Peter O’Neill recently interviewed Todd Mozer, CEO of Sensory. The conversation begins with some highlights of Sensory’s exceptional year before diving into the details of a recent major milestone for the company: the integration of its TrulySecure biometric software into LG’s latest flagship smartphones. LG’s move came in anticipation of Apple’s recent Face ID announcement, and Mozer speaks about taking a leadership role in an era when Apple can no longer keep its secrets. The conversation goes on to talk about speech, voice, and face biometrics in the world of IoT and smart homes, how decentralized biometrics can prevent major data breaches like the recent Equifax fiasco, and Sensory’s impressive virtual banking assistant technology.
Read our full interview with Todd Mozer, CEO, Sensory:
Peter O’Neill, President, FindBiometrics (FB): Has this been a good year for your company? What have been some of the highlights?
Todd Mozer, CEO, Sensory: It’s been a great year for Sensory. We have a couple of product lines that are really doing well. Our fiscal year just ended in September, and we grew over the previous year by more than 50 percent, and we are expecting continued growth in that range for 2018.
Things are going well with TrulyHandsFree, our low-power speech recognition engine. The IoT market segment has really created a lot of demand for voice controlled products, and we are getting a lot of traction from companies using cloud based digital assistants that want embedded wake up words and commands. On the biometric side of things, we are seeing a lot of opportunity and growth in both the face and the voice side of our TrulySecure product line. Obviously, mobile phones are a big area, and vertical markets like banking are taking off for us, so overall things are doing quite well.
FB: You just recently announced that LG’s latest phones will feature your Truly Secure facial recognition. Can you tell us a little bit about that news?
Sensory: The whole mobile phone industry – Led by LG, Apple, and Samsung are moving to face authentication as part of their biometric stack. Apple’s is moving towards removing the fingerprint and Samsung seems to be doing a layered approach with a lot of biometric technologies. Most of the mobile players are following this trend. LG put Sensory’s TrulySecure in the G6, Q8, and V30 lines. I just bought a V30 and it’s a fantastic phone!
FB: Now your comment about Apple, we are in total agreement with you. When that news came out we felt that it would have the same impact as Touch ID several years back had on our industry, and I guess you are feeling that almost immediate impact it would seem?
Sensory: Well, everybody knew it was coming. I think one of the big things about Apple’s latest announcements were that there were no big surprises. They were so historically famous for keeping things secret and then surprising us with one more thing and this time there were no surprises. The whole mobile industry knew that face was coming, and LG took a leadership role in setting the trend in this face authentication space.
But even on the voice biometric side we have released our higher end voice biometrics that we call TrulySecure Speaker Verification in both the Moto X and Moto Z, and it’s been used by Samsung, Huawei, and now LG as well, so we are getting a lot of traction for both voice and face across the mobile phone segments.
FB: You started to mention IoT and voice. It’s an area that we are very interested in – home automation etc. – and the two biometrics that you focus on, face and voice, are really considered explosive in terms of their growth right now. One, because of Apple, but on the voice side, what is driving that? It seems to be going through quite a growth period, what are the drivers there?
Sensory: Amazon’s Alexa did for the IoT space what Siri did for voice on the mobile phones and everyone and their brother are now shipping a speaker that is voice controlled. In fact, just today Sonos announced their new speaker which I’m kind of excited about just because Sonos has really the highest quality stuff out there and I’m a big fan of theirs for audio quality and connectivity. The Alexa phenomenon is driving a lot of adoption across a wide range of home consumer type of products. A lot of these IoT devices use TrulyHandsfree which gives embedded voice wake-up. And as the products get more advanced they are using TrulyHandsfree for embedded commands (fast response and wifi not required) and many are layering TrulySecure speaker verification on top of our wake-up word technology so that when you talk to the product it knows who you are. They’re a little slower in adopting that than just the wake word, but I think over the next year what you are going to see is more and more vendors having wake words that do entail speaker verification. Amazon just announced it in their echo line. In home robotics JIBO, the home personal robot, recognizes your voice when you say, “Hey JIBO”. Jibo identifies who you are because it has Sensory’s speaker verification built into the wake-up.
FB: Now you also recently launched your bank teller app and I have had a chance to demo it – it’s so cool with the avatar – and there is a lot of cutting edge technology involved in this product, can you please describe it to us?
Sensory: Sure, let me first give a little background on some of the technologies involved, because you are absolutely right: we have really combined a whole lot of AI technologies and I suspect it is the most advanced demonstration of AI that’s ever been done at this point, and I mean from anybody. We’ve combined not only text-to-speech and wake-up word speech recognition but we put a natural language speech recognition engine into it too and we have biometrics in it which combines face, or voice, or face and voice, so we can identify people. On top of all that we have an animated talking avatar so your assistant has a face; the lip synchronization is driven by the TTS engine, which uses a proprietary non linear morphing scheme to change the mouth shape and image to match distinct visemes that accurately models the way we speak.
