| Robert |
Hello and welcome to Globally Speaking. I’m Robert Jelenic and I’m the marketing director at RWS Language Services. Today I get to welcome Jim Compton, my own college, onto the show. Jim gave a really interesting presentation at LocWorld43 in January of this year 2021 and the topic was the locale of 1. And specifically the implications of hyper personalization on localization. I thought it was a really great presentation; it caught my imagination, and I asked him to come here today so we could maybe dig a little bit deeper into it and have some fun. So I hope you enjoy it. |
| Robert |
Jim, before we dive into our crazy topic today, would you take a second to introduce yourself to the audience? |
| Jim |
Thanks for inviting me on the podcast. I’m Jim Compton. I am a technology partner manager at RWS and I’m super interested in just seeing the evolution of the language industry and how it intersects with bigger technology trends or just bigger global world trends, mega trends. |
| Robert |
Great, thanks. So it’s gonna get kinda crazy today. So, I saw your presentation on the recent LocWorld on this idea of hyper-personalization and the market of one or the market of all. It really captivated my imagination. And I think we could probably take this in a really fun direction today.
So for those who weren’t at the LocWorld session, maybe you could give us a quick synopsis of what, like, the idea you presented? |
| Jim |
Yeah. So speaking of these, like, mega trends, one of these trends that was really innovative at the time and fueled by the capabilities of the internet was this idea of hyper-personalization. Which, like, back in the ‘90s, it would seem a radical concept that you, if you’re a marketing person or a sales person, could sell to a huge audience in a one-on-one manner. That’s what they mean by hyper-personalized, so target market of one. We see that today all the time, right? Like when you’re doing some shopping on a website and it’s like, “Hey, you bought this a month ago. We think you might be interested in this also.”
Or, “Hey, you’re probably running out of filters for your coffeemaker; here, we can send these to you.” You would almost think of it as a personal shopper experience. Right? And that in the past was an extremely unsalable thing to be able to do. But with technology, like, we’ve seen this, right? Every person can get that kind of, like, experience. So that’s what’s meant by hyper-personalization. And I’m interested in the intersection between that and localization or the practice of localization. And maybe this is the big question: how would you bring hyper-personalization to the table in a way that it could equally serve every single person on the planet, regardless of their geography, their language that they wanted to operate in and their culture?
So that was the basic thrust. It was meant to really ask questions and have people thinking about this. Because I’m always thinking, how can we evolve the localization industry to be more useful and valuable? |
| Robert |
When I saw that I was sort of thinking, I mean, marketing 101 is segmentation, so trying to cut up the population of, you know, whole world or companies or whatever into some meaningful groupings. And this is sort of taking it to its extreme and essentially having 7.5 billion markets of one on the face of the Earth.
Now, you need some engine that can serve 7.5 billion markets, right?
And a big part of that, obviously, is language and a big part of that is culture and a big part of that is usability and a big part of that is customs and payments. And, I mean, the list goes on. So I mean, it is a mammoth idea to undertake. |
| Jim |
This is true in product design also, right? Amazon Echo is a really good example, because, I think, a digital assistant like this really is a perfect example of what a hyper-personalization engine is supposed to do.
But they’re still quite segregated by locale, right? Like you’re in Germany, like you have a German version. And it’s a different universe than if you were to come to the US and use the US version, right? They’re literally different things. And, you know, you could say, “Yeah, that’s fine.” But if you’re trying to create, like, an equal, like, level of capability for everyone around the world, you probably wouldn’t wanna design it that way. You would wanna make it instead to where you could be anywhere on the planet and the product would sort of work in exactly the same way.
So, it’s an interesting challenge. Like, how would you get there? Do we actually want to get there? Right? I’ve come to this as a real kind of, like, technology optimist, right? Like the way everything should work would be these intelligent assistants that sort of permeate products and communication and everything around us, just sort of know what language we wanna be speaking in that minute, even to the nuance of, like, do I wanna be talked to in kind of a cheeky way? Or do I want this person to be serious with me? |
| Robert |
I was thinking about this the other day. And I have kids. My son is almost four and we’re attempting to raise him bilingual. So I speak to him in English and his mother in German. So probably a primitive recommendation engine would spit things to him in German, because it seems, you know, he’s got a German IP address and he’s logged into German versions of these e-commerce platforms and all.
