Breaking Language Barriers with AI with Steve Rotter

Russ Fordyce 🏆 (00:01.21) Welcome back to the Winner Circle. I'm here with Steve Rotter from DeepL. They are a really cool AI company now doing deep language learning. And Steve's going to tell us all about the intricacies of learning languages in this AI world. So Steve, welcome to the Winner Circle and congratulations on the award. Steve Rotter (00:19.212) Thanks, Russ. Great to be here. Russ Fordyce 🏆 (00:21.2) So before we go too deep into all of this AI language translation stuff, give us a little bit of background about you and DeepL. Steve Rotter (00:30.06) Yeah, sure. So, Chief Marketing Officer here at DeepL. It's been two years since I joined the company. This is my fourth time as a CMO. So, I've had a wide range of experiences from marketing technology, supply chain technology, developer technology. And this has been a very, I mean, just a fascinating journey working for an AI company. And you know how incredibly fast moving this market is. and it's kind of fun to be in the middle of the vortex. Russ Fordyce 🏆 (01:02.64) Well, and I was just talking to a colleague, a vendor colleague of mine, and talking about how amazing it is just in the marketing realm, being in marketing for 30 years, how impactful AI is. Everybody said it was going to be the biggest impact on health and healthcare and stuff. I think right now we're in the golden era of AI marketing. Hopefully it's not all slop. But you guys are outperforming the big guys in this area. What do you think is the biggest advantage you guys have in kind of helping organizations transform using AI? Steve Rotter (01:40.236) Yeah, I you know, it's interesting. You know, I think both of us being here in the US, we kind of have a this tendency to underestimate the impact of language. And when we think about, you know, just as a baseline for people to wrap their heads around, you know, 30 % of US households don't speak English as their native language. So even if you're just marketing in the US, you're an international company, whether you realize it or not. So language is critical. It's one of the most fundamental things in terms of marketers getting their message out and getting the right story and all of those things. But when you ask about like why DeepL, why the advantage, it's been this just absolute maniacal focus on specialized models focused around language. And this has been a debate that's been in the market for at least the last 24 months. Between these kind of general purpose models I can use it to write a poem about my dog or I could use it to do my homework or whatever you want to do versus specialized models and that's where we've decided to focus with you know training data everything aligned around just capturing the nuance of human languages and Doing that, you know Pretty well in you know over a hundred languages Russ Fordyce 🏆 (02:59.066) Yeah, I liken it to, think you're dead on right with these kind of specialized models. You I always tell people, we don't, you can go to Walmart and get a tennis racket and you can get a fishing rod, right? And you can get some clothes, but you're not getting, you know, world-class tennis racket that you're going to go win Wimbledon with. And you're not going to get a fishing rod that you're going to catch, you know, the best fish with. So you do have to go to these specialty shops and it's lot like that with AI, isn't it? Steve Rotter (03:22.67) You Steve Rotter (03:26.382) Yeah, really is. And again, think language is something where you're trying to capture the unique nuances of specific phrases, specific grammatical structures. And doing that in kind of just like this general purpose model setting just doesn't work well. mean, you can look at the data that we provide on our website. I'm not going to get into too much of it. But time after time, when you compare the blind test results with professional translators. mean, you know, we just most recently did one with over 44,000 blind tests comparing DeepL to chat GPT, Google, et cetera. Like almost 90 % of the time, they're preferring the language or the results of that translation. So what does that mean? I mean, you know, some of us, when we think translation, we think, I'm going on vacation. I need to translate a menu. It's not a big deal. Right? No one's going to be upset if you get that wrong. But if you're a business translating contracts, if you're a business translating critical healthcare information with your pharmaceuticals, if you're a news agency translating into multiple languages near real time, well, you've got to get that right. And it's going to have a huge impact on your business, not just the quality of the language that goes out. Steve Rotter (05:18.028) Russ, I completely lost the audio on that one. I don't know if the, did the mic die? Ha ha. I was trying to read your lips and I'm like, well, I'm not going to get that right. Russ Fordyce 🏆 (06:39.849) You still there? Steve Rotter (06:41.325) Yeah, there we go. Russ Fordyce 🏆 (06:42.771) All right. Let's. Russ Fordyce 🏆 (06:49.169) Get back in here. Russ Fordyce 🏆 (06:57.449) I'm going to pick up kind of where we left off. Steve Rotter (07:01.474) Yeah, no, it's fine. Russ Fordyce 🏆 (07:04.595) Haley is still backstage. So I was talking