Helping Healthcare with Voice Technology During COVID-19 with Aaron Bours, Head Of Marketing at Hyro
Your host Kerrie Roberts. Today my is Erin Boers. He's the head of marketing at high Welcome and thanks for being here that you so much. Carry happy to become a new election. Tel Aviv yes saw ULA. Were telling me you grew up in New York. You've lived in Israel for a while and also worked for Google before joining Jairo. Can you tell us a little bit more about yourself and your journey into technology in AI? Yet so I moved to Israel when I was nineteen years old to join the idea of Israel Defense Forces but before that I grew up in great neck long island I'd always been fascinated by technology is one of those middle schoolers who is always going to how stuff works. Dot Com to see how various tools functioned and mechanisms in that kind of thing so check spoke to me and marketing was also a natural transition from having loved to right from an early age and also due to some creative advertising alongside my father so advertising marketing and technical blended while for me and took me where I am today. been in Israel for about ten years now which is really convenient place to be since Tony is considered one of the premier startup. Hugs globally and done yet. Technology is just something that I doing now for five years working in high tech and it's an exciting space to be. Yes so you worked with Google Duplex for a while so yeah. Tell us about that and for those. That don't know Google duplex is. Can you talk about what it is and the work you did with that. Of course I distinction to be made I think between the Google assistant team or school duplex to you have Google's actual virtual assistant the Google assistant and group will duplexes a project at work in the background won't so consumers anybody who has an android or his downloaded the assistant APP could go in and requests a restaurant reservation and Duplex tragic with them. Be fired in the background would happen. Is You'd have a very human-like. Ai Powered call being placed out to a restaurant or a hairdresser among other types of businesses. To make an appointment or reservation. So that's kind of the difference between those two and they would work in tandem in order to achieve that user. Object Google duplexes brilliant brainchild from first name. Yaniv Le Eon Also Israeli and the project itself focused on new technologies for conducting natural language based conversations and again that's execute these real world tasks over the phone simple objectives scheduling certain types of appointments. That although aren't necessarily come in scope actually time consuming considered you know things that may be consumers. Don't need to do themselves. And the idea behind this technology specifically was to allow people to speak normally like they would to another person without having to adapt to some kind of rigid flow from something like you know traditional VR. Some other type of machine and the product actually begins so convincingly human like that it fell below one percent detection rate which is essentially almost like passing the Turing Test. Right so people really cannot fathom that they were speaking to the artificial entity became almost unbelievable that came to a head. The major moment for duplex was actually when Sundar the CEO we will stood on stage at the IO in two thousand eighteen and he exposed these discrepancies. That made Google duplex standout for what it was in some examples of those. I think maybe a couple of listeners to the podcast are probably familiar with this just because it was big in the voice. Say I stays but a couple of those like Gotcha okay cool and those kinds of nuances in the English. Language are really tough to teach. But when you have them. It turns off people's defenses concerns of being able to assume. Okay I'm talking now to something. That is not my species like this something that Sony artifice show so that was a very special moment and presentation. Obviously a big win for the duplex project went viral. A lot of technology is based on strong teach. Es Taxi Speech Technologies synthesised. I think the engine was run through Taco Tron and wave net and those were used to create these new recordings automatically for parameters the date and time and also make sure the timing the lag all these things that are intricate in human language were framed correctly. Says the conversation really flowed properly? You felt like you were talking to a human being and then I think one of the mentioned about Duplex Kind of move on but this is important into the segue now what we do it at high rope. One of these key decisions for duplex is to contain it to close domains so domains. That are narrow enough that you can you know master than quickly but also had enough attributes are nuances. That you can explore them extensively for duplex typically. This was restaurant reservations. That's where it started and that meant that it couldn't carry general conversations so what I said before about passing the turing test. That's in one domain. But that doesn't mean if you start asking I grew will duplex. Hey isn't going to be sunny tomorrow that it will all of a sudden be able to answer you with the same efficiency that it could if you're talking about things that were related to the restaurant industry so that's an important distinction. It's specific to its use case