DayCast discussion with Brian Garr on the state of AI in customer service. Brian discusses how Virgin Voyages is leveraging AI to provide exceptional service to its customers/sailors. Brian also talks about his time with IBM in voice recognition and how that technology helped transform the way businesses operate.
Announcer: Welcome to DayCast, where we discuss how AI is really used in business, no hype, discussions with real AI users. Here's your host Corey Dayhuff.
Corey: I’d like to welcome to DayCast Brian Garr. Brian is currently the AI Business Solutions Innovator at the Dayhuff Group. He has a twenty-five-year history in technology, including twenty years in AI. Brian was part of a team of IBM that first created the WebSphere voice recognition software, which I personally ran on my computer back in the day. He then ran a couple of other technology companies and most recently was a Senior Creative Technologist for Virgin Voyages. His machine translation team won a Heroes in Technology Award from the Smithsonian Institute in the late 90s. Welcome Brian.
Brian: Thanks Corey. It's awesome to be here.
Corey: So, you had kind of an interesting education background prior to getting into IT and a short career before that. Can you tell the audience here what you did before IT?
Brian: Sure, I took the traditional route. I got a degree in theater at college and tried my hand at acting and even did some sketch comedy in Georgetown for a while. And I’ve discovered two things: One, I really enjoy theater, and two I wasn't that good at it. So, I ended up taking a job as a property manager to manage a 930-unit rehab of apartments, and I said I told them, this was in the mid-80s, I knew the computer and I was going to do that. And so, an Apple 2E was basically all there was in the market, and I bought it and took it home and learned how to write database programs over the weekend so that I could manage 930 units. It wasn’t long after that that the head of technology for company brought me in to start writing software on their mainframe, the early days with Cobol and Databus. Then from there I was lucky enough to get a job at Globalink which was in the machine translation space. And from there it just sort of took off. I spent five years there, then we sold that company, and then IBM came along asked me to help with their speech recognition business and from there on to LinguaSys with Mark Cuban as my chief investor and Virgin Voyages and you.
Corey: So, how do you think an acting degree helps you out in technology today?
Brian: You know I'm so happy I got that degree for lots of reasons, but the main one is that I have been giving speeches all over the world, from China to Madrid to Disneyworld, and Paris and all over, and I get invited partially because I hope and what I hear is I'm entertaining speaker, because when I go in front of a large audience, I do think of it as entertaining as opposed to reading slides. So, I am very animated when I give a lecture or speech, and I also think it helps just with personal skills on being able to meet people and talk to people and entertain them at the same time that I might be trying to educate them or at the same time that I might be getting buy in from them for a new solution I’ve come up with.
Corey: So, I wonder if AI in particular is more suited to a different background than traditional IT, because it is so interactive. You're actually training an engine to try to act more human and be more human. In your perspective, on things we come from an angle that is more about communication, more about interpersonal skills than it is about numbers.
Brian: Very true. Also, I look at AI as an area that has no, thinking about it as a car metaphor, it has no road to head of it to drive down. You basically have to create the road as you drive somewhere, and that requires more of a creative mind that a strict rule-based mind. My theater background, I think you know, my background in general, has always been around the creative process, more so than the dogmatic part of logic, not that you don't need the other part, but basically, you're, saying what's some new ways that I can use this technology. As an example, when Netscape came out in mid 90s and all of a sudden people could search the web, I was CTO for Global Inc, and we were in the machine translation business and the first thing I thought of was wouldn’t it be great if people could surf websites in other languages. So, we actually produced the very first website translator in 1996 I think that was, and released it on the retail market. So, it's that kind of creative leap that's required not with machine translation so much but with artificial intelligence that really kicks it to the next level.
Corey: And how do you think you view things differently than other traditional IT folks do when going out and either talking to the business owners, the systems owners within the business or even going out and talking to potential customers, do you find yourself kind of bringing a different perspective and talking differently than traditional IT individuals would?
Brian: I still have the same basic problems that most IT people have when they're talking to people in that they use to many acronyms and use some of the lingo that was created in the IT world to basically scare off everybody that’s not in IT, to make them think that they're more valuable and smarter than everybody else and that sort of comes with the territory, not that something that I'm proud of, that I was working on that
Corey: You're giving away our secrets Brian. What are you doing?
Brian: It's true and that's one of the things I love to do is show how magic is done. I actually was an amateur magician in my teens, so that I do like to what they say straight talk, and I've written a book about it actually called Introduction to AI which expels a lot of myths about AI, how difficult it is, and it's available on Amazon, it’s a nice little plug there.