about trust. Give me one second. Is my audio okay? I know it all picked it up. Thank you for letting me know. Steve Rotter (07:12.236) Yep, sounds good. Russ Fordyce 🏆 (07:37.619) Gotta love tech. Steve Rotter (07:40.024) There we go. No, it sounds good. Russ Fordyce 🏆 (07:41.863) Okay. Alright, let's pick it up. we're recording still. Okay. We'll edit all this out. Alright, it will be great. Steve Rotter (07:53.568) Okay, sure it'll be great. Russ Fordyce 🏆 (08:03.335) All right, so we're gonna talk about trust and accuracy. All right. Steve Rotter (08:09.016) Okay, sounds good. Russ Fordyce 🏆 (08:12.435) So I know you guys have developed a reputation on trust and accuracy. And obviously, we don't want robots talking for us, and we don't want them acting for us. How did you guys go about kind of building that reputation? Steve Rotter (08:26.69) Yeah, I think it comes down to a couple fundamentals. One is knowing the problem you're trying to solve, and starting with what I would say is a tremendous amount of data. I all the LLMs, they're certainly trained on, I'll say, immense amounts of data, including kind of the garbage of the internet, if you will, which is unfortunate. But that is part of the challenges and resulting in hallucinations and things that have happened in the past. although that's certainly getting much better. We have over almost seven years worth of proprietary training data. So massive amounts of data that have gone into very specific human language translations, perfecting those, tuning those, tweaking those in over a hundred different languages. So if you imagine the wealth of information that's gone into training our models, it's completely different than what you might get if you're just, again, dropping something into your kind of run rate. chat GPT or something like that. So ultimately, you know, that's what's building the trust, right? If you think about, you know, the quality, professional translators rely on that information all the time. And if they can get a, you know, I don't want to say, you know, perfect, but near perfect translation when they're translating, you know, large content, whether it's a, you know, 50 page document or legal agreement or things like that. Well, then all of sudden, you know, they're going to start recommending that they're going to start saying this is a better translation. but again, it's, it's a very subjective thing as well. So when you get into translating language, it's never, it's not like a math equation, right? It's not like you just, you know, you always get the same answer every single time. again, building on that, you know, years of data and quality translation with, that human nuance of expression. is something that's really built our reputation up. Russ Fordyce 🏆 (10:22.771) Well, it's interesting you say that because you think of language, and I always think of language when we were dissecting sentences as kind of a math exercise of you got a noun, an adverb, and all of that. But you're right, because the difference between math and language is context. It's all the words around it. It's where you are. It's the setting. It's everything. So how do you guys get that context? Steve Rotter (10:30.428) yeah. Steve Rotter (10:34.893) Yeah. Steve Rotter (10:49.57) Yeah, I mean, it's it's fascinating, right? If I said, meet me at the bank. And all of a sudden you're sitting there going, well, I'm going to meet him at First National, what have you. And I'm standing on the corner with my fishing pole saying, you know, let's go. Come on, Russ, we're going to go out and catch some catch some trout in the Blue Ridge Mountains here. It's a totally different story. And that's that that context that certainly builds up over time. It's having. looking at larger quantities of information when you're doing that translation. But it's also other things. One thing we introduced last fall was this ability to highly customize the models based on your language. Because if you're a Fortune 500 company and your particular brand terminology or product terminology is something that is unique to your business, well, you want that to show up every single time the same way, regardless of language. We built that customization into the models where you as a CMO, if you will, or a marketer or a legal expert or operations can fine tune that to be exactly what you need to say every single time. So again, building on that trust. Russ Fordyce 🏆 (12:02.121) And so in practice, what does this look like? I'm coming to you as an awards company who wants to market itself to 100 different languages. I tapping into your API and pulling all my site into your site? How does it actually operate from kind of a customer standpoint? Steve Rotter (12:10.851) Yeah. Steve Rotter (12:23.544) Yeah, so there's a couple of different ways. There's certainly what I would call the two box translation that you would imagine if you go to a tool like Google Translate. Put it on this side, tell it your language, and it comes out on this side. That's pretty basic. That's where short form translation is just being done for things like emails and things like that. Then there is what I would call the document translation lens, which is You you publish your annual report and you want that to be in 30 languages