For the Finovate conference, it’s one of those demo-based shows where you get up on stage for seven minutes, we put together a bank teller app so it’s like having a bank teller on your phone that you talk to and it talks back to you. So, you are basically interfacing with what looks like a person, and you talk to it and it talks back. We’ve designed the natural language interface to be specific to the domain of banking so I can say things like, “Transfer $134.17 from my checking to my savings account” or “Pay my phone bill from checking” or “What’s my savings account balance?” or “Where do I go to buy checks?”, those kinds of things. I can send money to other people that are in my contacts list. It is a very smart and intelligent interface where if I leave things out it can do what is known as ‘form filling’ to prompt me for the things I left out. For example, if I say, “Transfer $100 to Peter O’Neill” it will say, “From which account would you like to transfer”.
FB: To me this is a major step forward. When I saw the demo, I was very impressed, I think mostly with the user interface and the comfort that is created with this avatar and lip synching technology; it is so natural. I think people will feel at ease using it. Are you finding that?
Sensory: The lip synch and the avatar is something that we had developed a long time ago and we just put on the back burner because we didn’t see a good business model. I do think with the rise of personal assistants we are going to see more use of avatars as personal assistants, but it hasn’t happened yet. From the demo that we gave there were a few banks that approached us that were interested, but it’s really not a big focus of ours today. We do think we want to get a few customers out there with this sort of technology to highlight that it is feasible but is not an area that we are putting a lot of development on.
FB: You mentioned one of the big news events that occurred last month and that was the Apple launch. The other was the Equifax breach, which really shook up our industry. What is your view on that and how can biometric technology help solve these breach issues?
Sensory: Well, it just highlights the sort of the philosophy that Sensory went into biometrics with. And, in fact, it’s not just biometrics but it’s what we do with our speech recognition too. We do it all on-device so it never gets sent out to a centralized database or cloud where a hacker can steal information from everybody at once. In fact, with the Yahoo breach it was just announced that they had totally underplayed the quantity; it turned out that all of Yahoo’s contact information was stolen, so it just wasn’t some percentage of it. If there’s a bunch of data sitting in the cloud, then with more information and the more personal it is, the more incentive hackers have to try and break in and steal it. If there’s enough data the value gets so high that professionals and governments can try and take it. So, keeping more on-device, in your home or in your pocket, it is just a safer approach, and that is the tactic Sensory has taken with both speech recognition and biometrics.
FB: In several of our recent webinars and panels it appears that end user education is becoming critically important in our industry as the it starts to move faster and faster. We are trying very hard to make sure that biometric experts such as yourself are speaking about some of the challenges in the marketplace. How critical do you think this education aspect is, and what are some of the challenges that you see as biometrics start to move ever more quickly into the marketplace?
Sensory: When you think about the taking of personal information there are two areas that you can think about. One is the legal theft by companies that have you sign an agreement when you click on a button and say yes take all my personal information and you can use it in any way you want. In this area the education is about how well the consumer understands the full consequences of what they are signing off on. The second area is the illegal theft, the Yahoo and Equifax kind of breaches. I think they both require a different sort of education. I think it is very hard to quantify how safe different websites or different places are that we trust our personal information in. I saw a really interesting documentary a while back, I think it was Terms and Conditions May Apply, and it basically talked about the ways that we sign off on giving away all our data in return for the free software. They make it so you don’t want to read the agreement. They put it in bold case, six-point font – they make it very hard to read so people don’t read the EULA’s. To the extent that there was legislation that would force them to quantify it in a few simple bullet points like, we are going to use your data for X, Y, and Z and not use for A, B, and C. I think that would be really valuable for consumers.
FB: I’m aware of that push and it is a long road to get to that, but is it needed because nobody reads that 17 pages of fine type that you talked about. And really it can probably be boiled down to the four or five points that are critical for the end user, you and I, when we sign up for anything. Privacy advocates are definitely heading in that direction and I hope it happens soon.
Sensory: Yes, it is a tough and challenging battle. And you know we get free stuff, right? Software has become essentially free by giving away our personal information.
FB: Well Todd it’s been an incredible year for your company, what’s next for Sensory?
Sensory: We’re just trying to keep up with our current customers. We’ve really gone from having a small base of really big customers to a much broader base of customers. So, we need to keep growing our support staff, we need to keep growing our technology staff to be able to deal with the increasing demand that we are seeing. But we are moving in some exciting new directions: we’re going to move toward 3D cameras with the face authentication – we think Apple’s move in that direction was the right one. We’re working on sound and scene analysis because we’re in so many always-on products we want to be able to identify the different sounds whether it’s the glass break, or the baby crying, or an alarm going off – things that can help the owners of the homes where these devices are deployed.
FB: Thank you Todd for taking the time to speak with us today. Your products are fantastic and I think you are really in the sweet spot of where the industry is heading, so congratulations on that.
Sensory: Thank you Peter, I appreciate it.