon Echo devices and, in a conversational way, ask what their bank balance is like, if they are spending wisely this month, or to order some milk and bread from the nearest Tesco. It is not beyond the realm of possibility for your goods to arrive by drone minutes later. That is the near future.
At present, researchers at TSSG in Waterford are working hard to iron out the kinks from a security and privacy perspective, to enable the compelling AI and voice-based experiences of tomorrow.
TSSG has a track record of building start-ups around evolving technology opportunities. Two years ago, Red Hat acquired TSSG spin-out FeedHenry for €63.5m and recently announced 60 new jobs.
Just this week, TSSG virtual reality (VR) spin-out Immersive VR Education raised €1m to continue to develop breakthrough VR games, including the recreation of historic episodes such as Apollo 11 moon landing and the sinking of the Titanic.
The sounds of commerce
Eric Robson, research unit manager of data mining and social computing at TSSG, told Siliconrepublic.com that the company is working on the next breakout area of AI, involving the area of voice or conversational commerce.
“We are starting up an AI initiative at TSSG all around how people want to communicate with their applications and services,” he explained.
Robson said that conversational interfaces hold the key to future services in the home, at work and on the move.
But he said the solutions have to be both discrete and secure. “They have to be a very passive, unobtrusive mode of engagement. Examples of conversational interfaces that are very real and that are taking over include Amazon Echo’s Alexa and Google Home, which will be coming to Europe soon. The accuracy of services creates by Amazon, Facebook, Google and Apple are numerous.”
However, Robson warned that these services are not perfect.
“What needs to happen on the AI side of things is the technology needs to be able to break down the conversation better, understand the intent and give the user the result they truly wanted. The technology is there,” he said, but it is far from ideal.
“What we are doing is we are building a whole tranche of apps and services that will be of immense value to consumers on a day-to-day basis.”
A recent debacle, which saw a six-year-old girl in Texas ask Amazon Echo, “Can you play dollhouse with me and get me a dollhouse?”, resulted in Alexa ordering a mansion dollhouse and four pounds of cookies. This was compounded when a local TV station played the story on TV and it triggered similar orders on countless viewers’ devices.
“It begs obvious questions around security,” Robson pointed out. “Data privacy is also a hot issue. We are seeing cool apps coming out but we have to make sure that we underpin the technology by addressing core security and privacy concerns as well.”
Robson said that his team of TSSG researchers are working with a number of SMEs in Ireland to advance conversational commerce.
“The challenge is ensuring the machine understands what a person said, and aligning that back to an API call that either gives an answer or triggers a service.
“A lot of that comes down to machine learning and the cognitive power of computing.”