It’s the household device people love to hate. Ever since Amazon unleashed Alexa back in 2014, over 100m of its Echo and Dot gadgets have been installed in homes around the world. From Durban to Dubai, consumers have embraced the voice assistant devices for everything from playing their favourite music to booking movie tickets, finding recipes or keeping track of their sports teams.
But for Alexa’s creator and head scientist, Rohit Prasad, this is only the beginning. In future, brace yourselves for a far brainier Alexa – which intuitively understands our needs and desires and becomes a proactive companion, prodding us into things we should or shouldn’t do rather than serving us as a passive assistant.
Alexa has already become ‘much more intelligible, much more fluent, much more natural in terms of achieving human level qualities [of speech recognition]’, claims Prasad, 45, a machine learning and artificial intelligence expert from India who joined Amazon – one of the world’s most valuable companies – in 2013.
Despite its runaway success, Alexa’s rise has been shot through with controversy, amid concerns over privacy, surveillance, intrusive data collection and ingrained bias embedded in the AI on which it relies.
Many will be appalled at the prospect of a bossy Alexa calling the shots in their home. But Prasad remains an unalloyed optimist about the future of the technology.
In the five years since Alexa was launched, he claims there has been a fourfold increase in Alexa’s ability to correctly detect individual words and understand their meaning. And with hundreds of millions of dollars flowing into Amazon’s Alexa research labs, which he oversees, the technology is becoming increasingly powerful. What comes next?
In future, he says Alexa will evolve from being relatively limited in its scope and capacity for reasoning into an increasingly intelligent and prescient presence in our lives, able to understand us better and deal with more complex queries and situations.
Prasad, who lives in Boston with his wife, gives the example of organising a night out with your family. In future, Alexa will not just tell you what time the films are showing at your local cinema. It will purchase the right number of tickets, help you book a table at a nearby restaurant, order a taxi, or advise you on what time to leave your house to beat the traffic if you plan to drive.
These are all ‘very complex tasks that require a lot of cognitive burden’ where Alexa ‘will practically help get you connected with the right skill or service’, he explains. From an engineering perspective this is challenging stuff, admits Prasad, who grew up in the impoverished Eastern Indian state of Jharkhand, and studied at the prestigious Birla Institute of Technology before moving to the US to study electrical engineering at the University of Illinois and MIT.
He likens building the AI that underpins Alexa to teaching it to play a game of 3D chess – only far more complex. ‘You can make an analogy with chess but this is much harder because the goal is ill defined,’ says Prasad, who worked for Raytheon, the US defence manufacturer, for over 13 years before joining Amazon. ‘In a game like chess, it’s very clear what the end game and objective of the game is, whereas in a human-machine [interaction] the objective is unknown.’
It’s not just rapid advances in machine learning and AI that will drive this progression, he says. Prasad also hints at equipping Alexa in future with new sensory capabilities that will enhance its understanding of the world surrounding it.
So is the next stage for Alexa to sprout arms and legs and turn into a walking, talking robot?
Prasad demurs but acknowledges that if Alexa is to become better at understanding human speech and behaviour, it needs to be equipped with new sensors and deeper powers of perception. Fitting smart assistants with cameras and physical bodies would ensure they have more data on the world around them and help boost their common sense. ‘It’s not just about spoken language, it’s about contextual reasoning,’ he says. ‘That’s how I think of it: what other knowledge sources can you bring to make the reasoning more powerful?
‘If you think about our own intelligence, without eyes we can only do a limited set of things... But I wouldn’t speculate on whether Alexa needs arms and legs and eyes.’ The commitment of Amazon to Alexa is clear – and for good reason.
The gadgets not only generate oodles of valuable data about our lives and preferences, they also represent the company’s best chance to dominate the operating system for a new and emerging medium – voice – which is already transforming the way we interact with technology.
That helps explain why Amazon’s strategy is to sell its Alexa devices as a loss leader – offering them to consumers at a fat discount on the cost of the hardware. This ‘pile ‘em high, sell ‘em cheap’ approach is designed to ensure Alexa becomes as widespread as Google’s search engine – the go-to product for voice assistance delivered on the internet. The company has never revealed how many Alexa-enabled devices it has sold but in September the Seattle-based company doubled down on its strategy.
From earbuds to glasses
It unveiled a string of cheap new devices aimed at driving Alexa into a multitude of products that are used outside the home – from Echo ear buds to an Echo Loop smart ring and frames for prescription glasses called Echo Frame. But while Amazon can see a business case for all this, what about the obvious privacy concerns?
‘Customer trust is paramount to us,’ he insists, claiming that Amazon is working to ensure maximum transparency and the ability for consumers to delete their data whenever they want.
‘If you are uncomfortable, you say “Alexa delete what I said” or “Alexa delete everything I said today”. We also launched auto delete where you can set deletion on a periodic basis like after every three months or 18 months based on your choice.’
Whether that is enough to assuage growing concerns over privacy is unclear but Prasad admits there is a need for an ethical approach. ‘With any transformative technology you have to have the right mindset of responsible use, and also make sure it’s devoid of any bias,’ he says.
The Daily Telegraph