But I want him to learn English as well. So, you know, a primitive engine would probably not make that distinction, or would rely on user preferences and so no. It’s almost like you’d have to feed that engine tons of data so that they understand actually my goal as a parent, and hopefully his goal as a child, is to learn a second language. And so it, like, you actually can’t get there with a primitive engine; you need to almost be hyper-personalized in that context. Because in this context, we have a child who probably will want to speak one language over the other, out of self preference. Or, you know, parents have different hopes for what they’re doing. So basically, the engine needs a deep profile of that human being to make the experience, I guess both in the kid’s case, for the kid and for the parents, the optimal one. Which is… |
| Jim |
Yeah, yeah, no.
The dealing with, like, multi-lingual households I think is, in and of itself, something that is a huge barrier if you take the approach of that there are such things as different locales. Right? Like I mean, you almost have to throw away the concept of the locale or this idea of, like, source and target language, right? And say it’s all just permutations of human language, right? With different expressions. |
| Robert |
Well, well, is it? Because I’m just wondering, when you say get rid of the idea of the source language, then I don’t know if you’ve heard of this example where two AIs, I can’t remember who built them, but they were speaking to each other early on. And then they learned from each other how to use the language. It might’ve been English, but it might’ve been something else. And then the researchers looked at what they were saying to each other after a day or a week or something and it was gibberish to a human, but it was actually a far more efficient, much richer language than what the source language started. So I just wonder if you kind of, you know, you get away from the idea that this needs to be a language that’s understandable by humans. Maybe it needs to be something that can actually take a humanly understandable form in the moment of consumption. That’s the part of your presentation that really, like, got me. Because, like, yeah, this sort of lingua franca, the true lingua franca, the recommendation engine, why does it have to be English or any language at all, actually? Maybe it’s just bits and bites or symbols or something more abstract. |
| Jim |
This is a really interesting question, right? And we are gonna take a little bit of a detour but I think it’s relevant. Okay, if you look at, like, one of the big revolutions in publishing, like electronic publishing, it was this idea of, like, component content management, or separating content from form, right?
And that’s enabled all sorts of efficient, you know, delivery of content that you wouldn’t be able to do, right? But if you think about it, some written piece of, let’s say English text, right? That’s still form. Right? It’s not content at all. So, how could you, I mean, and it’s an interesting question, basically, crack open the text and decouple English text from the content and the form. And you have to ask the question, what is the content? And it’s kind of more like the idea or the feeling or something like that. Right? Or the intention or something like that.
And I’m really interested in, like, what Elon Musk is doing with his Neuralink company. Right? This idea that you have some literally capturing of a waveform, right? And that waveform, you know, with some kind of mathematical precision, represents what the idea is, right? Maybe that’s the language, right? Like the full language of human, collection of all possible feelings and, or intentions within the human experience. And then everything else on top of that is just some kind of, you know, presentation layer, right? That would include the language, or is it in writing? Or, like, is it cheeky? |
| Robert |
Is it even a language? Maybe it’s just a delivery to your doorstep? Who says it’s content, per se? It could just be some sort of action or experience. |
| Jim |
Yeah. Or outcome. I think it’s interesting, because, and this is just with, like, as AI gets to be more embedded in the tools and stuff we use. I mean some of the conversation goes away and it’s just, like, delivering an outcome to you. And that can get kind of uncanny. Where you’re like, yes I did run out of coffee filters and they, you know, they just showed up on my doorstep, right? There’s not even a question, do you want these? It’s just, like all of a sudden, your wishes are fulfilled. I don’t know. I like that model. It sounds very Utopian. |
| Robert |
Or dystopian. |
| Jim |
Yeah, exactly. It sounds like it might have some potential for having the opposite effect. I mean, part of it is, like, oh, crap, I got what I wanted, right? Like there’s almost a terrifying element of you know, be careful of what you ask for element that if you’re too well served by these things, that the real risk is that you get exactly what you want. If that makes sense. |
| Robert |
I was watching some podcasts with Elon Musk on it, back to your point of Neuralink. And I was intrigued by the way he put it. The bottleneck is bandwidth between us and devices, right? So for me to interact with my smartphone, which is more powerful than any computer ever on the face of the Earth in, like, the ‘70s. My interface is these two sausages on my hand called my thumbs and that’s how, you know, that is the upper limit of the bandwidth between me and my device. Now we’re just getting into, you know, voice and things like that. But really, I think the idea behind Neuralink is to open that floodgate. Is to basically link you right to the device. And just imagine, if you could peer under that hood and see, what does that language look like?
