The Alexa Prize
Amazon's $1million Socialbot Challenge

25th April 2018

Voice technology has taken the world by storm over the past decade, becoming increasingly smarter and integrated into more every-day appliances. While the capacity to follow basic commands has long been achieved, perhaps most commonly with the iPhone's inbuilt personal assistant Siri, significant money and resources are now being invested in the development of more natural interaction with conversational artificial intelligence.

Today, almost every major tech giant has launched their own brand of voice technology, such as the Apple HomePod or Google Home, in the race to create software that can not only carry out commands but is also capable of natural conversation. Leading the charge head-on, however, is the Amazon Echo with its voice-controlled intelligent personal assistant service Alexa.

Since its launch in 2014, Amazon has sold over 22 million Echoes worldwide and employs over 5000 people specifically to work on the Alexa platform. In 2017, the company launched their $3.5 million answer to the Turing Test: the 'Alexa Prize'. In the same way that the Turing Test measures a machine’s ability to demonstrate intelligent behaviour equivalent to that of a human, almost 70 years later the Alexa Prize focuses on developing a new highly intelligent 'socialbot' skill for the Echo.

The challenge set by Amazon was simple: 15 teams of the world's best computer science graduates would use the Alexa Skills Kit (ASK) to build "a socialbot that can converse coherently and engagingly with humans on popular topics and news events" for 20 minutes. The areas of conversational AI to be advanced would include knowledge, understanding, context, planning dialogue and common sense reasoning, and the winning team would secure academic glory and the promise of brilliant future careers, as well as taking home a cash prize of $1 million.

Following ten months of competition where the teams worked tirelessly to advance human-like language generation and expression for Alexa, in November 2017 the overall winners were Team Sounding Board from the University of Washington, whose socialbot raked up an average conversation time of 10 minutes and 22 seconds. This just surpassed the halfway point of Amazon's 20-minute goal, so it was quite fitting that they were awarded half of the grand prize: $500,000.

Czech Technical University in Prague and Heriot Watt University in Edinburgh were awarded second and third place, taking home $100,000 and $50,000 respectively, and all university teams participating received a $250,000 research grant, Alexa-enabled devices, free Amazon Web Services (AWS) to support their development efforts, and access to other tools, data, and Alexa team support. With the 2018 competition already well underway, it appears that Amazon will continue the Alexa Prize as an annual competition until their goal of 20 minutes of engaging, high quality conversation with Alexa has been achieved.

With so much money and resources being poured into this research, the likelihood of voice technology transforming new spheres of life and business seems inevitable. In insurance, companies, such as DAS UK for example, have launched new Amazon Echo skills to aid the accessibility of insurance information, enabling customers to ask Alexa a variety of questions regarding legal expenses, insurance and a range of common legal issues.

CDL has taken this a step further, integrating its insurance retail solution, Strata, with Alexa to enable policyholders to make mid-term adjustments, such as adding a driver or registering a change of address or vehicle, as well as retrieving policy details, including the number of years no claims bonus or excess amounts.

Beyond this, project teams in the CDL Incubator have made great strides in creating voice interfaces to deal with routine calls in the contact centre environment and support the drive towards realising a 'virtual call centre'.

On the one hand, the benefits for consumers are evident: greater convenience, ease and accessibility in managing their affairs and contacting retailers at a time that suits them. Meanwhile, for retailers, there are equally apparent gains in efficiency and improved customer service, as routine queries are automated and operators are freed up to deal with more complex queries.

Replicate this across other sectors and the potential value riding on the quality of this technology is clear; and it's significantly more than Amazon's $1 million reward.