Brian: So back to your question I like to educate people without them knowing they're being educated. It’s not an ego thing for me to stand in front of them and say I'm going to teach you something. I'd rather have a conversation and make it go down a path that helps someone understand why something will benefit them and how it works and why it works and why it makes sense to go down that path.
Corey: Absolutely so tell me a little bit about your time and virgin voyages which had to be fascinating. It's unfortunate they have never actually sailed, having really created their first ship right before covid hit, but anyway you get into some fascinating AI technology, while working with them.
Brian: That was just the most enjoyable job I’ve ever had and so unlike any other job I've ever had, because I was living in Italy, where they built the boats, where the built the ships. I'm sorry, you can have boats on a ship but you can't have ships on a boat and if I call it a boat, I get docked. So, it was fascinating living there and it was fascinating building the ship. And, yes, we got into so much out there was creative technology, so I was supposed to think of cool stuff that the sailors, we call the people that were going to be on the boat eventually sailors, and our job was to make technology seamless. So, for example, and some of this was not AI, brute force technology, but we had sensors everywhere and you wore a wrist band, and you would walk down, and as soon as you get close to your door, the door unlocked and you walk in and when it's your first entry, the CD comes on and play something that’s uniquely you, explaining how the ship works and the lights go on and the curtains open up on your patio door. And we also did some things that were unique and not finished yet, because they’re proof of concepts, but one of them was taking the live CCTV feed and using voice analytics to determine people's moods in the restaurant and the idea behind it, and we haven’t proven that it's working yet because we haven't had any people in the restaurants. But the idea was to prove that we could tell when a table was upset. We didn't know if it was because of service or bad food or whatever, but still the maître d’ would get a notice to go over there and offer them of complimentary bottle champagne or whatever, and too to make sure that they were enjoying their time in the restaurant. So that we could very easily extend to other areas of yet obviously our goal in life on the ship is to make sure passengers are always happy and always having fun. So those kinds of techniques help us to do that. Another area that I spent some time on was using voice analytics to not only determine mood but beyond that. We just asked some questions, and this is classic AI. We ask questions like can I determine the gender of a person using their voice, can I determine whether they’re sick or not? There are some diseases we now know have particular characteristics in a voice utterance that would help us detect if they have certain diseases. So, it goes beyond there. We just kept adding onto that saying what other interesting characteristics can we find to try to identify in the voice. And of course, this is highly dependent on finding enough data enough people that have that particular characteristic you’re looking for and you have voice recordings of them to even test your theory to see if it’s possible.
Corey: So, was there a concerted effort from a business perspective to try to leverage as much AI technology as you possibly could? If your goal is customer satisfaction and customer experience was the way to implement that trying to find as much AI technology to implement as you could to do that?
Brian: So, to be honest this was the first ship for Virgin. The emphasis was more on making sure we had the basics right, that we had a blocking and tackling down, that we knew how to check people in and we knew how to assign them rooms and take reservations to restaurants. So, a lot of the creative stuff I did was behind the scenes, and if something came of it then I would bring it forward. So, I was given a lot of leeway as far as budget and ability to go do cool things, but there was always the thought that these would probably go into a future ship, not in the current ship, because the current ship was all about blocking and tackling.
Corey: Crawl before you walk right?
Corey: You know, that's probably true, from an AI perspective on most organizations and on many of my podcast with CIOs, that's really where they're at from an AI perspective. They are at the crawl phase of AI. I think I see that across majority of your organizations that we deal with is they’re really trying to figure out how do I start implementing.
Brian: Right. I did a really cool AI project for probably the largest distributor of toothpastes and mouthwash and stuff like that around the world, and the idea was they were a Tableau shop and they wanted their executives to be able to go into the conference room and talk to Alexa on the desk and say Alexa show me the report that shows we have in the and then they wouldn't say whatever they want to know that is important for them to visualize, and we wrote an app that would take that utterance and send it to at time cognitive codes AI engine, and we would then turn that intent into a SQL statement and then send it to the APIs to Tableau and bring back a pretty map or picture that Tableau created, a finalized report right on the screen in the conference room. So, it worked and we got pretty far with it and then, as you were saying, before, blocking and tackling we, the POC came to an end because they were pulled off of that and needed to concentrate more on a new analytics project.
Corey: So, what have you seen is kind of some of the most successful use cases as you worked across your career in AI, what has really flushed itself out?
