Let’s chat about AI and chatbots

How do we teach computers human language in all its complexity and context to the point where business travelers can book each element of their trip via messenger or chat? Ziv Baum, an A.I specialist in Tel Aviv, discusses the ABCs of artificial intelligence and chatbots and how they’re being used in business travel. 

Visit mycwt.com

Transcript

00:10

Christine Kashkari (CK): Hello, and welcome back to on the fly, the monthly podcast that dives into issues affecting those of us who spend a lot of time up in the air, out of the office and away from home on business trips. I'm Christine Kashkari. Whether you're a seasoned Road Warrior, or on your first work trip, you'll hear from experts in the studio, who will unpack the biggest issues affecting business travelers and the future of corporate travel. On this and every episode, we'll start off with top tips from travelers to and from their work trips. Then we'll go to Tel Aviv to serve up today's hot topic, artificial intelligence and chatbots with a self confessed AI specialist, Ziv Baum. And finally, we'll end the show with some insights from behind the wheel as drivers dish about clients on business trips.

01:05

We start the show by rolling out some top tips from seasoned travelers. Since we're talking AI, our top tip is related to technology. Specifically, some of the best apps for people who are traveling in a country whose language they don't speak. We caught up with Sandra in Madrid who's trilingual, but still needs help for languages outside of English, Spanish and French.

01:28

Sandra: When I travel to a foreign country, I want to make sure I know the basic words of etiquette. Not only it's polite, but I believe also it takes you a long way. In the past, I used to jot down a few sentences in a notebook and learn them by heart. But that did not work when it came to understand street signs or navigate menus. Thankfully, technology has come to the rescue and to date smartphone apps like translate now in Google Translate will help you with the more complex interactions. You don't even have to type the words you can just take a picture of the sign or the menu with your phone, et voila!, the apps do the work for you. A bit lazy I know but very handy, right? Did you know that even Siri can do basic translations?

02:14

CK: Siri, what's Spanish for Thank you? Gracias. While we're on the subject of language and technology, today's special guest basically teaches computers the human language, so they in turn can chat with us through AI and chatbots. Today, chat bots are everywhere. They send notifications about your pizza order, update your bank account balance or service your hotel concierge. Joining us on the podcast to talk about how AI and chatbots are being used by business travelers, please welcome Ziv Baum, an AI specialist in Tel Aviv. Hi, Ziv. Thanks for being on the show. 

02:55

Ziv Baum (ZB): Hi, Christine, thanks for having me. And it's pleasure to talk about this subject.

02:58

CK: Before we begin, I think it's useful to take a step back and talk about the basics. AI and chatbots are terms that get thrown around a lot as a catch all for all things. How do you define AI and chatbots?

03:11

ZB: AI is not a machine that thinks or even a computer that works the way that the brain works. AI is what the machine does, it's not how they do that. AI it's basically a software system to try to perform tasks that usually takes human intelligence to accomplish. So this includes things like cognitive tasks, planning, reasoning, learning, and sometimes perceptual tasks. So things like recognizing what people are saying or understanding, what are we seeing in any image, those all fall under AI.

03:40

CK: So how many times do you think people come across AI driven technology in any given day without even thinking about it?

03:47

ZB: I think that we see it all over us. When you get an email generally suggesting some answers to you. So things like "yes, it will work for me" or "it's no longer available" that you can just click and respond with to an email. And this is based on AI, the Gmail or Google is actually looking at the content of the email you just got. And it's trying to understand what it is that they're asking you for, and providing you some answers. So they're doing a lot of thinking in the background. Another example, I use Spotify all the time, whenever I listen to a song, Spotify at the end of that song suggests a new song that I that I should listen to the next one. And they base it on what I like, the types of music that I listened in the past, about based on the time of day, maybe in the morning they give me one type of music and in the evening some other. When training, maybe I'll get another recommendation. This is all a lot of decision making that is being done by Spotify or any other music service. This is definitely AI.

04:42

CK: I'm curious to know how did your work lead you into this area?

04:45

ZB: You can call me kind of a geek, I like technology, I like those concepts. So really started to look into this area to understand how it works, what drives them and the AI behind them. And what's the best way to deploy them in a way that makes sense for a traveler, so that's how I got to learn about this industry and those concepts. 

05:03

CK: That's really a nice segway as we chat about chatbots. There seems to be some confusion around chat and messaging, can you help us understand the difference?

05:13

ZB: I can understand why people are getting confused. By live chat, the conversation is a synchronous conversation. So if the agent doesn't respond within a few seconds, the traveler will walk away. And if the traveler doesn't respond to the agents question, in a minute or two, the agent will close the session, and the traveler will need to study it all again. This is usually the service that you get when you go on a website, and you click on a 'chat with us' button. Messaging, on the other hand, is like Facebook messenger or WhatsApp. It's an ongoing conversation, where the traveler doesn't have to wait for the agent to respond, he can send the message, put the phone back into his pocket, and whenever the agent respond to him, the traveler can continue the conversation from that point,. This is pretty convenient for business travelers that are making a booking while in the office, for example, you can send a message before you go into a meeting, and get a response from the agent while in the meeting and respond back after the meeting. 

06:06

CK: Then there's also what is called a hybrid, which uses a man and machine combo. It's something I know you're passionate about, why is this better?

06:13

ZB: Business travelers can ask us questions about many things. And they would expect the chatbot to provide a very good answer on all of those things. And AI is not there yet, it still has a long way to go. So we really like the concept of hybrid bots, where you have a chatbot, that starts the conversation with the traveler, and he's able to provide answers of some things. But whenever it fails, we always have an agent on the background that can jump in and continue the messaging interaction. I think the traveler gets the right answer and is able to fulfill his needs. This is what we call a hybrid Chatbot. And I think it's the right way, in today's state of technology to provide great service to travelers. 