and that's a fundamental difference from what we're trying to do now at higher. I was just talking to someone else for this show about the importance of what you're talking about the nuances using certain slang using timing so that it feels more natural and I think a lot of companies kind of pushed that aside or feel like it's not as important or maybe they don't know how to tap into that and so I think when you're talking about with Google Duplex and the difference. It makes that little bit. Make somebody feel more connected. The I WANNA use. It is a really important piece so I thank you for sharing that in bringing that up and now you said you joined Tiro last year and I love for you to tell us what the company is about and why interested you to join them especially with your background with Google Duplex as well. Yeah absolutely so Hirose at its core and natural language and conversationally. I company it was founded by Israel crush Ron Cohen out of Cornell. Check they were doing your masters together and Who is the? Cto CAME FROM DUPLEX. Has a crazy resume in terms of voice technologies and linguistics? So the three of them came together and started this company. What we do re power virtual assistance through natural language. And there's an emphasis there in glisters to understand some belief is you can understand you know basic structure. You can understand a paragraph you can understand the entire peace. Our focus is more on understanding as opposed to something. Like you know other projects where it's supposed to be human centric and trying to make personal connections. We understand that we are agent and the idea is to get users the help they need as quickly as possible beyond that. I think what makes us different and why I'm very high on the company and the trajectory is the capabilities that we have in terms of. Really SEAMLESS LIGHTNING FAST DEPLOYMENT. So not sound too much of a marketing guy here. We are extremely quick in terms of how we can automatically understand information by scraping it and it's a very sources so like you can give us a database. We can scrape your information. Which might be disorganized on a bunch of different websites? You can give us. Api's you can give us all of those things and we can scrape all that information and add natural language understanding layers. You Map that to our knowledge breath which is essentially a bunch of nodes that understand different parameters and are able to context and then we deploy on a website or call center or an APP or really. It's NAMI channel solution. So the idea is with one line of code you can embed conversational interface extremely quickly scale that as well as you wanna add more channels or different use cases entirely so while Google duplex for instance when domain by domain. We're trying to actually focus understanding of language at its core and then extracting from that the different types of use cases as opposed to vice versa. Doing use case understanding the language. And now we're using those same types of technologies and we we were able to apply this to helping with cove in nineteen so quickly in the last couple of months higher developed a hybrid voice and text virtual assistant for free to help healthcare organizations deal with Kobe nineteen. Can you talk a little bit about how and why this started and how it's helping right now? Yes it's my favorite topic since it's literally happening in real time. We're really proud. That makes going so well actually. So it started in February when one of our senior engineers guy laying Idea to offer a chat. Bot FOR HOURS. And he's kind of obsessed with the Crowbar. He's made his own graphs for the last couple of months. She's like a covert nineteen. No Sir dommage predicting the different trajectories of this and that anyway at that point wasn't yet full in the US. And so we kind of shelby idea. We predominantly worked with. Us healthcare organizations and their patients weren't demanding information but then once the explosive numbers started to hear out of Washington. We jumped into action. When I WANNA clarify people say it's important nowadays to kit allow for us. This wasn't a pivot. We were already in health care. This is more of a heart reflex. And the reflex naturally represented our company values to pave forward and kind of create this good products for the greater good kind of thing and took about twenty four hours to craft our messaging around free covert nineteen conversation assistant for healthcare. We revamped the website and immediately started emailing all our friends clients partners in healthcare in within. Maybe forty eight to seventy two hours. We had our first deployment is almost instant at least in the world of beat Abi Sach. This is almost like an instant reaction from us which we're all very proud of but we were able to do this so quickly and the idea to give it for free was almost a no brainer. I don't think anyone in the company discussed ice models. So what was the motivation through? January developed to help these overwhelming healthcare organizations currently dealing with unprecedented spike in traffic across their communication channels like for instance websites or call centers. In addition we noticed slew of Misinformation. That was coming from various sources across the web for social media but even chat bots just the wrong facts rumors purposefulness information so we.