Brian: Well, surprisingly, it's the easy chatbot type of events that usually create the most usable. Sure, there are robots now that can take a drink order. We actually had a drink order machine on the ship, on the crystal sorry on the Scarlet Lady, and so that kind of event is cool and everything but chatbots that can offload sixty percent of the efforts of a call center so that they can concentrate on higher revenue products and work towards improving customer sat and revenue comes out much better than the fancy robots and the cool things that are usually tied together with a little bit of duct tape and sometimes don't work that well.
Corey: Right. The old adage, keep it simple comes to mind. Right?
Corey: Seems to be the case for all new technologies to come to market. The promise never seems to be realized in totality, but the core of the value almost always finds its way to the business right.
Brian: And a good parallel to that is speech recognition which everybody today takes for granted. In the early, late 90s, I’m sorry the late 90s IBM had Via Voice out there and then they created Voice Server, which is the enterprise product, and we were never quite good enough, and there are a lot of companies that tried us and we poured our heart and soul into our efforts, and usually there are pilots, and usually the pilots didn’t quite get off the ground and because the technology was in its infancy, it took a lot of effort to get to the point where now everybody just assumes that Alexa’s gonna understand them when they say turn off the light, and it’s gonna understand them when they give all sort of commands to it and they ask for sports scores or whatever it is they're doing. I think that most popular command right now for any of the speaker, AI Speakers, this music related. But it does a really good job at it. And not only that it does a really good job with accents now and with gender, and that shows how a technology that’s AI based can evolve over almost a decade. You get to the point where it’s almost ubiquitous now. And the same with be true and is true for natural language understanding. We started off basically using rules as opposed to algorithms, and then algorithms became the mantra and then algorithms that were fenced with rules and so where we are now is that you expect to Alexa understand what you want. There are literally thousands and thousands of pre-written intents to cover everything you can think of, but it's still an intents-based technology. I mean some programmer somewhere had to write code for that specific question you were going to ask.
Corey: Yeah, it's not the first time I've seen technology that was good. I mean the WebSphere product was a good product. When I installed it on my machine, you could actually talk to your machine, you could get it to do things. But not the first time I've seen where it failed, the first time to market going to businesses, but then, when individuals start to use it for personal use such as your phone, such as you know, doing searches on Google things like that, that it eventually worked its way back into business as adopted holistically. I look at the PC in general right? The first days of the PC business didn't adopt the PC. You didn't see it on the desktop. You had the remote dumb terminals that hooked up to main frames, because that functionally is how business operated and it took a good decade before everybody had a desktop at their desk and was able to use it and hook into their corporate systems. But it took really adopting it at home before you then adopted at business.
Brian: And then businesses for years, I mean ever since, really since the 50s had to use mainframes to do big kinds of things that took a lot of processing power. But the idea of a computer being ubiquitous that could save you time every day in your job wasn’t until the 90s or the aughts really.
Corey: And we'll probably see the same thing with AI Technology, that the capabilities of AI will way out-pace the adoption of AI and then it will eventually sift itself out into the use cases within business that can easily consume that AI technology, will consume it embedded within applications, and people won't even know they're using it. Like you said, most people don't understand that they're even using voice recognition technology when they're talking to their phone.
Brian: And everything should be seamless, because you don't want your user thinking about the technology they are using, they just have a need, and the idea is to fill their need, where they want it when they want it, how they want it in the language they want.
Corey: So that being the case going out and in the multiple AI roles you've had in your career, going out and talking to the business about AI, how do you sell those concepts to the business? Or did the business come to you and ask for specific use cases to be solved by AI?
Brian: I think it’s a matter getting them excited about the possibility. So, I have to admit I still love the technology so much and I get so excited about it, when I'm talking to people about it that I'm almost contagious, which is a strange thing to say in the Covid-19 world, but the idea is that there's some real value here. We can do some really cool things that aren't just cool but have effect on the business down the road either on the PNL or on customer sat or on HR retention. All those things have an incredible amount of value to companies, so, we can do some very interesting solutions using things like Watson Assistant and other IBM technologies that really can have real stone effect on the company.
Corey: So, past the point that you’ve identified the use case within an organization, and you've helped them or even at Virgin implementing the AI technology, what were some of the challenges that you faced in implementing the technology?
Brian: So, this is an easy one. The answer really, the biggest challenge is expectations because most people of my generation, their first exposure to AI was on Star Trek or maybe 2001 A Space Odyssey. And, of course, we’re nowhere near today in AI where they were thirty years ago on TV.
Corey: I don't know they didn't have touchscreens so who was more advanced?
