AI: What We Know So Far, Part 1 - The Startups

It’s been a little over a year now since ChatGPT burst on the scene like a cannonball into a pool, soaking us all. The initial waves have settled down, giving us a chance to see what kind of ideas and applications are making the best use of this fascinating technology. And while AI isn’t exactly new, the democratized access to it that ChatGPT introduced means virtually anyone can make use of large language models and generative AI tools without needing a PhD.

This expanded access in turn is driving the early commoditization of a sophisticated technology, making it even cheaper to incorporate into all aspects of business. It’s unlocking innovation and reducing barriers to entry in a similar way that cloud computing did in the Web 2.0 era, and the internet did for Web 1.0. The ability to hoard teams of engineers and data scientists is no longer the moat it once was.

It’s still early days, though, and there is a lot of runway yet to travel. Joining us for part 1 of our deep dive into the state of AI in travel are two startup leaders operating in this rapidly evolving space, Andrei Papancea, CEO and Co-Founder of NLX.ai, and Stephanie Daniel, CEO and Co-Founder of  Legends.

Follows:

Gilad Berenstein - host

Cara Whitehill - host

Andrei Papancea - guest

Stephanie Daniel - guest

Go Deeper:

• The promise of travel in the age of AI - McKinsey
• 9 Months Into the GPT Era, What Has Travel Learned? - PhocusWire
• Humans at the heart of generative AI - MIT Technology Review
• Ask Skift: the AI Chatbot to Answer All Your Travel Industry Queries - Skift
• Travel Early Adopters Share Lessons After Year With Generative AI - PhocusWire
• How AI is disrupting the luxury experience - Jing Daily

Transcript

Episode 8 - AI Part 1

Guests: Andrei Papancea, Stephanie Daniel

[00:00:00] Cara: Well, hello there and thank you for joining us on the Travel Tech Insider. I am Kara Whitehill. 

[00:00:04] Gilad: And I'm Gilad Berenstein. We are thrilled to have you with us as we dive into the topic of the past 14 months, artificial intelligence. I want to confess that this was probably the episode of season one that I was 

most excited about creating.

And at the same time, I found this to be one of the most challenging to put together. And there are probably multiple reasons why, but there are two that really come to mind. The first is that this new era of AI that we're in what I like to call the great commoditization is so new that we're still in chapter one.

We're still trying to figure out what is happening and things are changing so quickly that it's hard to get an overview, a holistic perspective of what's going on. And the second reason is that AI is an umbrella term. That encompasses an entire wide field of study, and thus talking about artificial intelligence as one thing is often misleading and can really be unhelpful.

But Kara and I were not scared, so we decided to face the challenge. So here we are, and we're thrilled to jump into our first episode about artificial intelligence. But before we go all the way in, Kara, what's new on your end? And I'm wondering, do you have any new favorite AI tools you've been using? 

[00:01:11] Cara: Oh gosh, um.

You know, I, I've been playing around a little bit, but not enough to really have a favorite at this point. I've, you know, like everybody else been poking around with, uh, chat GPT. And, um, mostly my husband and I just come up with really goofy love poems to write to each other. But, um, you know, that's, that's probably the thing I've engaged with the most.

I have of course, poked around with a few of the different travel AI tools, but nothing has really overwhelmed me yet with accuracy or. Really solid capabilities. And I think that points to what you alluded to at the beginning. We're still in the very, very early days of this. And I think there's a lot of interesting stuff to come, but, but we're not there yet.

So, um, that's, that's sort of where I am, but what do you, uh, what do you see on your 

[00:01:55] Gilad: end? Well, there are similarly a lot of tools that I like, but there's two that I actually use on a regular basis. Um, the first one is speechify. And I grew up in Israel, so English is my second language. I often feel like my reading is just a little bit slower than I would like it to be.

So not only does a tool like that help me read faster, but it actually increases my comprehension and increases my recall of information I read. Um, so I find that really useful. And then the other one I really like is mid journey, which is an image creator. And I give a lot of talks at different events and it makes it really easy for me to create like a custom graphic.

For the audience I'm talking to or the city I'm in to make it more relevant for the audience. So those are two that I use regularly, but just like you, there's so many out there. And as I was thinking, you know, this is a good illustration of one of the key points I like to make, which is that AI is just a tool used by people to accomplish their goals.

It's that simple or said slightly more accurately. AI is a field that brings us a set of tools used by people to accomplish their various goals. And it really is that simple. 

[00:03:00] Cara: Yeah, I've, uh, I'm actually in San Francisco this week. And so I've been marinating in, in a lot of the different conversations with folks out here that are, you know, of course, knee deep and all of the ideation for that.

So, uh, anything else we need to know before we dive in? There's 

[00:03:14] Gilad: probably a lot I could say, um, but to get started, just one more thing I do want to add, which is in addition to, of course, the fact that AI is just a tool used by people to accomplish their goals, um, while many people like to refer to this era of AI as a generative era, as I referenced before, I think that's a bit of a short sighted way to think about it.

I like to refer to this era as the great commoditization of AI. And it seems to me that when we look back at this time in kind of the artificial intelligence story, um, We will likely have forgotten about the specific new tool, you know, generative AI in the same manner that most people have forgotten about natural language processing, which was the biggest tool in 2014 that everyone was talking about.

And the thing I think we will all remember from this era is the great commoditization, the fact that open AI followed by lots and lots of other companies made these technologies so easy to utilize. That the thing is, I think when we look back at this period, we'll remember this decade as a decade where we started with only the largest companies in the world and top universities having access to this technology.

And this decade is very likely to end in a place where every single business, and I really do mean every single business, We'll be utilizing AI capabilities within their company and within their organization. And, you know, I'd like to think about this as kind of the fourth chapter in the macro story of AI.