It’d probably look a lot more like waveforms and hexadecimal than it would look like anything you could actually tell, right? It’s just the implementation of that or the manifestation of that would be something that a human experiences. |
| Jim |
So what’s so interesting about that is so much of what we can understand from AI is based on what we have words for, right? There’s some kind of statistical correlation between this thing and this thing and we have a word for what that correlation is, right? Or that concept, like, acceleration. And because we have a word for it, then it sort of makes sense to us. But with just pure finding these correlations, like there was a podcast that I had heard where there was this device, I think called Eureka, which was a room that you would put objects in and start, like, modeling their physical behavior. And it would spit out, basically, these formulas that sort of can predict different attributes, right? If this thing’s mass is this. And I don’t think it was, like, probably pure, classic physics, but it was able to punch out some famous classic physics formulas. And then all sorts of formulas that the scientists had no idea what even it meant, because there was no word for the concept.
So you think about that, like, our world, the amount of words that we have, even in all of our collective human language, is, like, just scratching the surface about what actually is existing in reality, that could be, let’s say, predicted.
And if you remove the need for the language, you can do things, like predict things and no one knows how it’s actually, I mean the actual humans in the mix who brought the AI to the table have no idea how that prediction was even made or why it’s true. It’s a super interesting, potentially scary thing. It reminds me a bit of the series Westworld, where you have these kind of programmed people who have this sense of independence. And then finding the script of the thing that they just read, you know, realizing, no, they’re actually programmed. |
| Robert |
It raises the idea of free will, right? Because, I mean, if these things get so good and language-independent, but I mean, certainly it’s interesting with the language component. How much of the randomness of life do you still get to experience if you’re constantly nudged in one direction? So for example, I’ll go back to my son’s example. So maybe an ecommerce platform would find that he is more likely to buy something in German language. But I’m trying to nudge him to speak English more. So when do the commercial priorities start to actually nudge you away from other factors in your life that you may want to take in other directions? Be it his father trying to get him to learn another language. Or you just bumbling around an e-commerce shop or walking through a store and just stumbling upon things.
I wonder, as we go to hyper-personalization, where are we going with that? What do you think? |
| Jim |
That’s a great question. ‘Cause in the case of a company trying to sell to you, it’s still selling, right? Like, but it’s this interesting thing where we become such prolific users of the technology that it’s no longer about us, like, buying things. It has to do with our sort of human experience. And there is, you know, potentially a conflict of interest between what you wanna do as a person or what you wanna do as a parent, right? And what a company wants to do who’s trying to, you know, maximize what they’re able to sell to you.
So, I think for sure, it’s probably true, people have just become more consumers in general, thanks to how fantastic of an experience it is to buy things now, but yeah. What if it’s in your best interest to not be served directly in the way that you want, right? Or, like you say, you want, or even would benefit from, some different experience that’s maybe not even like the optimal… |
| Robert |
Exactly. |
| Jim |
… experience. |
| Robert |
Not optimal experience.
Maybe there’s utility in that for me. I wonder I’m being denied that utility, if I’m being nudged in directions so subtly and so efficiently that I almost didn’t even realize it. |
| Jim |
Yeah. It brings up this question of also identity is very much based on your experience. Right? And if your experience is being reinforced and supported, right, it’s in a way kind of making your identity, pulling you along some predetermined identity. And I do think that’s a really important thing for people to even disrupt themselves every once in a while, right? And go do things that they wouldn’t otherwise do, just to be more well-rounded, right? I mean that is one of the things, I mean, there’s a self interest in the engine that’s trying to sell you something, to have you be as simple as possible, right? Like it’s going to be easier to serve you if you are simpler. |
| Robert |
For now, right? If you take that into its limit, I mean, technology’s looking to get better, they’re going to have a lot more data feeding these things. And at some point, they may even care how nuanced it needs to be to serve you well. It can serve the most sophisticated individual just as well as the most simple individual. |
| Jim |
The most sophisticated person that’s just changes their mind all the time is hard to predict. Yeah, you’re probably right. I feel like the language part of this is a really interesting element of this, right? And, like you say, maybe it becomes less important as it’s just getting not to having a conversation with you, but trying to serve you, something like that. But there is also an element of, I think people really do now have conversations with these systems, right? I mean, very literally, in the case of like your digital home assistants, it’s designed to have conversations with you, to be funny and yeah. Watching that play out, like, I’m sure you’ve seen this, I don’t know if your son is old enough to be interacting with the assistants at all. But, like, as a parent, watching that happen is really interesting. ‘Cause it’s actually part of the teaching of your kids how to interact with other people, maybe? I’m not sure on that. |
| Robert |
It’s almost like education. I mean, they’re learning how to interact with something.
