Voice Assistants on Brink of More Meaningful Interaction, Says Analyst

Amazon has been making “steady progress” toward more interactive experiences with Alexa through conversational AI, Forrest said. This year, users should be able to ask Alexa to join a conversation, and the voice assistant should be able to “interact intelligently in a multi-person discussion,” he said. In practice, conversation will still be limited, but in demonstration, “Alexa appears to distinguish between the questions intended for the humans versus those directed towards the AI system,” said the analyst.

Assuming conversational voice platforms develop intelligently, Forrest said, "this should naturally elicit deeper and more meaningful interaction," which would allow virtual assistant vendors to learn more about users and personalize responses. Beyond that, "we may experience 'intelligent interjection,'" where the virtual assistant might start a conversation or make contextually relevant suggestions, "therefore influencing a user’s decisions and behaviours, potentially towards monetizable activities." He envisioned an opportunity for location-based services with voice interfaces to deliver "precisely-timed and contextually relevant information." Having a virtual assistant "in the secure channel that is your ear unleashes several possibilities, and one can imagine advertisers and service providers wanting to speak to individual users via that interface, capitalizing on the opportunity to influence what they’re doing at the time." He imagined a virtual assistant directing a user via earbud "to a restaurant across the street or reminding them of a special offer at a nearby clothing outlet."

Voice-based e-commerce is currently restrictive, Forrest noted. Screenless devices often have to resort to reading out lists of options, an interaction that’s both “unnatural and distracting.” Users aren’t accustomed to allowing AI to make purchases on their behalf. The AI software must become highly personalized beforehand and “know” what users’ preferences are, “since users will quickly become unforgiving if a virtual assistant orders the wrong items.”

Futuresource forecasts 80% of CE products sold in 2024 will have voice assistants, dominated by Google, Amazon, Apple, Baidu, Microsoft and Yandex. Forrest also sees growing opportunities for “domain-specific” assistants that are more task-orientated. They will be used for playing and discovering music, for smart home control, and to address specific tasks in the car, for instance. In the U.K., the BBC is creating an assistant called Beeb that’s designed to interact specifically with the broadcaster’s content and services.

Qualcomm’s announcement Tuesday that it will integrate Alexa Custom Assistant with its Snapdragon Automotive Cockpit platforms will enable OEMs to equip each brand with unique wake words, voices and capabilities (see report, this issue). “Many companies are already working to place their brand identity into voice with their own 'wake words,'" Forrest said, highlighting automotive vendors that want to maintain strict control over the in-vehicle experience.

In the past decade, the industry has transitioned from a voice-controlled platform with AI to the virtual assistant as an AI-based tool controlled by voice. In the future, virtual assistants promise “a frictionless way to interact with products and services,” though the industry is several years away from having “voice first” interfaces, and it may never become truly independent of screens. “After all,” Forrest emailed, “text-based internet search is still far from optimised,” delivering pages of results that humans have to scrutinize to find items of interest, he said. Voice-first interfaces “must deliver the top answer, each and every time,” he said, “so the internet giants must discover new methods and develop AI capable of surfacing precisely the right result.”

Forrest also referenced ethical questions on how far the personalities and behaviors of virtual assistant AI might be allowed to develop. “Much of what is achieved with virtual assistants today surrounds factual questions with definitive answers, or instructions to control products and services within the immediate environment of the user,” he said. The virtual assistant is working within a tightly defined frame of reference, "aiming to be realistic, honest and truthful in handling various intents.”

Virtual assistants don’t have true personality or exhibit genuine interest now; neither is the AI particularly friendly nor tactful in building relationships, Forrest noted. Many of these aspects are holding back the next stage in virtual assistant development, he said, but whether machines could ever mimic emotional intelligence and empathy, or even approach anything that might convince humans, “is still an open question.”