You know, if we think back to 1970s, AI was really an academic exercise within universities. And then in the 90s, early 2000s, we got the optimization era. Which is what brought my family to the United States. And then in the 20 teens, we got the machine learning era, where natural language processing, sentiment analysis, collaborative filtering, and all these other things came into the forefront.

But in all of these various stages of AI, You had to build the capabilities in house every single time. You had to go build out an entire AI team to be able to do this. And like I said, the reason I think of this as the great commoditization is that when we look back to the 2020s, we'll remember this as the era where anyone Could begin to access AI capabilities.

[00:05:14] Cara: So to help us dive into this arena, we've got two really awesome guests joining us today. First, we have Stephanie Daniel, who is the CEO and co founder of Legends. Legends is powering a new era of personalization by transforming the phone and photo data from your phone into actionable preference profiles.

So it's really about the DNA that you, that you have as a traveler and then helping travel providers improve revenue and loyalty and experience by, by taking advantage of all that travel DNA. And then our next guest, Andre Papantia, who is the CEO and founder of NLX. NLX is the conversational AI platform that enterprise teams are relying on to deliver multimodal chat and voice automation.

So thank you both for joining us today. We'll start with you, Andre. Can you please give us a brief intro about NLX and how AI is playing into what you're, what you're building, what you're delivering? 

[00:06:12] Andrei: Certainly, Kara. Nice to see you both. And well, I guess all three of you. Nice to meet you too, Stephanie.

And thank you for having me. Um, and Alex specializes in conversational AI in a nutshell, and simply put, we help enterprise companies streamline their operations by driving automation through any form of conversational channel, chat, voice, and as you mentioned, our multimodal, we have some patent capabilities that allow our customers to, um, synchronize two or more channels to drive a richer and more effective, uh, self service experience.

Um, so anything from, let's say, if you call, um, your, your favorite airline to, uh, change your flight or, um, you know, adjust anything related to your trip, add a pet, a service animal, uh, talk to your bank to make a payment, uh, open up a new credit card. We power those types of experience. It's in practice at scale today.

[00:07:15] Cara: Wonderful. And then Stephanie, tell us a little bit about the AI component of what you're building with Legends. 

[00:07:21] Stephanie: Yeah. Well, great to be here today. Thanks for having me. Um, so exactly as you said, Kar, in the introduction, uh, Legends transforms phone and photo data into what we call your travel DNA. Your profile of who you really are as a traveler and your, your preferences.

Um, and to bring a tangible example of that to life is imagine in your favorite hotel loyalty app, there's a button in onboarding that says create my travel DNA powered by legends. With one click, uh, the traveler is consenting for our technology to, uh, to process. And then we deliver back the travel DNA profile.

Call for the traveler to experience that. And at the same time, we deliver insights to the brand so that they can now finally know their customer better and personalize across every stage of the journey. Example being, we now know that Cara loves to surf in the morning in a Costa Rica type climate in January, uh, with her family and loves boutique hotels and sushi just from that one click of a button.

Uh, and so, you know, how. AI plays into our process. Our company is really think about us as a data company, and that's a lot of where we've been focused is really the data foundation, because that's the pain point that we're really trying to solve here is that our customers and travel providers don't know their customers.

And so, um, We, uh, we're solving that by unlocking this new source of data, and zero party data is, uh, is the concept that consumers are, uh, consenting to share it, uh, and then our process is essentially, uh, capturing, ingesting that data, transforming it into something that's just useful and then delivering it.

And AI can play a role in every stage of that process. But the way we think about AI is, um, you know, our roadmap is to leverage it where it delivers, uh, an optimal, uh, experience for our end customer. And not just employing it for the, for the sake of, uh, you know, employing it. The other place, uh, that we're using it is as it relates to operational efficiency, so yes, within the product, but also across the 

[00:09:24] Gilad: company as well.

Yeah. It sounds like you're eavesdropping on Karen, my conversation before we jumped on the pod here, because we made that exact same point at the day. AI is a tool. It's not the end. It's the means you get to the end. So, to help us dive a little bit deeper, Andre, help us understand, how did you get started when you started an AI startup?

What are some of the things you did? How did you get training? Like, how did the whole thing get started? We always start 

[00:09:46] Andrei: with what we're trying to solve. And similar to what Stephanie said, I think there's, there's a lot of, um, misconception, especially in the segment of AI that then Alex operates in, where Um, I think a lot of companies or people just expect magic and you just use AI for the sake of using AI.

And here's a fun fact. Technologies like ChatGPT are largely impractical for customer facing use cases. Here, I said it, uh, simply put, because you cannot control its output. It's, it's a statistical, all AI ML models have statistics on their core, statistics are imperfect. Therefore, any output out of that cannot be fully trusted, right?

And when you're a brand, you can't rely on that. So we, at our foundation, um, sets a lot of the experience that my team and I have had in the past, including my, my, um, my, my time at American Express where I built their conversational AI platform. From the ground up and there I learned and thereafter that the AI component is not the main driver of automation It's actually the content simply put it doesn't matter if I can tell you're trying to cancel your trip or upgrade your seat If I don't have conversational content to engage you through that.

So as a result, we've built a platform It's SAS. It's no code, but most importantly It has all the bells and whistles that an enterprise company might be looking for, which has nothing to do with AI for any type of application, whether you're talking a website, the mobile app, a chatbot, the voice assistant, you need a life cycle to build and release these applications, you need versioning, you need, you know, language management, you need all kinds of different things that, uh, should be there irrespective of whether you're dealing with AI or not.

