Conversation is the ultimate user interface

We may be living in the golden age of information, but finding the right information is always a challenge. To meet this challenge, my team and I at Amazon Alexa are building what we believe is the next-generation user interface that will redefine the way we interact with technology and find information.

We spend hours every day hunched over phones and laptops. We open, close and reopen applications. We parade. We type on tiny QWERTY keyboards. And we click on an endless sea of ​​blue links every time we search the web. The internet is indeed amazing. The user interface is not.

We agreed to these terms because, since the dawn of the digital age, this is all we know. But these modes of interaction with the digital universe have been developed in the service of economic models, and not of the user experience. They are designed to increase the time you spend online, generate clicks, and maximize engagement time. But it’s unfair to force humans to find information this way. And it’s time to move on.

Conversation: the secular interface

The first step is to change the way we interact with the Internet. And thankfully, recent advancements in AI make a whole new user interface possible. In fact, it’s the original interface, the one we’ve been using for almost two million years. This is called “conversation”.

No speech, watch out. We’ve been using it for nearly a decade already, interacting with our phones and digital assistants like Alexa. I’m talking about a real human conversation. The kind you might have with a friend over a beer, in which vague or poorly worded questions are understood. Conversations in which intent is inferred and answers to questions are summarized and personalized.

When two people converse, they understand each other’s context and incorporate visual cues. Conversations can be concise and effective. Or they may cover a variety of topics, change direction, and lead to a chance find. Humans do this without even thinking about it. But teaching a machine to do this requires significant advances in the science of AI. It’s not just about natural language processing (NLP) capabilities, which rapidly improve with every voice interaction (Alexa alone receives more than a billion requests every week from hundreds of millions of devices in more than 17 languages.)

AI in milliseconds

On the contrary, for a machine to learn the give-and-take nature of conversation requires fundamental rethinking of our current system of information retrieval, including the ability to crawl billions of web pages in real time (retrieval of web-scale neural information), concisely summarizing information from the enormity of the web (automated summarization), and the ability to recognize an end-user’s intent and recommend additional relevant content (at the using contextual recommendation engines).

Conversational interfaces require these systems (and many others) to work together seamlessly and instantly. For example, if you ask an AI assistant, “Where is the oldest living tree in the world?” he should be able to not only answer this question quickly and concisely, but also understand that you are currently only an hour’s drive from said tree, and follow the instructions and trail recommendations from hiking in the region.

Or if you watch the Dallas Cowboys on Thursday Night Football and vaguely ask, “Who just caught that pass?” he should be able to deduce which game you are watching, which team is on offense, who caught the pass and for how many yards. All in milliseconds.

These are difficult and unprecedented problems. As such, Amazon has assembled a team of world-class AI scientists dedicated to solving them. We invest in these resources because we believe these capabilities represent the future of human-computer interaction. And we are not the only ones.

“These give-and-take interactions build relationships that will shape both the user and the system,” said Hae Won Park, a researcher with MIT’s Personal Robots Group. “Relationship agents can disrupt areas such as personal assistance, health, aging, education, etc. We are only just beginning to realize the benefits for users.

Towards “ambient intelligence”

Indeed, conversational AI can benefit any business that wants to change the way its customers or employees interact with digital information. And like so many AI advancements first developed to serve Alexa – like Amazon Lex and Amazon Polly – we hope to make these capabilities fully available to any business, in any industry, through the AI services available on AWS.

The end goal is to shift the burden of retrieving and distilling relevant information from humans to AIs. And by integrating this conversational capability into the spaces where we live and work – our kitchens, cars and offices – we can reduce the time we spend scrutinizing phones and laptops. We call this concept “ambient intelligence”, in which AI is available everywhere around you, assisting you when you need it and even anticipating your needs, but backing away when you don’t.

In other words, we can still take advantage of all the power of the Internet while spending much less time on it. As for business models that depend on small screens, endless scrolling and a sea of ​​blue links? It’s up to them to adapt to us, not the other way around.

Vishal Sharma is Vice President of Amazon Alexa AI Information.