A conversation with Adi Stein

Voice chatbots – which science fiction has presented to us in versions that are efficient (Starship Enterprise), bold (KITT from Knight Rider), captivating (Samantha from Her), and horrifying (HAL9000 from 2001: A Space Odyssey) – have emerged this decade as a consumer product, initially on cellphones (Siri and the like) and then in digital home assistants (Amazon Eco and its competitors). “The main differences between a textual chatbot and a voice assistant are that with a chatbot, the user can handle more visual information, the user experience is more familiar to people, and therefore it is more difficult to get confused and lost while using the product, whereas the voice experience allows for curtailing complex actions and enabling additional functionality when hands are not free,” says Adi Stein, technical projects manager with i.am+, which launched a smartwatch personal assistant platform, and is now offering a virtual personal assistant platform for tasks such as conference call summarizing and meeting scheduling, and chatbots for company and corporation service centers. “The voice interface allows one to learn in an experiential manner and make the best use of time (for example, while doing house chores or driving). One can ask the smart assistant to catch up on the news, learn new languages, ​​and listen to Ted lectures or podcasts for enrichment.”

“If we look at the timeline of human-computer interfaces, we started typing somewhere in the 1950s; in the 2000s we adopted touch when we moved to mobiles and tablets; and in recent years we’ve witnessed a change in interaction with the use of devices such as Google Home, Alexa, Siri and more – we’re in the midst of a transition from the era of contact to the era of sound, from phones to smart home appliances and voice-operated car entertainment systems,” says Stein, detailing the evolution of the human-machine interface. “The aim of the current technological developments is to allow us to interact naturally with the machines, an interaction that will allow the machines to understand how to communicate with us – not only laptops and phones, but also cars, refrigerators, lamps, TVs, etc. And that’s the goal of all the tech giants.”

For the full Magazine >> Click Here

This transition is fraught with technological challenges. “Understanding the user’s intent, and language analysis in particular, is a tough problem, and this is where artificial intelligence comes in. To decipher the speaker’s intent, we need to understand natural language in a noisy environment, to understand what the speaker said, the double meanings, a linguistic understanding of the sentence, a reference to the context in which it was spoken, and also to try to decipher what a person meant when he said what he said, because people don’t always say what they mean.” Stein gives a possible scenario: “It’s dusk and you’re in your house, in the kitchen, thinking about what to prepare for dinner. You’re sophisticated, so you have a number of personal assistants (in the kitchen, in the living room, and in a few other rooms). You are pondering what to cook while watching TV, which is in the adjacent living room, broadcasting a cooking show. The children are in their rooms, doing homework with loud rock music playing in the background, and just a minute ago your neighbor walked into the house to ask if she could borrow some eggs, and stayed to talk. Then you turn to your personal assistant, and in the midst of the cacophony of sounds and noises you say, ‘Alexa/OK Google, give me a recipe for steak.’ The personal assistants have to overcome all the noise, understand that they were spoken to, make out which of them will answer you according to your location, and direct the answer to the speaker. Assuming that the personal assistant was able to filter out the background noises and listen to you, it also has to deal with different accents to convert the voice into text. ”

As with search engines, it’s not enough to understand what users have said, but it also needs to understand what they really want. “The right answer to a user’s question is a challenge for voice application developers,” Stein admits. “For example, let’s take the weather question ‘Is it hot outside?’ The answer to such a question is yes or no, but the trick is to understand the real meaning behind the question. Usually when we ask if it’s hot or cold outside, we want to know what the temperature and chances of rain are, so we can dress appropriately. Or if you say, ‘Personal assistant, this sushi is not tasty,’ the personal assistant must try to understand why you told it that: Do you want recommendations for a nearby sushi place? Do you want to post a negative review on Yelp? Or are you at home and want it to order you new sushi? It should be able to answer you with the options that it deems most relevant.

Another challenge in having a conversation with a personal assistant is maintaining the conversation’s context. In a normative conversation between two people, questions or information items are usually exchanged within the same context. For a machine, it’s more difficult to follow the conversation’s subject when it is not explicitly written, and to understand when the subject has changed (for example: ‘Who is Barack Obama? Who’s his wife? What did she promote during her term of office?’).”

The speech interface makes voice chatbots suitable for teaching pre-reading-age children. Stein name few such apps: “The Sesame Street app, where kids can talk to their favorite character and learn about letters and play educational games; an app where the child can ask to hear the sounds his favorite animals make; a NASA app that you can ask about Mars; calculus learning apps; and apps that teach one a new word daily and tests the child’s use of the word in different sentences; etc. The innovative aspect of these apps lies in the user experience, which fascinates and intrigues children and allows them to consume more educational content.” Stein herself has created an educational voice chatbot in the framework of a personal assistance and artificial intelligence hackathon in July in Tel Aviv, organized by members of the Facebook community, “Personal Assistants – Alexa Google Home Echo Siri Cortana HomePod,” which won the people’s favorite award, a monetary prize, and flight tickets to a related Boston conference: “My team created an Alexa skill, a personal assistant for parents called Mary Poppins, designed to help parents get answers in real time to parenting questions from podcasts, TED lectures, and other resources. Parents need to ask, in natural language, a question about their kid’s development and the app’s logic knows to search for the relevant segment in a large database of podcasts and play it for the parent.”

For the full Magazine >> Click Here