And then it so happens that AI and in our class, uh, both natural language processing and, um, large language models slash generative AI can be meaningful. I can have meaningful value in allowing our customers to better scale their operations 

[00:12:03] Gilad: using it. Yeah, I love that answer. One of the things that Ornette Cioni, who was the founder of the Paul Allen Artificial Intelligence Research Institute, likes to say is that AI is often doing this statistical last mile.

It's really code and normal product that's doing most of the heavy lifting, and the AI is doing that last piece, which is, of course, very complicated math. I'm not trying to trivialize it, but it really is a statistical last mile. One more quick question for you, Andrei. What is multimodal AI? There's 

[00:12:32] Andrei: probably a whole bunch of different definitions right now.

I don't know that there's a universal one. Uh, we've been using it for much longer than it's been a buzzword in the, in the market because We've patented our technology, the multimodal technology, I'll dive into a little bit further, uh, that we patented, uh, not too long ago, has been around for like three to four years at this point.

Um, I think more broadly speaking in the market, multimodal refers to any, um, interactions with AI, especially conversational kinds where the inputs to the AI and even the outputs go beyond just a simple, uh, Textual or pure voice and interface or integration where I might ask, I might send a, and I think a lot of us have already been used to this, just maybe not in the context of how it's referred to in the market today.

Remember when Google first released the ability for you to, to search using an image, this is not any different than. What we're seeing today, otherwise through like chat GPT and Google Bart, where you can input, uh, you know, context to the AI system in different mediums. And then you get, you can converse with it, whether you're, you're starting in chat and then you're moving to voice or.

Our double click in that regard, like we, we, we can do all of that, but then NLX also has some additional capabilities that are patented that allows us to synchronize any audio channel to say a digital asset or a mobile device, even an IOT device. And it creates this rich voice guided experience where basically imagine you're on the phone and you're, you're calling up again, your favorite airline.

And. And you might interact with, you might need help with a business process or some, some problem that cannot be solved through a voice only interface. Something as trivial as resetting your password or even selecting your seat is impractical over a voice only channel. So at that point we can, let's say send a notification, a text message to the, to the end user with a link to, you know, go to a website, the mobile app, but we don't drop the call.

So now the voice assistant. In a multimodal fashion is guiding the customer, the end user all the way to the end goal. And if anything goes wrong. You send them to an agent alongside with all the context, so it's not like a broken experience or there's no drop off along that that 

[00:15:06] Gilad: customer journey. Yeah, and it makes perfect sense in the travel setting, of course, because some things you have to see on a map and you can't really explain a map on the phone.

I loved your example of a seating map on an airplane. It'd be very hard for someone to explain exactly where the seat was and where you could just take a picture of that. So thank you for explaining that. 

[00:15:22] Cara: Totally. Yeah. And so. You know, the thing that we, we all sort of know and are learning about AI, it really needs a lot, a lot, a lot of data to train the models appropriately.

And so Stephanie, I know one of the things that you've been working on with legends is the ability to collect zero party data. And, you know, when you were talking about the, um, The hospitality example, right? So I have my, you know, my hotel mobile app and the particular hotel brand may know a lot about me in terms of my relationship with that.

Hotel, but they don't know anything else about me other than the things, you know, if I've taken the time to go into my, to the profile of their app, which hardly anybody ever does. And then you click a bunch of, you know, check boxes like, Ooh, I like these kinds of hotels and those kinds of things and whatever.

But those are so removed from context because I might like a certain type of hotel when I'm traveling for work. But when I'm traveling, or, you know, maybe like is, is not the right word. I endure a certain type of hotel when I'm traveling for work. And then I prefer a different type of hotel when I'm traveling for fun, right?

And so, there's this interesting, as you were talking about that, I was thinking there's this interesting layer of the data that I provide myself, right? That, that zero party data. There's the data that is inferred by the brand based on my relationship and transactions with them. And then there's all this other data.

Like the stuff that I take pictures of when I'm traveling, um, that's captured in my phone, that, you know, We don't have easy ways to tie that together. So, can you talk a bit about how Legends is, is addressing that, and that nature of zero party data in conjunction with the other elements of data that a brand may have about me?

[00:17:04] Stephanie: Yeah, great question. Um, so, um, that was essentially one of our insights. The behind when we first started legends was that there's all this uncaptured data, um, that sits in our phones because we are, we carry our phones every second of every single day, right? Every tool that we use for our lives is on there and we take like 5 billion photos a day.

Uh, and every photo has all this metadata associated with it. And it felt to us as consumers at the time that that data wasn't being harnessed for our benefit. It wasn't making our life. I was still digging out my spreadsheet of recommendations to share with a friend, or like, trying to find a photo of the cool taco shack to recommend.

Um, but it wasn't structured in a way that was useful for me. Then, through our journey, we realized that, um, if we could build a technology that could harness that, that for the benefit of consumers, we could do so and deliver it in a way that was beneficial for brands. And as we kind of dive, dive very deep into the industry and got to know it better, uh, we realized that that was really a fundamental problem.

Like you said, they're stuck with first party data. So the couple of times I stayed with that brand before, maybe the couple of clicks I did on their website, if they track that and organize it, or they pay a fortune and a reliant on third party data sources. Um, trends based, transactional, and by the way, with cookies going away next year, that data sources.

Um, you know, going away as well. And so that's why the future has to be zero party. And that is where, like you said, a consumer consents, they intentionally and proactively share information about themselves with a brand. That's the official definition by Forrester, um, in return for a value exchange, which can include personalization and rewards.

And so that could look like. filling out of a manual form, like you said, but no one wants to do that. And so essentially our technology automates the collection of zero party data in a way that is a seamless, easy, trusted experience for the consumer. Uh, and easy, uh, way for the brand to now, uh, instead of having two data fields about their customer, they now have thousands.