So in so far as it can recognize which person’s talking to it, it’s already creating a profile. You know, which language does this kid use? Or what kind of stuff does he or she like to order? You know, so, I mean, early on it’s, these profiles are being built. |
| Jim |
And you know, this is really interesting. ‘Cause, and you were talking about would there be more of a universal language that isn’t anchored at all on any, like, traditional human language. It’s like its own language. I feel like that’s a real trend, right? I mean, certainly, my kids, when they’re interacting with, like, the Echo, they’re speaking to it in their own slang, right? Which is super interesting. And being very cheeky with it, incidentally also, right? Like, trying to almost, like, trick it.
Which, as being a thing that’s training the engine, right, like, you can see it that having really a strange form at the end, where you can really only interact with it most effectively if you’re being, you know, sassy if enough people are doing that kind of training. But I’ve also noticed that just in, like, so you know Roblox, the game Roblox.
It just went public, some massive thing. My kids play that all the time. And it has a super interesting idea of, like, it’s a culture that’s creating its own language. Like I’m definitely seeing that. First of all, there’s different sort of, like, linguistic servers. Like, you can go on a server and it’s like, you know, mostly German kids or mostly Spanish kids or Chinese kids and what I’ve seen is all the kids kind of don’t necessarily try to stick to something that’s their language. Like, at least my kids, they’ll be, like, on the Spanish server. And they’re figuring out how to communicate without any, like, real training in one another’s language.
But they’re using their own language in the system. So there is actually like a sort of like a new internet language, like, arising. At least on this system. Part of it, also, is because the system tries to police the language quite a bit. I mean it’s a bunch of kids, so it’s trying to protect, you know, but in doing that, having to bypass the, these sort of language filters, this just entire new language has arisen. And it’s really interesting to watch.
And it also makes me think, like, all these systems being trained with, like, “natural human language.” Right? What are those gonna look like in a few years when the language is now this kind of like hybrid language influenced by Roblox, right? Like I have no doubt, well, maybe, I don’t know. It would be interesting to see if my kids use that Roblox language outside the context of Roblox as they’re interacting in the business world or things like this later on. I mean, I wouldn’t be surprised. I mean, it actually, you know, becomes part of the language.
And I know, I took this completely off on a tangent with hyper-personalization. |
| Robert |
My brain just went, you know the movie Arrival, right? Where, like, the aliens talk in that, like, ring language. I looked this up the other day and it’s like little flares off the rings and they have meanings. And I think the actor’s a linguist in the movie and she kind of puts it all together. I mean, it’s almost like why should the storage and delivery and synthesis of ideas be coupled to any one particular human language? They’re probably all far less efficient than something else that, you know, anything I would come up with, or kids playing a game, right? |
| Jim |
The thing I also love about that movie is this hypothesis that, like, reality is defined by the language, right? And if you’re not able to time travel, maybe it’s just ‘cause, you know, your linguistics schema doesn’t support that concept of time travel. And if you have a language sophisticated enough it’s like, oh yeah, of course. But it’s kinda true, right? Because things become real when there’s a word for them, right? I mean, in my own personal life, I’ve found this true. Something that’s, like, a mystery, something that you maybe don’t even have any power over. The minute and you’re like, oh, there’s a word for this. It’s like this sort of exponential level of power that you have in the universe as a person, right?
Like, ‘cause now there’s a concept. But I like that extrapolation from Arrival a lot. Where it’s like, yeah, you can actually just, like, have science fiction level capabilities just by having a word or the flare. I love the idea of the flare in the… I didn’t know that, by the way, those, like, coffee rings sentences that they would send out. That was such a, so great. |
| Robert |
So back to hyper-personalization. I think we went way, way off. That was fun, I had fun. I mean, what would have to happen? Like, specifically from a localization language point of view, in order to weave that into an experience that starts to get close to 7.5 billion markets of one.