Brian: Good point, good point, but Spock could call the computer whatever he wanted. I don’t know if he called it the same thing every time but, in any case, and then has a different view of AI, and they got on the movies as well, but if you look at something like, the movie, what was it called? The movie about the guy who fell in love with his operating system. And in that movie, there's this wonderful scene where he has his personal AI that’s part of his operating system, and he pulls up a video, a 3D video game and his AI jumps into the video game starts playing the video game, which is incredible concept. But we’re so many light years away from something like that right now. But that doesn't mean the generation of people watching it don't say wow, that's great, when can I have that? So, I think expectations is number one item you’ve got to corral when you're trying to talk to a company about implementing an AI solution. And number two is be careful of who you hire to do it. I get something the other day which floored me. I get tired of seeing everybody hanging the shingle out as an AI expert when they have two years working on some kind of AI or they’re going to Google and or use Tensor flow, or something like that and went to LinkedIn, and I went to People and I searched for the phrase AI expert. It came back with 546,000 hits. I don't believe there are 546,000 AI experts on LinkedIn. And in fact, I would be shocked if there are more than five hundred people that really are experts that can hold their own and talk about AI and develop and lead AI companies, for example, that had that kind of expertise. I just, we live in a business society now that everyone claims to be an expert as soon as they can because I think it adds cache to their resume and to their value, and it's become almost meaningless. It's like, and I don't mean this for you, but a lot of small companies, it’s so easy to start a company now that everybody is a CEO of something. I guess I would imagine if we tried to check how many CEOs there are on LinkedIn the numbers would be astronomical as well, because all you need is a couple bucks to start in AWS server and you’re a company and you’re a CEO. And those titles, Vice President. CIO, CTO at least in my mind, have all become fairly common until you look into the person’s background, and see what they really are capable of.
Corey: You know it's interesting when I get to people and talk about our competition, Dayhuff Group having been around for twenty-three years and in the AI space for really seven years, although some of the AI technology, as you have over the last fifteen years, really came from other areas right, so we did visual recognition back in the early 90s, that's AI, based technology. We did you know, speech to text activities back in the mid-90s, that's all AI. I think the concept of AI and how AI could help business really coalesced over the last ten years into a meaning all of its own and the thought processes around, it's not just a computer in technology, it's artificial intelligence and seeking out artificial intelligence really hasn't manifested itself until the last ten years.
Corey: When we talk about who our competition is in the market and you're absolutely right, and I describe a lot of our competition as PhDs that will go into companies and say you know I've got a PhD in artificial intelligence and machine learning modeling. What kind of problems do you have? And the value in AI is not that you know how to do machine learning. The value in AI is you can solve business problem. So, if you don't understand how an industry works, if you don't understand the value of data, if you don’t understand how that data is collected within an organization, how it's stored, how it's managed, then you really can't produce good AI for an organization. You can't solve real use cases until you have that understanding which takes years of experience. So, if you're, like you said, if you're hiring somebody who has six months-worth of experience in the market because they know how to code in Python, then you're really not getting an AI expert. You're, getting somebody who can do a tactical activity for something specific, but can't solve a use case.
Brian: Right and you don't, you end up with a bifurcation, where you have companies that have some gray hair and understand business problems and how to solve them, and then you have technology companies that know how to write Python know which algorithms to use for which problems. But the former is the one that's going to solve your problem and the latter maybe the ones who coded it at some point. But it's the former that's going to actually solve your business problems.
Corey: See you talked about voice recognition. Do you think that's the most heavily adopted AI technology today? Or what would you consider the most heavily adopted?
Brian: I think that the most heavily adopted is text to speech, which is its cousin. Text speech has been around a long time. It used to be a formant technology meaning we created all of sounds without any human recordings, and it could say whatever we needed it to say. But the quality was always very much like this and people didn't like the sound of it. But we have very natural sounding text to speech now and you can tell when it's text to speech, but text to speech applications have been around, I would say twice as long as speech recognition applications, because there is such a need for it with the blind and other areas where you have handicapped people that can't hear, and they need to have it out in text, so they can see the text. Or you have people that are blind and can't read so they have to have it out in speech. So, text to speech is the most successful and most implemented of the AI technologies I think thus far.
Corey: Yeah well, I wouldn't have guessed that one so you're probably right, now I think about it. Well Bryan thank you so much. I appreciate you doing this. Good conversation. That's our time for today.
Brian: Great talking with you today and great being here at the Dayhuff Group with a great bunch of people, and technologists and a really interesting place, doing some really incredible things, thanks.
Corey: Outstanding. Thanks Brian. We’ll talk to you soon.