Um. So that's really our foundational premise and studies are showing that consumers are ready for it. Like we're at this paradigm shift where people, you know, 83 percent of people want to share more information in return for personalization. But to your point, we just don't have easy, trusted ways to 

[00:19:27] Cara: do it yet.

Yeah, and then that becomes, it's, you know, when I think about the AI implications, it's sort of Twofold. One is, you know, the ability for you to capture all this data and make sense of it and then feed it out to your brand partners in a way that lets them, in turn, leverage AI in their own applications to make sense of that data and use it for, you know, customer service, like the things that Andre talked about, you know, or marketing or what have you.

So, Um, you know, it speaks to the fact AI is like a tool, but it's based on knowing a lot and having access to a lot, a lot, a lot of data for it to be relevant and interesting. 

[00:20:01] Gilad: So, yeah, the other thing I love about what you said, Stephanie, is that even if someone was willing to fill out the survey, which of course most of us are not, the survey is always going to be finite, no matter how long the survey is.

It's never going to include all the possible questions that we should ask to understand a traveler. But of course, the number of photos on our phone are effectively infinite. Um, and there isn't a limit within there. I think that's a really interesting insight you guys had. 

[00:20:26] Stephanie: Totally. And to Kara's point, the context behind something is so important.

And when you, when you View a large amount of data. You can look at those patterns and, and kind of that context can be part of what you can deliver. Yeah, 

[00:20:37] Gilad: absolutely. Now we were speaking about some of the partners. So Andre, I'm wondering, you know, NLX works with a number of very large clients. How has AI adopted changed over the last couple of years within the corporate world?

And how do you think about licensing technology from a startup, as opposed to from a large player like Microsoft? That's a 

[00:20:55] Andrei: good question. Um, ultimately. What we've seen, especially since the beginning of last year, is an acceleration of interest. Uh, I'd say adoption too, but depends of which class. So, contrary to what you might imagine, um, or what most people might expect, the adoption of generative AI has been relatively, like, uh, meager.

Because a lot of companies are tinkering with it, but they haven't really adopted it in the true sense of the word, because There's still a lot of impracticalities around it for, for production use. Now, uh, on the other front, it's, it's accelerated a lot of conversations because now you have the C suites of all these major companies, uh, pressing down on their teams that they need to adopt AI, whatever AI means, right?

Cause it's, it's being thrown at this just generic all encompassing term. And, um, NLX sits at this perfect intersection between. Practicality and we play at the cutting edge and we have all the latest and greatest capabilities, but we always ground them in what we believe is right and real and practical.

And in that regard, we differentiate quite significantly from a lot of the players in the market, especially the large ones, because, you know, the large ones, for better or worse, maybe they have to play into the hype and whatnot, and we routinely find ourselves. Locking in large customers in a way where, you know, we, we position our technology and our approach on how we might solve their, their, their problems and pretty frequently they'll turn around and say, well, so and so told us this other stuff.

And I, I normally at that point, I tell them, well, then go with so and so I'm telling you, this is how the technology works. This is what I can do. Anything beyond this is. Unrealistic for the following reasons and ultimately just kind of grounding all of this, this, you know, all the expectations, but also like keep them at an exciting level that's practical today drives the conversation for because ultimately all these companies are trying to solve problems.

So if you show them a pathway to solving those problems instead of like just painting them with fairy tales and magic. Moves the conversation forward. So frankly, easy as that. 

[00:23:30] Gilad: Yeah, I love, uh, you did a nice job highlighting the challenge of selling AI or licensing AI in this era of having to keep the excitement and the hype while also being realistic about the results that the clients are going to get from that.

I can't imagine that that's a Difficult balancing act. The other thing you touched on that I think is incredibly important is that the adoption of generative AI has actually been a lot slower than a vast majority of people think out there. And there's this graphic that's been going around. I'm sure many, many people have seen it.

That shows like how many days it took different apps to get to a million users. And it has like Netflix, it took a couple years, and Facebook that took a few months, and then GPT Chat that took like five days or six days or whatever it was. But that graphic is incomplete because it doesn't ask the question of, how many of these people kept using the GPT Chat app after creating an account?

And like you said, we know there's been a huge drop off there, which I think relates to the point we all made earlier about the importance of productization. A good model, an AI, open AI model is very good, but a good model is just not enough. 

[00:24:31] Cara: That's right. And Stephanie, what are you seeing in terms of, you know, I can see this being really applicable and valuable to a lot of different types of companies within travel, right?

And so what are you seeing when you're talking to some of these bigger brands and agencies? Are you finding that they're receptive? Are they reticent? Where are they on the spectrum? of, you know, poking around with AI and deploying it and using it in their applications. 

[00:24:59] Stephanie: Yeah. Um, and I think, you know, we obviously were optimistic that we would receive great results, great feedback from the market as we started talking to customers.

But honestly, we've been very positively, um, Almost surprised, I would say, with the speed at which and depth with which large enterprises have engaged with us, um, and, you know, I think, and also have galvanized multiple resources across their firm. To kind of work with us, because obviously we touch data, we touch marketing, we touch, can touch loyalty, we can touch experience.

Um, and really it's a topic at the C suite level, and I think, you know, the, um, the holy grail of personalization in travel has been talked about for so many years. And I think people are excited that finally we have a new solution that also helps deal with some of the upcoming data and privacy challenges that, um, uh, that brands will be facing.

[00:25:54] Cara: Yeah. And, you know, it's interesting because large enterprises are not known for moving quickly on much of anything, unless there's some FOMO in play, right? And so I could see how, you know, bigger brands are looking at each other and kind of thinking, well, geez, if that company is really kind of at the forefront and cutting edge of making use of this stuff, that's going to make them better at attracting customers, retaining customers, and mitigating some of their own internal cost of operations.