06:50

CK: Can travelers tell the difference, whether it's man or machine on the other side of their screen?

06:55

ZB: People are not looking to talk to a chatbot, they're not looking to talk to an agent, actually, they're looking to do stuff. They're looking to book a trip, they're looking to get help when they're stuck. And this is what you're interested in. And they're less interested on who provides the service as long as the service is good. Whenever the system works, travelers don't really care if it's a chatbot, or an agent talking to them, as long as they get what they need.

07:18

CK: So I understand one of the biggest challenges, though, is a natural language processing, right? Where machines still have a ways to go to understand human context or language context. Can you give us an example of why that is and how people like you are working to teach machines to learn? 

07:35

ZB: Yeah, I'll give you a simple example. When I say I'd like to book a flight, or can I bring a book to my flight. And those are pretty similar terms. And it's very, very hard to a machine to understand that I mean, two different things. This is a very simple example, as we look at more and more of what people are saying, it becomes more and more confusing for machine. The way to go about it is to have as many interactions with the machine. So as time goes by, we get better by looking at many more examples. And learning from what humans are responding to machines, in different cases. Human language is very complex, and it will take time for machines to be perfect at it. The other thing that comes in our way is context. So while the machines are being pretty good at understanding a sentence and responding to a sentence that someone throws at them. Being able to look at the next sentence and the third sentence and providing an understanding the context and the relationship to the rest of what has been said in the conversation is where machines are failing today. If you talk to Siri today, for example, you can ask her a question. But the next question will most likely not take into account the first question that you ask. And that's not how humans speak. So we're pretty limited at the moment. 

08:45

CK: So let's talk about the elephant in the room, or in this case, the dragon, China, are they the most advanced in terms of chat application. I mean, they have WeChat, which has almost a billion users. Imagine the amount of data and machine learning that goes on in there on a daily basis.

09:02

ZB: China, and in a way, a lot of other countries have gone first into mobile, and kind of leapfrog the entire web concept. And this made them a lot more advanced on the mobile side. And when you look at how they use messaging, well, it's amazing. Everything happens inside WeChat. You have payment application, you have shopping, you have everything that a consumer would like to do on the web today, you can do within WeChat. And it's built a little bit differently than what we used to see on Messenger or WhatsApp. It's a lot of different functionalities that we don't have on Messenger and WhatsApp. China is really investing in AI as a way to leapfrog the headstart that the West had on China with regards to investment in software and innovation software, they're using AI to overcome that. So we definitely should and do expect to see a lot of investment in this field coming from China and when you go after your clients in China, it's a must have because this is what the travelers in China would use for everything else so they should have their business travel need fulfilled on WeChat as well.

10:01

CK: So is this the future of messaging for business travelers to be able to book their flight or hotel, for example, within the messaging tool that they're using whatever it may be?

10:10

ZB: A big part of our vision for text based channel for travelers is to provide a service wherever the traveler is, rather than forcing to go into our channels to find the service. So again, it goes into discoverability, and how frictionless the experiences. If we need the traveler to come to our website, or to download our app, we are successful in doing it sometimes. But we are definitely going to miss a big chunk of travelers who are not going to do that. If we put this functionality inside the tool that they're already using, WeChat in China, or WhatsApp in Europe, or Facebook Messenger in North America, then we have a lot more chance of getting into use that and we're even taking it one step further. We actually think that in the business travel context, it made sense to put this functionality inside the messaging application used in the corporate. So applications like Microsoft Teams, Slack, Facebook, workplace are all b2b messaging applications that corporations are deploying for their employees to use internally. And we believe that putting our functionality within those applications will make it easier for business travelers to find the functionality and to use it and to enjoy the benefits. 

11:20

CK: Thank you so much, Ziv. Before you go, can I ask you what advances in AI are you most excited by?

11:27

ZB: I think that 10 years from today, we will get off a plane and jump into a car without a driver, that would be amazing.

11:34

CK: Before drivers totally disappear in a world of autonomous or self driving cars. Let's give the last say on this episode to a Lyft driver in Minneapolis. I asked Julie if there were any gender or generational differences when it comes to travelers using technology while they're in the car.

11:51

Julie: It really is an individual thing. I mean, some people get in and they want to talk or you know, they just start talking and then I talk with them. Some people get in they barely say hi. And then I know to leave them alone. And they start staring at the phone sometimes it's probably more women that do the phone thing. More women, just straight code, go to the phone and I can kind of hear their nails sometimes tapping on the phone. And so I know they're texting like crazy.  

12:23

CK: What about for age then or generations? Whether someone's a boomer or a Gen X or a millennial? Is there a difference?

12:32

Julie: I don't think I've ever had an over 60 year old person getting in my car and just starting to text. So the younger I think the younger you know they go, the more likely it is.

12:48

CK: Thanks, Julie. So talk to your drivers once in a while before they become extinct. And that wraps up this month's episode of on the fly. Tune in again next month. When we check in with Dave Holmes. CWT is Vice President for payments. If you're someone like me, or many of us who dread expense reporting you may want to listen in because Dave will talk about the future of payments specifically, how to make expense reporting disappear. 

We hope you enjoyed this month's episode of on the fly. If you did, please subscribe to on the fly on iTunes, Stitcher, or wherever you get your podcasts and help us spread the word by visiting my CWT on Facebook, LinkedIn and Twitter. 

Remember, we're all on this journey together. Until next time, when we talk to business travelers on the fly.