I mean, you mentioned before this paradigm of, you know, target and the source and targets. My brain’s thinking, going to the place that, like, maybe that’s anchored in the past and maybe there’s something else in the future. |
| Jim |
Absolutely. I mean and this is just true of, I think, localization in general. And I’ve been beating this drum for, like, years. It can’t be an activity after you do your product design or your message design or these things. It really needs to be baked in, embedded into your product or your campaign or your business, frankly, right? And I think part of this would be, yeah. ‘Cause I, ‘cause you can see how this, and this was part of my LocWorld presentation, just showing some visual examples how, like, if you’re gonna build a model for like some hyper-personalization engine that captures all the experiences you want to provide, captures all the attributes that might influence that experience and all the outcomes, I guess, that you wanna be able to serve with your engine. And if you think that, like, language and culture and geography are a factor in being able to, maybe, like, capture what someone wants, or sort of, you know profile them in culture time space, right? This idea of taking everything that I might wanna say or hear or not wanna hear, every cultural experience, for me, in Denver, right? And then take that thing that you designed and then go, like, quote unquote, translate it. I just think it will never work that way, right? Because you have experiences that I have, that aren’t relevant to someone else in the world at all. There are experiences that they’re gonna have that you’re never gonna capture through me.
So I think you have to start with the assertion, I guess, that localization as a paradigm, being this idea of taking a source and turning it into target will never get us there, to the idea of, like, target market of one. I think you have to start with treating language and culture as something that you model in the actual engine itself, or the business itself, right? It has to start there. And, linguistically, I think that this is something that my daughter brought up, which I thought was a really interesting question and made me think, maybe, is this kind of how you can do something like this?
She’s like, “Is there such a thing, dad, as a dictionary with every word in every language?” Right? It’s like, no. But can you imagine? Why? Why not? Right? Why not like a dictionary of all of human language or all human concepts, right? The challenge with having that as an actual book is we don’t really know how to encode ideas without some form.
Like, you look at the internet even. The, internet, which should be, by definition, this language agnostic thing. It’s written in English, right? I mean if you go look at like HTML, you look at source, it’s like, I mean even, you know, metadata. You would think, oh, metadata can bring everyone to, like, what they’re looking for. Metadata is all encoded in English, right?
So, you really do have today and if you go to, like, Google and search for the word, I wrote an article about this at one point. Search for the word kitten in, like, six different languages, and you’ll see that the kittens that are returned are different kittens. Even though I think the concept of a kitten is the same, right? It, I mean everywhere; that goes to show you have an internet that’s right now segregated by language. And it’s still true with businesses. Businesses are, I mean this is where localization is helping to bridge the gap. But it’s also, like, it’s also too late. Like, by the time you’ve designed something that works, you know, in the context of, let’s say Silicon Valley, right? It’s not gonna work everywhere in the world. You need to design it in the context of the world. |
| Jim |
And yeah, what if, like, maybe you do have something like a dictionary that has every word in it. Every concept in it. And how would you write it? Well, maybe this is where, like, the waveform of what feeling that has, or the waveform of what intention it is, that becomes the main thing that you search for and then everything else just sort of anchors on that. I mean, I don’t think it would be a real physical book, but that would be pretty cool if you could figure out how to do something like that. |
| Robert |
… it almost seems like the whole source, the target paradigm, is a bit analogous to the two sausages on my thumb as the bottleneck of the bandwidth is, like, I think as soon as you are in that position, you can’t be hyper-personalized, or pivoting to that is gonna be unnatural. It’s gonna be sort of, like, forced and pushing water uphill. |
| Jim |
Yeah, I mean, I think you will have, in the Venn Diagram of concepts that were relevant to me but not relevant to someone else culturally and concepts that I don’t have that are relevant to someone else that you’re missing. There is something kind of middle there. Which is, like, this sort of happy accident of, like, a hyper-personalization being localized, but it’s not very efficient, right? I mean it is, like you say, it’s the sausages on the screen.
And I think what would probably happen is the people for whom something was designed at source are always gonna have a better experience than the people for whom that experience was quote unquote localized or translated. It’s just gonna be worse, I guess, for them, by definition. |
| Robert |
Right. So the answer is 7.5 billion sources and 0 targets? |
| Jim |
There you go. Exactly. 7.5 billion sources, 0 targets.
Yes. That’s exactly it. |
| Robert |
Jim, I’m looking forward to your next presentations at LocWorld or wherever they might be. That was a great talk today. Thanks a lot for joining. |
| Jim |
Thank you, Robert. I appreciate it. |