instantly, know, drag, drop, done. It's literally that simple. Now that sounds easy, but if you think about the formatting of like a, an annual report with the page breaks and the images and wrapping all around that and doing that in a way that doesn't create a massive break in the, in the translation of the, of the content, because words are longer in certain languages. So you're to have German maybe 20, 30 % longer than English in many cases. All of a sudden, your document is being reformatted instantly on the fly. Or languages that read right to left versus left to right. Again, very, very different in terms of the translation. So that's another way in which companies use our technology. Russ Fordyce 🏆 (13:30.601) graphic designers nightmare. Steve Rotter (13:46.479) And then last is, like you said, the API. So company like Notion, in fact, I just recently did an interview with their head of localization and they're using DeepL real time, right? So their website, their apps, their services, 20 languages near real time using the DeepL API. And it's literally just saving them massive amounts of time. Russ Fordyce 🏆 (14:10.633) Well, yeah, and it's saving massive of time and energy, but it's also expanding that universe as you were talking about. If you're only in the US and only in English, you said it was 30 % non-English speaking. That's new to me. I was thinking about, as soon as you said that, I was thinking about MrBeast. And MrBeast, the famous YouTuber and now Netflix star or whatever channel he's on doing this game show. A few years ago, he started doing Steve Rotter (14:30.888) I'm Steve Rotter (14:35.428) Yeah. Russ Fordyce 🏆 (14:40.755) translations of his own content, I think using AI and other things because he realized that his audience was global. And it radically changed his business because now he was Mr. Beast in every language. So I imagine these companies, companies like Notion, where you're not limited. You've only been limited by language, it seems like. Steve Rotter (14:47.748) Yeah. Steve Rotter (15:01.629) yeah. I mean, you think about it, you think about a company like Notion, certainly an amazing example of great collaboration software. Well, by definition, their product serves people that need to work together no matter where they're located. We also see this with e-commerce companies, right? You think about a large e-commerce company, for example, one of our customers in Europe, you know, well over 100,000 SKUs. Well, can you imagine, you know, translating that into, you know, from from Dutch to Polish and you want to reach that new market. Well, that would take years and by the time you get done with that, you're probably out of the market. exactly. So again, that near real time immediacy, giving customers access to new markets, it's absolutely a game changer for them because all of a sudden they're not saying, well, we launched in our primary language, let's say it's English. Russ Fordyce 🏆 (15:39.305) Your product's out of date. Steve Rotter (15:56.377) But we have to wait three months before the German content is available and the French content and the Italian content and the Japanese content. You you're losing precious time. In today's market, you can't do it. Russ Fordyce 🏆 (16:07.869) Yeah, and now I know you guys have also turned to an agent tech model. And I think that's really cool because like you said, you're not looking for assistance anymore. You're actually taking stuff into action. So what does that look like? What does that agent model look like for you guys? Steve Rotter (16:25.901) Yeah, I mean, we are, definitely a multi-product company. In fact, you know, last, last fall, as we introduced some of these new capabilities, we also introduced things like DeepL Voice. I'll spend a few minutes on that because that's actually fascinating as well. When you, you know, when you have conversations in multiple languages with multiple people, regardless of where you're located, that's hard to do. I will say that's, you know, that's one of the most challenging things from a language perspective, because unlike a document where you can read the entire context of the document while you're translating in a live conversation like this, you don't know what I'm gonna say next. And picking up that translation and nuance is something that's difficult. But you asked about agentic. We've been exploring ways in which we can help support kind of that full translation workflow. So we've introduced capabilities to... help with not only the kind of ingestion, the translation, the routing, human in the loop translation or human in the loop review, all the way through to publishing. So that's kind of our vision is much more of a seamless flow from the time that you said you need something translated, not only to a system that can do that really well, but the systems that ultimately where that stuff resides. Like you said, if it's marketing content, does it just automatically push out to the website? Is it a PDF that you're automatically putting into a repository for your sales representatives? Russ Fordyce 🏆 (17:57.469) Yeah, now you guys have also, I I think in your nomination, you were talking about AI swarms. What are you experiencing with AI swarms these days? And kind of give us an explanation of what