Then. You know, they're going to be more incented to run quickly in that. So, um, very interesting space. Any, any particular concerns that you're hearing from them as you're getting more deeply embedded with their teams? 

[00:26:33] Stephanie: That's a good question. Um, you know, I think one of the things is making sure that it's implemented in a way that the data becomes actionable.

And so that's a big conversation around the right implementation, and that depends on a brand's systems, and obviously our, our goal is that we are seamlessly integrated with their existing systems to make that process as easy as possible, such as integrating with a CRM if necessary, um, and as well as seamlessly integrating within their existing flow with a, with their customers, and Getting the right messaging in place such that the, um, the permissioning process with their customers is done, uh, in a really easy, clean, seamless, fun way that adds to the user experience as well.

So I think, um, with our first customers, you know, we've, we've, we've, um, had some great, uh, great, um, great implementation processes so far, and, and we'll continue doing some great testing with some of the people we have in our 

[00:27:29] Gilad: pipeline. What about you, Andre? Any concerns you're hearing from your clients or prospects out there?

They 

[00:27:35] Andrei: tend to be pretty consistent in the form of, um, new clauses in MSAs, um, and contracts. Like, basically, we've even had an MSA that was sent to us with a clause that we have to guarantee we're not using any AI in the system. I was like, well, you do realize we're an AI company, right? So that's not going to happen.

Obviously that got adjusted or taken out. But yes, I, I've, I've started seeing a significant increase in just legal protections that companies are, are, are looking to put in place. Um, you know, which points to the concerns and, uh, that they have, especially around, they, they classify them all again around AI, but what they truly mean is protections against the likes of open AI and chat GPT.

Um. And then, let's see, there was another one I had in mind that's escaping me right now, but ultimately, um, it's just the general reticence on using some of these technologies if they impact data privacy, security, IP, like anything like that. Again, because we sit at that intersection of practicality and hype, we, we, we show them the pathway that, that leads to, uh, to a successful implementation that doesn't jeopardize brand or, uh, you know, customer data or 

[00:29:06] Gilad: anything like that.

Yeah, I mean, a lot of the big companies I work with have been really concerned about IP. And basically, by putting things into an algorithm, does that somehow make them fair use? And there's been a lot of questions around that with that. As we said at the beginning, it's still early, and we actually don't know what the answer is going to be to some of these questions at the end of the day.

One more question for you, Andre. In addition to being a CEO and an entrepreneur, you're also a professor. So how do you encourage people, whether it's your clients or your students, to think about AI when they're building mental models around this? 

[00:29:38] Andrei: I, I like to be as pragmatic with my students as I am with, with you in the audience today.

Um, and it's, and it's, it's all about being transparent. I was actually having a conversation with, um, you know, some of my former students the other day and, um, They were, they were just trying to understand, they asked me about, like, what's our, what's NLX's, um, stance on, um, responsible AI. And what I told, what I told them is that, well, ultimately our, our stance is only use.

This, these technologies in ways and for use cases where you have confidence that it's not going to go awry, right? So don't, don't, don't do what Expedia did or what Chevy did or many of these other companies kind of like put out these literally like thin layers around ChatGPT and creates these, these weird experiences that taint brands and It's like you're not even I don't even know why anyone sees this as innovative in any way because it takes like five minutes to put out an experience like that out, right?

So ultimately it's it's it's a lot around. I I just love to ground people in practicality like There's no point sometimes to Stephanie's earlier point again, you might not need to use AI at all. And if you use it, just know why you're using it and know what value it brings to the problem, the solution, the stack.

Otherwise, just drop it. Like sometimes a good old if statement to just get a little technical here is all you need. Like it and who cares what you're using if it works and if it checks the boxes on safety, security. Uh, and, and overall like, um, privacy, like those, those things should just be held to the highest degree.

And if you have concerns that introducing a certain class of AI system might jeopardize that, don't 

[00:31:46] Cara: use it. That's a, that's a great, I love the practicality aspect of it. And that's a great way to frame it. I think, because there is a lot of hype and I think people want to run fast on something that has buzz to it.

But if you really stop and think about what is the end goal. Of the use of the technology and then walk it back that way. It'll it'll tell you what you need to know, right? So Stephanie when you look at the the bigger enterprise players that are out there Who are you intrigued by? Who do you see that's at the forefront that's actually doing really cool stuff and, and, you know, the AI realm, whether it's generative or otherwise, um, that, that you're seeing that you're thinking, Ooh, this is somebody I can really learn from.

[00:32:26] Stephanie: Well, like you said, it is such an exciting landscape and moving fast. But I think if we look at the big picture, Guys who I think has done well from the beginning and continues to do so is Microsoft, right? An obvious one, but from what they've done for consumers and also for enterprises, I think the way they've implemented Um, in a seamless way with tools that people were already using has made it very accessible.

And Gilad, to your point, like, if this is the democratization era of AI, I think that they've been a leader in, in doing that. Um, from a consumer perspective, uh, something that I find fascinating. Do you guys see the keynote about the rabbit? Yeah. It's this new, uh, I don't want to call it a smartphone, but essentially they launched it at CES and his, you know, their premise is that you, they're using large action models.

And so they're going one step further that when you tell your rabbit to do something, it can actually take that action for you. Um, so that's, I think watching that space is, is, uh, is going 

[00:33:29] Cara: to be interesting. Yep. We all need, we all need one more, one more device to carry around with us. Exactly. Exactly.

It's interesting to watch what, what emerges and like new ways that we hadn't even really thought of. And then something random pops up and takes hold and then we're off to the 

[00:33:45] Gilad: races. Yeah, absolutely. I love your answer, Stephanie. You know, the company that I often think about. doing a really good job is HubSpot.

