When one of the world’s largest AI companies rolls out dozens of new products, services, and features in the span of roughly an hour as Amazon did on Wednesday, what it all means can be tough to discern.
Of course Amazon, the inventor of smart speaker and smart display consumer electronic categories, had more speakers and screens to share, and the rollout of invite-only wearables like Echo Frames prescription eyeglasses and the Echo Loop smart ring delivered a bit of wow factor. Alexa is getting smarter in a dozen major ways, and the company continued its long-term, years-old strategy to make Alexa available everywhere, an effort that’s made an ecosystem that VP David Limp now refers to as the “Alexa economy.”
But the trend that stuck out to me more than any other among Amazon’s new wares are products and features intended to bring Amazon recommendation engines deeper into people’s lives.
Amazon’s Alexa has been slow to proactively speak up in the past, but that effort took some big steps forward this week.
Proactive recommendations from Alexa started last year with Alexa Hunches. At first, Hunches told you if you left a light on that you typically turn off at night, for example. In an update introduced this week, Hunches will begin to recommend the automation of certain tasks based on your habits.
For example, if you set an alarm or check the weather on a regular basis, Alexa might suggest you add it to a Routine. Or if you typically play music after you check the weather, Alexa might chime in and ask if you want to make a Routine or add to an existing one. Beginning with recommendations around alarms, weather, and music makes sense because each use case is among the things people do most often with AI assistants. Recommendations in the future could extend to music or streaming video content.
Routines are important because one of the major learnings from the early growth of the smart speaker market was that adoption rates are higher among people who incorporate features or voice apps into their daily lives.
Routine recommendations will be a challenge because they require that Alexa learns intent from each exchange and then stitches together your favorite things to do, when you want to do them.
“As an AI scientist, it’s one of the hardest problems to get right,” Alexa AI chief scientist Rohit Prasad told VentureBeat shortly after the news conference Wednesday. “This is where you’re taking the friction away and dynamically suggesting routines and completing through voice, but again, I want to set the expectations here: [It’s an] exciting feature, but this will be the one where we keep learning a lot.”
Recommendation engines further expanded this week with a growing understanding of the world outside your home. Amazon’s line of wearable devices must understand context about where you are and base its answers on location awareness, Prasad said.
“If you ask ‘What are the pool hours?’ it should know you’re in a hotel rather than in your local community, so this is where it’s now going into what I think of as on-the-go use cases. I’m very excited about it, because it’s pushing the boundaries of intelligence,” Prasad said. “Now you have to apply a lot of intelligence in order to say what’s the best decision to make, acoustics notwithstanding.”
For instance, Echo Buds, Amazon’s new Alexa-powered earbuds, will chirp in your ear about Whole Foods’ inventory while you’re browsing the store. If people actually begin to use this feature, proactive voice recommendations like sales or coupon offers might not be too far behind.
Amazon’s recommendation engines will also work to power a new feature for its Smart Home Skills API that predicts when consumable products like toothpaste or printer ink cartridges need to be replaced.
Those are predictable routines that can be monetized. Amazon is selling the idea to developers to entice them to use inventory sensors. When the inventory sensor detects that a consumable is low, the system can order a resupply — through Amazon, mostly likely. It’s these kind of steps that are giving Amazon the ability to increasingly talk about building not just an ecosystem around Alexa, but an economy.
A deeper relationship with Alexa
To entice people to invite Alexa into their lives, Amazon is pumping up its AI assistant’s personality.
In one of the first pieces of news introduced Wednesday, celebrity voices like Samuel L. Jackson will soon be able to read things like weather updates or tell you what’s on your calendar.
The addition of celebrity voices is an interesting evolution of text-to-speech synthesis being used to tie Alexa’s personality to those of icons and celebrities. Letting users, instead of tech companies, choose their virtual assistant’s personality could make for a powerful selling point, and it also gives users a way to associate their assistant with a celebrity or icon they like or admire. Imagine a constellation of hundreds of celebrities or well-known voices adapted to tell you the daily weather or be part of your morning routine.
Recommendations of any kind with Alexa will soon be helped along by automated frustration detection. Frustration detection, part of Amazon’s initiative to understand human emotion, uses two LSTM networks to recognize both tone of voice and the actual words used to make decisions based on whether a user seems satisfied with Alexa query results.
Alexa started out, as Limp said earlier this week, with the ability to respond to roughly a dozen commands to do things like set a timer or play music. But now Alexa wants to be your home security assistant, your pipeline to donate money to charitable organizations or political campaigns, and a helper to the visually impaired.
Amazon wants a relationship with you, and it wants it through Alexa. Celebrity voices deepen the personality; features like Alexa Guard for detection of human activity when you’re away from home deepen trust; and products like Echo Flex and low-cost speakers matched with far-field voice recognition make it more likely that people can speak with Alexa anywhere in their home.
Recommendations engines in the driver’s seat
If Alexa can make it through your front door, its various recommendation engines, using knowledge of your habits and your data alongside, can recommend that you buy certain products or make Alexa a bigger part of your daily life.
These recommendations could be enhanced in the future as Alexa better learns how to navigate the world, and how to glean data from home sensors with programs like Amazon Sidewalk.
Amazon CEO Jeff Bezos was excited enough about Echo Studio Wednesday that he made it a point to promote it in front of a gathering of journalists and analysts atop the rainforest biodome at the company’s headquarters. Echo owners, after all, are more likely to spend money in the Amazon marketplace, according to CIRP analysis, but the sale of smart speakers is only part of the formula.
But whether Alexa is on your face with Echo Frames, in your General Motors vehicle, or on a standard Echo speaker, the most powerful and transformative part of Alexa’s push into people’s lives is recommendations — the ability to predict your needs and remove all friction to fulfill them.
Thanks for reading,
Senior AI Staff Writer