you mean by AI swarms. Steve Rotter (18:11.575) Yeah, sounds scary. think ultimately, people have started to look at multiple processes happening in parallel. Let's call it maybe, let's take the AI-ness out of it for a second. And we're talking about multiple parallel processes going on. And just think about the life cycle of a typical document. Let's say you translate something in, you've got a new, great new marketing. white paper that you want to put out. And it needs to go to five different things. It needs to go in your website in 20 languages. Great. That's awesome. That can happen in one parallel path. It needs to go to your sales team with a communication on, this is now available. That can go down another path. Maybe you have a large partner network and you're putting that out on your partner portal and so on. And I think most marketers can get that, right? The idea of AI Swarm, it sounds a little you know, science fiction-y, but when you really put it in terms of the fact that, I guess when I look at what happens when I do X, all of these other things need to happen and I can automate that in a way that doesn't take, you know, I'll say whatever project management tool you use, I mean, we use one and I can tell you, like every single task seems to multiply tenfold into the things downstream that people need to do. Well, We're about taking that friction out of the process. That's it. Russ Fordyce 🏆 (19:42.717) It's super interesting because I think you've hit on something I've experienced, is when you try to give an AI almost too much multi-step processes, it seems to get caught up somewhere in that process. And I find myself running parallel processes now to get the work done so that it doesn't get jammed up. And it seems to work just fine. Is that what you guys are envisioning in Swarms? Steve Rotter (20:09.315) Yeah, I think so. I mean, I think that's how a lot of organizations are now starting to think about this, which is, if I can translate one document, why can't I translate 100? OK, great. And then take that next step further. Why couldn't I have those 100 documents possibly be in a way that are translated for multiple personas? Again, getting into that level of marketing precision that the holy grail, if you will, of of marketing is that high level of personalization for your audience. And if it's the chief marketing officer that's your target, but the buying team maybe has both financial people and operational people and other roles, then why not? I mean, if you can do it that easily, why not make it possible? Russ Fordyce 🏆 (21:04.071) Yeah, and you've hit on something pretty interesting too. It's just marketing's now in this kind of real pivotal role. It's always been kind of in the middle of an organization, you know, interfacing with all the different groups, but even more so now because as marketers, we have new capabilities that we didn't have before. You know, I'm a part-time coder now. I never was before. You know, I've got a graphic designer on staff, which I didn't before. And so all these things, I can have a finance guy. Steve Rotter (21:25.518) Yeah. Russ Fordyce 🏆 (21:33.661) So it's really interesting when marketers get those capabilities. And I imagine it's true outside of marketing, too. The finance guy can now be a little bit of a marketing guy, too, or a person. It is a little dangerous. But that's when I think, really, human judgment comes back into it. And I know you guys have raised some concerns over that, too, of putting humans in the loop and making sure there's human judgment and not just data. Steve Rotter (21:45.721) dangerous. Yeah. Steve Rotter (21:59.767) Yeah, I mean, think that for us, especially in the world of localization, there are things that, and we've coached our customers on this exact methodology where we have three or four layers, if you will, starting with things that you would use AI or machine translation for without even questioning, right? Maybe it's some of your support content, maybe it's low risk content that's user manuals or things like that that you know. there's a high degree of certainty that it's going to be accurate and certainly test that and check it. ultimately, you get to a point where you're like, okay, I can just drag drop done. Those things are easy. There's another layer that I would say would be kind of where things get a little bit more nuanced where you'd want to have a human in the loop for kind of a quick review and say, okay, is this right? Check, done. We've introduced some really great capabilities that you could take a 20 page document and it would flag. These are the six things we're not 100 % sure about. Mr. Human translator, please take a look at this or please review these. So it makes the editing a lot faster, makes the editing a lot more efficient, but ultimately it does say that at some point the AI isn't sure, which is great. It's a really great useful tool. And then there's things that you probably don't even want AI doing, which would be like very short form ad copy, for example. Let's say you're creating an ad on a billboard. Well, in every language, no matter what, if you're trying to keep to nine words or less, you probably want some specialist, some language specialist that's going to capture the heart of the message