I'm a big fan of Ben Thompson and the Stratech, Stratechery, um, newsletter. And one of the things he talks about is jobs to be done and how businesses have specific jobs to be done that consumers or clients are looking for them to solve, and there's jobs that they don't want that vendor to solve. And if I think about the integration at HubSpot, they basically took the open AI API.

And didn't build anything new with it. They just reinforce the existing features. They already have their consumers already love already use, and they made those features more efficient. So I think that's a really good example for big companies, rather than trying to come up with a whole new part of your business, a whole new job that your consumers are not asking you to do is using the technology, the tool.

to do a better job with the things you already are doing for your clients or consumers. I thought that was a really great answer. One more question for you guys, um, on this front, is what advice would you give to other entrepreneurs pursuing AI startups who are trying to sell into larger companies?

Anything you can, you've learned you can share with them, and maybe we'll have Andrei kick us off. Think 

[00:34:55] Andrei: about the problem you're trying to solve and work backwards from that. If along the way, The solution that you, you come up with could benefit from AI, incorporate it, but any, any sort of, uh, product out there should start with the problem that it's trying to solve.

So just work backwards from that and pay attention to the customers and technically everything else just kind of falls into place. I said technically because, you know, the journey is not quite as linear as anyone might want along the way, but persistence wins in the end. 

[00:35:34] Gilad: Absolutely. 

[00:35:35] Stephanie: And Stephanie? Um, I would add kind of, uh, get comfortable testing, exploring proof of concepts that don't necessarily need to distract from your core work, um, but so that you can, um, you know, think about how you can optimize or improve on what you're already doing using, using some of the new technology.

Um, that's something that we do. And then, yeah, the operational efficiency point. I think not enough companies are thinking about how they can leverage that, uh, as well. Um And you said advice for selling into companies. Uh, I think really clear value proposition. Uh, and I think that's something that has, as all startups learn, you, you, you evolve over time.

And that starts with listening to your customers and understanding what they need, but, uh, leading with the value proposition rather than like the features of your technology. 

[00:36:24] Gilad: And it's really hard, I assume, in this phase to differentiate yourselves because, of course, every startup in the world can now credibly claim to be using cutting edge AI.

So I imagine that really differentiating yourself and figuring out what your value prop is even more important, and it's always been important. 

[00:36:41] Cara: So as you both look into the future, what are you most excited about that AI can bring to the table that we're not necessarily doing or seeing yet? Anything that jumps out at you, Andrei?

What 

[00:36:58] Andrei: Gilad said at the beginning of the session really resonates with me. AI and technology in general, more broadly, is here to make humans more efficient, to augment our lives. So I'm super excited to see what, what I really like about what OpenAI has done or what it's catalyzed is it's brought the industry to the forefront, but more so, and actually not just OpenAI, mid journey, right?

I feel like all of these technologies just kind of exploded around the same time, even though they're different classes of AI and even different classes of generative AI. Um, but it's, it's made it really easy for people to get creative. So I think we're, I'm very excited to see about all the creative, like here's, here's a concrete example.

And frankly, this might not be the ideal example to share, but I've seen it and it's super creative. Um, I've seen this, this app that people can use for interviews. Um, to, to imagine like I'm interviewing for like aerospace engineering and then it's like listening in real time to what you're, what you're, um, you know, what the conversation is going on about and it's offering you suggestions on answers that you, you can give back and obviously that's, that's That's the nightmare of anyone that's operating a company or a role, but I gotta give it to whoever put that together.

It is super creative, right? So we're going to see things like that. We're also going to see like super interesting uses of these technologies that simply fast track learning. I mean, universities are already adjusting to this and they just have to accept that you're normal. They're like maybe just throwing students and just crafting an essay might not be the best way to assess their skills.

Cause Hey, you can put that into JGPT or BAR and then you'll get back an answer in like 30 seconds. So, um, I love to see what people come up with in the next couple of years and, you know, some use cases we'll benefit from, some not, but I'm sure we'll see a lot of exciting things. 

[00:39:15] Cara: Stephanie, anything particular jumps out at you that you're excited to see?

I'd say I 

[00:39:20] Stephanie: agree overall with the excitement about it should make consumers lives and businesses lives better, should make us productive, um, more productive, happier humans. Um, like some of the safeguard rails that might need to be put in place, you know, is a whole other conversation. Um, but separately, I think something that I'm excited about is I think there's been a realization that AI is great, but the value of proprietary data sets to power it is what's needed.

And so, uh, I get excited about that realization as I think about, um, the stage we're at with Legends. 

[00:39:51] Gilad: Yeah, one of the things that I'm really excited about is this vision I have for auto organizing databases. You know, historically, and I still think today that for a lot of businesses, it's really hard to get started with AI or predictions of any kind because the data is not structured correctly.

And that really can be a massive undertaking for a large business to get that in order. Well, very soon, I think we're going to get technologies where the database organizes data for you. It will use the logic and insight that it has from a generative model to be able to understand the relationships within things, within an email or any sort of a document.

And it will make it much easier for businesses to access this type of, this type of technology and to make predictions. And I think that's going to be a really, you know, a really exciting opportunity for many more businesses to participate and to benefit from this. The other side of the question for both of you, anything that you think is being overhyped or just totally misunderstood by the general public out there?

And maybe we'll start with Stephanie this time. 

[00:40:45] Stephanie: Um, yeah, I would just say, I mean, obviously chat GPT is great. It's been the thing that most people have been able to play around with. But I think from a. Business perspective. I think there's been probably too many businesses that have been created that are just another version of that, or a very light UI layer over that.

And again, how valuable is that if you're not adding more data or a different, um, you know, different element to it. So I think that's, that's some overhyped there. 

[00:41:11] Gilad: Absolutely. Andre? 

[00:41:12] Andrei: I think the doomsday, um, elements of AI are, are certainly being overhyped and Nothing about these systems, even their state of the state of the state of the art today.

