Max Headroom may have been a legitimate pop-culture sensation when he debuted in 1985, but a new breed of body-less digital avatars is taking deepfakes in a whole new direction, as an Australian entrepreneur debuts a new talking-head engine that he believes will push ‘conversational commerce’ into the mainstream.
Crowd Media – whose engine combines natural-language speech recognition with AI-based digital heads capable of carrying on natural-language conversations – has been investing heavily to build out a flexible customer-engagement platform that is now available in beta form as a digital double of founder Domenic Carosa.
The company – which in recent years has built a chatbot database of over 180 million user-submitted questions and answers from subject-matter experts – is now tapping technology from UneeQ Limited and VFR Assets and Holdings Limited to build a flexible platform that would allow celebrity heads to be integrated into all manner of fan, educational, recreational and e-commerce platforms.
Crowd Mobile is based in the Netherlands, reflecting its 2015 acquisition of Dutch mobile-payment company Track Holdings, but Australian executives including Schapera and Carosa – a serial entrepreneur who was once the youngest CEO to manage an ASX-listed company – have their sights set on global opportunities for the venture.
Executive chairman Steven Schapera already speaks openly about the possibility of allowing consumers to have a Q&A session with a digital Lewis Hamilton, cooking lessons from a virtual Jamie Oliver or “learning about the Battle of Trafalgar by Facetiming with Napoleon” – promising what he calls a “tectonic” commercial opportunity.
Digital human, or not quite right?
Despite their relatively smooth speech and almost-verisimilitude, the digitally generated heads suffer the same ‘uncanny valley’ effect that led digital motion-capture Christmas film The Polar Express to be resoundingly labelled as “creepy”.
Yet online users have responded well to previous simulated people – such as ‘virtual digital influencer’ Zoe Dvir and AI-based news anchors – but Crowd Media’s decision to set its sights on emulating actual humans, including ‘synthetic voice cloning’ from partner Aflorithmic, is poised to take deepfakes – and concerns about them to a whole new level.
“Although it may seem fake to you today, by the end of next year it will appear very real indeed,” Schapera said in launching the beta version of ‘Digital Dom’ this month and promising a platform that could ultimately be used to create an interactive version of nearly anybody.
Crowd Media is hardly the only company toeing the line between real and virtual people, particularly since new AI-based image-analysis capabilities are finding a whole new range of applications.
Online genealogy site MyHeritage, for one, this year debuted a ‘Deep Nostalgia’ feature that generates simulated, moving 3D photos based on any portrait they are fed.
Designed to breathe new life into often flat photos of long-lost relatives, Deep Nostalgia uses technology from D-ID to build a 3D model by extrapolating a range of facial features from a 2D photograph – and then animating the face along one of a series of pre-defined paths.
By putting a human face on well-established chatbots, talking-head systems are poised to take conversational commerce to a new level, with a recent Deloitte analysis suggesting that the Asia-Pacific region would lead the world in its uptake – with the market growing at over 20 per cent annually, to more than 10 billion interactions within a few years.
“For retailers that get conversational commerce right,” Deloitte Digital Conversational Commerce Director Simon Stefanoff said, benefits include “reduced customer churn and customer service costs against those who did nothing.”
“At its heart it’s about building brand trustworthiness and brand agility to be front of mind when customers are making a purchase decision,” he said, “and ultimately bringing that personal, in-store sales experience into a highly personalised and convenient ecommerce scenario.
“The risk of not building this capability is that retailers could very quickly look slow, inflexible, and impersonal.”