Amazon sends Alexa developers in search of "Holy Grail of the science of voice"



[ad_1]

At the Amazon: re: March conference last week, the company unveiled Alexa Conversations in preview. Conversations is a module of the Alexa Skills Kit that combines Alexa voice applications with experiences that help you accomplish complex tasks.

Alexa Conversations is perhaps the most intriguing and substantial speech of Amazon in years. Conversations will create skills with fewer lines of code. This will also eliminate the need to understand the many different ways in which a person can request the execution of an action, since a recurrent neural network will automatically generate a flow of dialogue.

For users, Alexa Conversations will facilitate the completion of tasks requiring the incorporation of multiple skills and will reduce the number of interactions needed to perform tasks such as booking a movie ticket or ordering food.

Amazon VP David Limp sees conversations as a big step forward. "It's kind of the holy grail of the science of voice: how to do a chain conversation when you do not think about it in a programmed way, end to end. […] A year or two ago, I would have said that we had not found a way out of this tunnel, but now, I think science shows us that [although] it will take us years to talk more and more, […] this breakthrough is very big for us, the tip of the iceberg, "said Limp.

It starts with an evening and an informal conversation

The Alexa Conversations journey begins for the first time with a night scenario. At a demonstration on stage last week at the address Re: March, a woman buys a movie ticket, makes dinner reservations and launches a walk in about a minute. (The Atom, Uber and OpenTable tickets are Alexa Conversations' first partners.)

The scenario of the evening is the first of what, according to Amazon, will become a collection of grouped experiences to get things done.

Conversations may one day fuel more difficult tasks such as the weekend scenario that Limp demonstrated last fall at an event intended to present nearly a dozen new devices powered by Alexa. Limp's speech on the Holy Grail is a transformation that all the major technology companies of the world with an artificial intelligence assistant are trying to evolve the voice interface assistants completing the basic tasks to an assistant who can manage complex and complicated tasks.

Two years ago, at a rare meeting on the scene bringing together current or former leaders of Alexa, Google Assistant, Siri and Cortana, the co-founder of Vivana and the co-creator of Siri, Adam Cheyer , a person who was questioning about the future of voice assistants since the 1990s, was wondering aloud about an assistant who can guide you through the scenario of marriage planning for your sister. (Samsung acquired Viv in October 2016 to enhance its AI Bixby assistant.)

At this event, Cheyer explained how the voice would define the next computer decade and that it was important to bring the services of the main AI assistants together with an ecosystem of third-party voice applications. "I do not want to have to remember what a car assistant, the TV system, the Alexa against Cortana can do against … too much. I hope that an assistant from each device can access each service without distinction between what is essential and what is third party, "said Cheyer.

Amazon is working in this direction, starting by reducing the number of interactions needed to get things done with Alexa. Last fall, Amazon introduced Tracking mode, allowing you multiple interactions without having to say the word "Alexa" once. With Conversations, the number of interactions needed to run the night scenario is reduced from 40 to around a dozen round – trip interactions.

To reinforce the perception that Alexa is capable of natural conversation, the artificial intelligence wizard has learned to whisper when a person whispers and can now answer a nameless skill summon. This means that you can say "ride me by car" instead of having to run the skill first by saying "Alexa, run the Uber skill".

Create the perception of intelligence

Amazon is not the only one to make an assistant a fluid conversation like the one you expect from another person. Google has launched Continued Conversations so you can not keep talking about something. Alexa Conversations also gives Amazon's artificial intelligence assistant the power to do business quickly or to trade, such as the new power of Google Assistant for Food Ordering and Google Duplex. Duplex for the Web and deep connections between Android apps and Google Assistant started last month. Microsoft also brings a similar intelligence to assistants at the workplace with Semantic Machines, a start-up acquired in 2018.

This highlights the fact that more complex tasks require more than a simple exchange, said Alexa Alexa product manager, Sanju Pancholi. "When you start solving more complex problems, there is more information exchange, there are more decisions every moment, and therefore many actions can occur in the context of Same conversation with different individuals, "he said.

He led a session on re: March to introduce Alexa Conversations to companies and developers, and spoke of an assistant capable of "solving their product and service needs as soon as they realize that They need it ".

To be considered intelligent, Amazon believes that an assistant needs to understand natural language, remember the context, and make proactive forecasting suggestions, characteristics that can prove that an assistant is smart enough to perform more complex tasks. Eliminating the need to repeat oneself is also essential.

"If you do [customers] Repeat the information again and again and again, you force them to believe that they are talking to a silly entity, and if that is the relationship you are building with them from the outset, chances are that They never delegate Higher-level tasks are your responsibility, because they will never think that you are able to solve higher-order problems for them, "he said.

The Alexa Skills store now has more than 90,000 skills and 325,000 developers have used the Alexa Skills kit, said Pancholi. Alexa is now available in 100 million devices.

Pancholi told developers that potential next steps for Alexa Conversations scenarios could include skill sets to help people watch content at home, get food delivered, or buy a gift.

Skills competencies

In an interview with VentureBeat, Alexa's chief scientist, Rohit Prasad, declined to give details of the use cases that could be addressed, but believes that this could include ways to help plan a weekend. end. Prasad, who led Alexa AI's initiatives for language comprehension and emotional intelligence, said Conversations was designed to link the voice ecosystem to increase the engagement of skills and that of Alexa. .

"The developers' proposal is that you start having more traffic and more discoveries as we develop cross-skills, such as the fact that the nighttime experience now requires you to order a taxi. So, Uber and Lyft will see more traffic and more commitment from customers. Thus, and in addition, the discovery of skills will naturally occur as part of that. It is therefore an important element of our value proposition in this case.

Even Blueprints – voice application templates for custom and private Echo skills – may soon incorporate Conversations, Prasad said. Customized skill sets for the home could, for example, guide children through multistep routines, do housework, and help count down important dates.

The first proactive features of Alexa – Hunches, which suggests reminders of events and actions of smart home, and Alexa Guard for the detection of broken glass noise or a smoke detector – have were launched last fall.

Conversations could also become part of Amazon's voice assistant if the module was integrated with Alexa for Business, which supported Blueprints in March.

Brands, independent developers and

In January 2018, CNBC announced that Amazon was in talks with brands such as Procter & Gamble and Clorox to enter into contracts to promote their products to Alexa users.

Steve Rabuchin, vice president of Amazon Alexa, insists there is no way for companies or developers to get priority from Alexa's voice app recommendation system, but the ecosystem Alexa's voice apps could face another problem. Because of the way voice applications often operate without a display, packaging skills mean that some skills may inevitably be omitted or not classified.

This is particularly important for voice applications. Unlike looking for apps on a smartphone, Alexa offers a voice application recommendation engine that only serves three skills at a time.

"Our vision is not to end up where it's just the biggest brands or the most popular," said Rabuchin in an interview with VentureBeat. "Our most popular skills are mostly independent developers, individual developers."

The Amazon skill recommendation engine responds when you say things like "Alexa, bring me a ride," recommend voice applications based on metrics such as engagement levels, for which Amazon has started paying developers in 2017.

Conversations will incorporate quality measures of skills such as user ratings, engagement levels, factors such as regional significance, whether a skill works on a smart display and personal information may also decide skills to display during Alexa Conversations interactions.

"I think we have a good reading book to start on, I do not think it's a perfect reading book, but it's a great book to start with," Prasad said.

[ad_2]

Source link