They're not sentient. Computers are rather stupid, actually. And they're just really good at executing a ton of instructions, including complex math, very, very fast, right? Which humans are not great at. Humans are fantastic at dealing with abstraction, on the other hand. So, um, I think it's just important. It goes back to that baseline, right?

Of like, what's really practical and what's just It's just purely hype, right? So um, I think the state of the art of this technology today is just phenomenal and it's leaped forward so much, even in the last couple of years and it's going to keep progressing. But I think we're very, very far away from any form of Terminator style scenarios that are 

[00:42:14] Stephanie: Have any of you watched The Creator?

No. Okay. It's a new movie out and if you want to explore what sentient AI is. Might be like in the future. Um, I recommend it. All right, 

[00:42:26] Cara: making a note of that 

[00:42:27] Gilad: one. Absolutely I'll just add that, you know, Yuval Romerari likes to remind us also that AI doesn't have to be sentient or conscious For us to be concerned about its implications Which I think is important people people forget that, you know, I agree with Andrea 100 percent I've seen zero evidence that any AI out there has any version of sentience or consciousness But that does not mean we have nothing to be 

[00:42:49] Cara: concerned about.

Totally true. I am I'm curious about what you're seeing in the fundraising side too, since you both are startups at different stages, um, you know, from the investor's perspective, this all happened very quickly and kind of came out of nowhere and then immediately you're seeing a thousand startups that are, we're AI for this, we're AI for that, we're AI for the next thing and, and you're kind of like, wow, you know, I'm not an AI expert, but it seems weird that you could suddenly, all be AI based companies for, for things.

So, you know, from the investor side, it's, it can be a little tricky to understand, you know, who is, who is doing real things and who is that light layer on top of. Um, so I think there's been a little bit of, um, in some, in some cases, reticence from the VCs on where to write checks into, um, taking a little more due to, and that is, of course, combined with the macroeconomic climate that we're seeing too, right?

So, you know, um, checkbooks have been a little bit slower to open up, but what are you both seeing from the investor side and fundraising side in terms of, um, interest in, in capabilities to evaluate it? AI based startups. 

[00:44:01] Andrei: I'm seeing a climate where we're just about to close on a series A. So for us, it's been great.

Um, thank you. Um, just a few more days. And, but what I've seen, and not just seen, but also heard from my founder friends and whatnot, like it's For anything that's not in AI, it's, the powder's a little dry right now, but for anything AI, investors are like super excited about it, but to your point, Kara, they're just, not as many investors have the right subject matter expertise in house to know how to vet these companies and these technologies.

They're really big players, they're really big VCs, they'll just kind of throw money at anything. I've seen a couple examples, won't name them. It's like, what, like, Why? And I guess the answer is because they have a lot of money to burn, but the smaller funds are being a lot more cautious, right, and, uh, probably putting even more effort into, into how they assess these companies, um, because of, of the, I guess, facade or the gray area of like, are you real or is this just, again, a thin layer around something else?

And I guess the other interesting thing I'll mention, um, In, in this round, and I think others will start seeing that too, there's been generative AI clauses inserted into the deal docs for, for the, uh, equity transaction, right? And certainly that wasn't the case a year ago. Um, so it's, it's interesting to see how generative AI has impacted, um, the legal sector, uh, beyond just, um, you know, consumer lives or, you know, the prospects of what you can do with 

[00:45:46] Stephanie: this technology.

Yeah. Um, I'll maybe just add, so I think definitely AI data company has generally been positive reaction to that. I think there's certainly layers of diligence, um, and. There is, like you said, almost that, uh, another AI powered company, you know, how real is it? There's definitely that lens to it, but I think we at least have experience that once people see what you're doing, they're like, oh yeah, okay, you know?

Uh, I think there's also still this lens that we all know of, that, you know, travel is, um, not necessarily a easy VC, per se, um. But I think at the same time, people see it as a sector that has a lot of data, and so they do see a lot of potential for data AI related solutions, where travel might be one use case, um, and so I think that's been interesting.

The other trend that we're seeing, which is not, you know, surprising, but is the level of traction or milestones that I think investors want to see at each stage of a company is just that little bit higher, right? Um, whether it's from customers or revenue, etc. And I think that's just a little bit of the The, I won't say new normal, maybe reset.

Yeah, and 

[00:46:55] Gilad: I'll add that one of the things that I've seen is people really trying to figure out is how much Value is being created or what kind of ROI are the clients of these AI startups Really getting and it's a really difficult thing to assess at this point Which is why I think we're very likely to see a number of companies that were very exciting AI startups a year ago Shutting down or doing aqua hires this year because it's just very difficult for them and their investors to assess Has their platform created any real value for the consumers?

And as we've all been saying this conversation, that's what actually matters reminds me a bit of, um, in our last episode, we made a point at some point that, um, business models still matter. Well, that's true. And AI as well, business models and value still matters, even if your AI is very cool and very new.

[00:47:45] Cara: Yep. Yeah. It reminds me of, uh, if you all watch the old, uh, HBO series, Silicon Valley. And, uh, remember the, the not hot dog, the visual recognition algorithm. It's like, okay, well, you know, AI is kind of fun and interesting when it's sort of a toy and then you kind of find these really random use cases. But, um, I don't know what the business model is behind something like that.

Um, so you, you never know, but yeah, it's, it all goes back to business models. 

[00:48:15] Gilad: Yeah. I guess that there's not a big market for AI in Shakespeare plays about avocados. Well, thank you guys. This has been an amazing conversation, and as we said at the beginning, it's really just the beginning of this conversation.