rather than simply the words that might have been translated in a larger document. So that's how we kind of guide our customers to think about this. Again, we fully expect that AI is going to make a lot of that way more efficient, but we recognize that the reality is with over 100 languages, the nuance of language, the different customers and what things mean in different languages, that you want to have that highly efficient human in the loop translator. Russ Fordyce 🏆 (24:18.185) Yeah, well, let's talk about your customers for a second because you guys have built a great customer base with nearly, I think, half of the Fortune 500 now using your services. So you've seen a lot of this kind of enterprise adoption going your two years with DeepL. What are they doing differently that the small guys aren't doing that the small guys need to catch up with? Steve Rotter (24:25.677) Yeah. Steve Rotter (24:38.819) Yeah, I mean, I would say, you know, it's not necessarily small or big. Maybe I'll reframe that, Russ, where I would say that, you know, it is half the Fortune 500, but it's also 200,000 customers worldwide and millions, literally, know, hundreds of millions of people on a monthly basis coming to deepel.com using our free service. So we make that available for, you know, again, huge, huge scale and and very, very large audience. I think what it really comes down to is there are situations where words matter, right? And I'll give you one of my lasting memories from when I joined the company was with the CEO. We went on a press tour. We were in Korea and we were doing a number of press meetings and we sat down. This journalist sits down next to us and he begins the conversation. talking to our CEO, goes, I want to thank you. And it's kind of a strange way to start the conversation. You know, yeah, right, exactly. Because you know, when you're when you're meeting with journalists, sometimes you don't know, are they going to hit you with a curveball question? Are they asking about funding or, you know, IPO or anything? But no, he says, I just want to thank you. And he went on to say that a year or so prior, there was a story that broke. was a critical story in their market about Russ Fordyce 🏆 (25:42.345) I'll take it all day long, your customer thanking you. Steve Rotter (26:06.095) a Fortune 500 executive and something had happened and all his colleagues used a kind of a free translation technology from Google and they got the story wrong. And he used DeepAlt, he got it right. It was actually a changing point in his career where he said, look, if it wasn't for that, I would have just kind of followed this path. I would have published something. It wouldn't have been as accurate as we wanted. And because of that, I was recognized for essentially my accomplishments in this. So again, I think what it comes down to is when words matter, they're the right tools to use. And there are certainly things that we've built up over the years that support both big companies, Fortune 500, but also small companies of any size, even individuals that want to get their message out to other markets. Russ Fordyce 🏆 (27:01.235) Well, to me it seems like we're moving from, you know, we're moving almost away from tools and kind of to your point into systems that act for us and that are really kind of managing work. How do you see that changing the work over the next five years? Steve Rotter (27:16.943) I mean, think our vision in many ways is that, and we've just launched a series of initiatives around this idea of a borderless business. So what does a borderless business mean? Well, today there are constraints. There are constraints for you to reach new markets. There are constraints for you to communicate with customers in a way that they want. But there's also constraints on how, and what not only how you communicate with them, but the employees you actually hire, right? And I'll use the example of DeepL Voice in a second here, but essentially what we're saying is with those language issues being solved, you could hire the best of the best, no matter where they are in the world, right? So imagine a software company and you wanted to hire some very specific talent in a part of the world that you didn't have that language expertise. We wanna make that borderless business possible and you and I could be having this conversation. You could be speaking English. I could be speaking German and we would know the difference. We would be able to communicate with confidence, with clarity. This is something we're pretty excited about. Again, I mentioned the DeepL Voice product. It's one of those where when you show it to people, it's kind of one of those wow moments where you just look at it go, wow, this is actually like, you know, science fiction coming true. because you're actually literally translating those conversations in multiple languages, not just English, German, but we could have five people on this call all speaking different languages, getting near real time accurate translation of what's going on. So, yeah. Russ Fordyce 🏆 (28:58.994) So tell us a little bit about DeepL Voice to give us an overview of how does it interact with my systems? What's that kind of look like? Steve Rotter (29:05.934) Yeah Sure, so we've built it to kind of three