Um, we're at a new technology epoch, and we really appreciate you guys being here to try to unpack some of this with us. But before we let you go, we do have a set of lightning round questions. So you're going to have 30 seconds maximum to answer these questions, and Kara is going to kick us off. 

[00:48:42] Cara: Okay, we are going to start with you, Andre.

Tell us about your favorite Halloween costume. That you've ever worn. Vampire. A vampire. That would make sense, right? Yeah, you are from Romania. So, that, that fits. That was, that was, uh, That was a softball question. Stephanie, what about you? 

[00:49:05] Stephanie: Oh, mine's, uh, much lamer than that. Generic, uh, Gothic. Uh, Queen of the Dark, I think, is my go to Halloween costume.

Nice. 

[00:49:19] Gilad: Yeah, and I think Andra gets the award for being the fastest answerer to any of our Lightning Round questions ever. He was just ready with the vampire right there. Alrighty, next question, and Stephanie, you're gonna go first. What is one AI magic trick you hope that AI can solve for you and it's okay if the technology does not yet exist?

So what's one magic thing that you hope AI can do one day? 

[00:49:41] Stephanie: Make time stop so I can get a few hours extra sleep sometimes. 

[00:49:44] Gilad: Me 

[00:49:48] Cara: too. Yep, that's a good one. I'm, uh, I'm plus one on that one. Um, okay, last one. Andre, if you could have one superpower, what would it be? 

[00:50:01] Andrei: Teleportation. 

[00:50:04] Cara: Mm, very good. Just across distances and current times, or would that include, uh, time traveling teleportation?

If I could 

[00:50:11] Andrei: have the bonus of time travel, I'll take it. But, but if it can be anywhere, you know, on the universe, whatever, I just Just let me, let me 

[00:50:21] Cara: travel. Nice. Stephanie, what about 

[00:50:22] Stephanie: you? Very similar. I, my instinct was to say fly so that I can go wherever I want, whenever I want. 

[00:50:28] Cara: Very good. 

[00:50:29] Gilad: Absolutely. Well, we'll have to come up with some pretty good AI to accomplish those two goals for you guys.

And we just want to thank you again for being a part of this conversation with us and wish you all the best and keep on innovating. Thank you so much for having us. 

[00:50:42] Cara: Thanks for having us.

Well, that was a really fun conversation and, um, you know, a lot of things to think about that. You know, as we're all getting more up to speed on AI, it's fun to hear from the people on the front lines that are doing the thing. Right. But, uh, what, what did you take away? What were some of your top, top hits from that conversation?

[00:51:03] Gilad: Yeah, there, there's a lot. I mean, one is exactly what you said. It's kind of crazy how broad this topic is and how even in the short conversation we touched on sales and HR and legal, and it really is touching every aspect of every business, um, which is a really interesting kind of. element here that's probably different than some of the past technological advances that we've seen over the last 10 or 15 years.

Um, another thing that I thought was, was, was great is both are making the point really clearly that value propositions and actual value creation for their clients is what matters most, not starting with AI as the answer to all questions, but using it as a tool, as we all spoke about throughout the conversation.

I thought they both spoke really eloquently about that. Yeah, I, 

[00:51:49] Cara: um, I'd echo that I think, you know, the past year and whatever, three months or so since, um, chat GPT came out, I think AI was a hammer and everything was a nail, right? And so, you know, we're, we were seeing a lot of hype around that and I think it's, it's dissipating a bit, but the one thing that I, um, I found interesting that both of them commented on was the.

The speed and intentionality and executive level focus that AI use cases are getting within big enterprise companies and their willingness and ability to mobilize quickly to explore and understand how it can be deployed within their, their different businesses. Right. And so that could be both from a revenue generation perspective or from a cost mitigation perspective.

So, um, I was encouraged to hear that. And, and, um, you know, this is something that's. It's still, like you said, very early days, but companies are being intentional about how they view using AI and putting it in place. And then I think the thing that Andre mentioned a few times is. You know, he described it as, you know, fairy tales and magic, right?

It's not the answer to all questions. And there may be some cases where you really don't need AI, you know, a good old fashioned, if then statement, right. Can, can accomplish what you need to accomplish. And so I thought that was a great way to frame it. And, you know, for startups in particular, thinking about.

the intentionality of the problems that you're trying to solve and making sure to your point, you know, you communicate that and the value prop that you're going to deliver for them and that the clients think about. Is this really necessary? Is this the best tool to accomplish my, my ends? So, you know, I think we'll, we'll see more of that and I think it will be shaking out quickly as we, as we move forward, but, um, good conversations all around.

I'm excited to watch both, both of them and, and what they, uh, what they deliver in the market here. 

[00:53:39] Gilad: Yeah. And as an investor in NLX, I was very happy to hear the announcement about the Series A that is about to be closed. Um, And of course, I'm very thrilled about that. And then they both want, uh, teleportation technology and please sign me up.

You know, I'm the guy who loves to travel, but would rather skip the flying experience. So, uh, I'll definitely sign up for that kind of AI. Good stuff. Yeah. Thank you very much. And we look forward to having you guys back with us. And next time we're going to look at the other side of the coin. Instead of talking to the startups innovating at the bottom.

We're going to talk to some of the biggest companies in our industry and see what they're doing with AI from the top. 

[00:54:14] Cara: Thank you for tuning in to the Travel Tech Insider podcast. We hope you enjoyed today's episode. A big thank you to our guests and to our producer, Zach Vanass of Dinosaur Trips. To stay up to date with all our latest episodes, make sure to subscribe to our podcast on your favorite podcast platform.

If you have any questions, comments, or suggestions for future episodes, we would love to hear from you. Find me, Gilad, and the Travel Tech Insider page on LinkedIn, or you can email us at TravelTechInsider at gmail. com. We'll be back with another exciting episode soon. Until next time, travel well.