different use cases, the most common ones in large companies, right? One is virtual meetings, right? So we could be on Microsoft Teams or Zoom, and we could have people from around the world contributing. just one of the things we've seen is there's a huge impact on just employees' confidence when they're not speaking in their native language, right? And we've seen this, you know, in... in throughout the world. But if you're not speaking your native language, you might be, you know, hesitant to say what's on your mind. You might be hesitant to express your ideas. You might miss critical information relative to your job. So we we've removed that with this kind of plugin where you literally would be having a multi language, multilingual conversation. I'm hearing it in English. You're hearing it in German. know, Sue is hearing it in Dutch, whatever she wants. and seeing that happen real time. So that's kind of one use case, but it's a real common one because we spend a lot of time on the virtual meetings. The second is face-to-face. you know, imagine you're a trainer. You go into a factory from one of your companies. Maybe you're a large automotive company and you're going to be training new people on the new machinery that's been installed for the new product. Well, those people might speak multiple languages and instead of, you know, maybe some kind of delayed translation or something that where they just kind of guess, pop open a DeepL app, people scan a QR code and all of a sudden, you know, wearing their AirPods, they're hearing you in that language of choice. And again, other technologies can kind of do some of this stuff. They can't do it accurately. It's really what it comes down to. And then the third one around DeepL Voice, Steve Rotter (31:02.479) is an API based. So imagine your call center and you're, you receive calls in multiple languages. That's kind of a tough business because you have to staff for all of those different languages. And sometimes that requires a tremendous amount of overlap in skills and staffing issues. Time zones, right? Exactly. All of a sudden, if that goes away, you can hire the best and brightest talent. Russ Fordyce 🏆 (31:21.064) time zones, geez. Steve Rotter (31:28.815) And now all of a sudden you're serving your customers in any language they choose, in any geography, in any time, and it doesn't cost you a fortune to staff and manage that. So again, this is one of the areas we're really excited about. think when we think about the progression here from text, documents, API translation as being the foundation, now the next level for us is going to be things like voice and those kinds of real-time interactions. Russ Fordyce 🏆 (31:57.298) So from a bigger perspective, I let you go, how do you think this changes that kind of marketing role? As a practitioner, you're out there, you've been doing this for years, how does this evolve? Steve Rotter (32:15.085) Yeah. I mean, I think in many ways, it's about speed and time to market, right? Because in many ways, marketers are constantly under pressure to deliver. There's no doubt about that. They're constantly under pressure to do more with less, right? And if a, I mean, you'd be shocked. mean, the translation market alone is an $80 billion market. It's an $80 billion market. It's one of the largest software and services TAMs in the world. Some of that is software. A lot of it is kind of old outsourcing services where you're dealing with tens, if not hundreds of human translators that are basically become this, I guess I'll say this service arm to translate things. It's slow, it's incredibly expensive. If you could take that and compress it in terms of time and cost, that's a CMO's dream. Because now all of a sudden, there's more money I can spend on campaign dollars. There's more resources available to get to the customers that I want to communicate with. And I'm not spending it on what I'll call language infrastructure. I'm using it to promote my product. So it's a huge advantage in the world of marketing. And again, I come back to that first point, which is time. mean, you know, there's... There's never been a CMO where you get the message, know, hey, we're launching this product, but take your time, don't worry, just get it out whenever you feel like it. Yeah, no problem. It's that time pressure. So if I can compress the time and say, I'm gonna not only ship the products and the launch and the campaign in our core languages, let's say it's two or three languages, I'm gonna do it in 50 languages simultaneously, that's a. Russ Fordyce 🏆 (33:47.481) Take a couple of weeks off. Steve Rotter (34:07.385) That's absolutely game changing for that business. Russ Fordyce 🏆 (34:10.75) Well, it certainly allows us to kind of not only work faster, but really be doing something different, right? I mean, we're kind of, taking that stuff that was super expensive, because I, God, I remember paying translators a lot of money in the nineties and ohs. And, you know, just taking that out, it really, I think you said it the best when it really does make it borderless. You're not, you're, Steve Rotter (34:17.721) Yeah. Russ Fordyce 🏆 (34:36.798) I never really thought about it, but language is probably one of the biggest limitations in marketing that we have. Steve Rotter (34:42.287) Yeah, it really is. mean, again, doing it well, right? I mean, that's the key is nothing will reduce or jeopardize your presence in a new market by not speaking the language, by not really knowing your customers. And that's kind Marketing 101. But if you think about there's plenty of those funny, if not horror stories of bad translations in other languages, failed marketing campaigns over and over, and then you become not, you it's the opposite of what you wanted to achieve. You no longer achieve that growth. You've actually eroded trust. And then all of a sudden you're just the foreign company trying to, you know, penetrate a new market. It's a very, very big deal for organizations to get this right. Russ Fordyce 🏆 (35:34.909) Well, Steve, fascinating conversation. I am now going to go spend some time thinking about how we make the business intelligence group borderless, maybe with some deep L help. Well, it is funny because we do one request we have every year is we run the Best Places to Work awards. Steve Rotter (35:44.257) I was going to say, great opportunity. Absolutely happy to. Russ Fordyce 🏆 (35:56.681) As you said, not everybody is English speaking. We largely deal with US audiences, but when you're doing even those surveys, employee surveys, we get asked most of the time for Spanish, but I can imagine if we go to other markets, it's going to be every other language. Steve Rotter (36:12.853) Absolutely. especially, mean, as soon as you go into either Europe or Asia, the number expands very, very quickly. Right. And I mean, but even in North America, you could say, well, you know, we're we're not only English and Spanish, but French, Canadian or, you know, there's Brazilian Portuguese or, you know, what have you. mean, it doesn't that North American, but like certainly in the Americas. you would see that very, quickly. again, yeah, reach new markets. think you're onto something there, Russ. Russ Fordyce 🏆 (36:49.701) Super fascinating. I definitely want to try out the DeepL Voice. That is super interesting to me because that has been, I think that has been almost that kind of holy grail of translation for a long time. And I have not seen it done well. I do remember seeing it impacted me a lot when I started doing, I think there's a company, Hey Jen, that will take your video and put your lips in different language. And you see that and you're just, and you're like, Steve Rotter (36:55.413) yeah. Steve Rotter (37:09.23) Yeah. Steve Rotter (37:13.101) Yeah, it's pretty wild. Russ Fordyce 🏆 (37:17.458) And your mind kind of opens up to what's possible. We'll all just have to follow that Mr. Beast method that he did with Span. Steve Rotter (37:25.847) Yeah. And just a short plug, because I think in many ways, like I said, there's technologies out there that do this in a world where the content is prerecorded and you're translating and dubbing and things like that. That's one thing, right? Because that is a level of the challenge, technical challenge that is certainly, it's not... It's not meaningless, but it's not nearly as hard as real time. Real time, what ends up happening is you're translating as the words are being spoken, you're predicting what is likely next. But one of the key things when you're doing that is the rate of flickering or delays, the latency has a tremendous impact on people's ability to understand. And at some point, if that technology is too slow, let's say I'm you know, like some of those old translated, you know, movies from the 50s, you would see my, you know, my lips move and then you'd hear the audio a couple of minutes later. And it gets to the point where it's actually so distracting that you can't pay attention to either. That's why when you think about translating real-time voice, the problems we're solving not only are the accuracy of the words eventually, but it's doing it incredibly fast so that the comprehension is there. It's actually pretty magical. Anyway, I just wanted to share that because I think, you know, again, when you really think about what's going on, technology under the hood, it's actually pretty impressive. Russ Fordyce 🏆 (39:02.643) Well, like you said, when you can read the whole document or have the whole video, understand the context. That's very different than a stream. And you're 100 % right on the delay. I came out of the voice business back in the day when voice and internet merged. And it was something that we tackled all the time, which was delay. I think the human ear, it's anything after a half a second, starts to notice the delay and notices. Steve Rotter (39:09.316) Yeah. Steve Rotter (39:19.004) yeah. Russ Fordyce 🏆 (39:31.941) starts to impact the conversation. So you guys got to be right. Steve Rotter (39:35.011) Yeah, yeah, that's exactly it, Russ. Russ Fordyce 🏆 (39:39.197) Well, Steve, thank you so much for taking this time with us. It was super interesting to hear about DeepL. And again, congratulations on the AI award. I can't imagine a better candidate than somebody that's breaking down all the barriers between all us humans. Steve Rotter (39:41.965) Yeah, it's been great. Steve Rotter (39:52.505) Fantastic. Well, Russ, again, thank you and it's been great chatting. Have a great day. Russ Fordyce 🏆 (39:58.344) You too.