Alex Elias, CEO & Founder of Qloo, joined us on the SourceForge Podcast to discuss AI and how AI can be used to predict and decode consumer taste.
Qloo is a taste AI company that offers a unique and privacy-centric solution for understanding consumer taste across various industries, and discusses the evolution of Qloo and the AI industry.
Watch the podcast here:
Listen to audio only here:
Interested in appearing on the SourceForge Podcast? Contact us here.
Show Notes
Takeaways
- Qloo is a cultural AI company that predicts global consumer preferences and catalogs cultural entities.
- The AI industry has evolved over the years, and Qloo has adapted to the changing landscape.
- Alex Elias’ background in privacy influenced Qloo’s focus on protecting user data.
- Qloo faced challenges with its initial business model and pivoted towards a B2B solution.
- AI technologies, such as language models, have been instrumental in enhancing Qloo’s API.
- AI is transforming knowledge work by enabling more efficient and creative thinking.
- Qloo transitioned from a B2C company to a B2B solution through the acquisition of TasteDive.
- Qloo offers a unique and privacy-centric solution for understanding consumer taste across various industries.
- Their proprietary corpus and ability to refine taste associations give them a competitive advantage.
- Qloo’s clients include global brands in industries such as financial services, travel, automotive, and consumer services.
- They help companies overcome challenges in understanding consumer taste and make informed decisions.
- Qloo’s API is easy to use and requires minimal data input, making it accessible to a wide range of businesses.
- They have competitors within specific domains but offer a comprehensive solution for enriching taste across categories.
- Qloo is constantly innovating and will soon launch a self-serve research tool and on-device learning capabilities.
- The future of the industry is moving towards on-device processing and differential privacy.
Qloo’s solutions can benefit companies in marketing, sales, content creation, and more.
Chapters
00:00 – Introduction and Overview of Qloo
03:31 – Evolution of Qloo and the AI Industry
07:00 – Alex Elias’ Background and the Privacy Focus
09:19 – Qloo’s Business Model and Challenges
14:40 – Utilizing AI Technologies in Qloo’s API
26:11 – Transition from B2C to B2B
31:03 – Qloo’s Competitive Advantage
32:11 – Industries that Benefit from Qloo
33:10 – Financial Services and Travel
34:05 – Automotive and Consumer Services
35:33 – Understanding Consumer Taste
39:29 – Challenges Qloo Solves
43:07 – Communicating Qloo’s API
45:40 – Competitors and Alternatives
48:10 – New Features and Announcements
52:32 – On-Device Processing and the Future of the Industry
54:48 – Qloo’s Solutions in a Nutshell
55:42 – One Big Thing for Potential Customers
56:20 – Getting in Touch with Qloo
Transcript
Beau HD (00:01.308)
All right, hello everyone and welcome to slash.media’s SourceForge podcast. Thank you all for joining us today. I’m your host, Beau Hamilton, senior editor and multimedia producer here at slash.and SourceForge, the world’s most visited software comparison site where B2B software buyers compare and find business software solutions. Joining us today is Alex Elias, the founder and CEO of Qloo, a really fascinating cultural AI company that is decoding and predicting consumer tastes across the globe.
Alex, welcome and thanks for joining us here with our first podcast.
Alex Elias (00:35.43)
Yeah, thanks for having me, Beau. And it’s an honor, an honor to be your first guest and a huge fan of SourceForge and everything you guys do. So appreciate it.
Beau HD (00:44.2)
Awesome, yeah, I’m excited. So I just want to jump right into this. Let’s start fairly high level. I see on your website, Qloo.com. That’s Q-L-O-O.com for those listening. Your company offers a privacy-first API that predicts global consumer preferences and catalogs hundreds of millions of cultural entities. That sounds incredible, but could you kind of just tell us what that all means and just give us an overview of what it is that Qloo does?
Alex Elias (01:13.086)
Yeah, absolutely. So at a very high level, if you think about all the ways in which, you know, tastes are currently being predicted and cataloged, it’s usually part and parcel or a small part of a large consumer service. You think about Spotify and the world of music. You think about Netflix and the world of film and TV. They’ve actually come around and become a customer of ours. But then there’s all these other areas like OTA with travel, Expedia, etc.
And one of the challenges over the last decade plus that we’ve been operating and when we started building the company is that a lot of that exists in these silos and developers, if they want to kind of build in an understanding about taste, particularly in this new AI era, it’s very hard to come by that sort of data while respecting privacy. And so what, what Qloo has built is essentially this massive kind of globe spanning database of cultural entities with a very kind of
deep conceptual understanding of what exists. So think all the music artists globally, every restaurant and hotel around the world, think of massive kind of catalog and database that we build and normalize. And then on top of that, we have a lot of AI around generating kind of inferences about taste, given totally anonymized input context. So…
That’s a mouthful, but at a very high level, what Qloo’s built is what we believe is the most nuanced understanding of sort of consumer preference and taste across these areas that matter. You know, the traditional media categories all the way over to travel, consumer brands. And so what part of what makes us unique is just the breadth, the fact that we have built this in a systematic kind of normalized way that
can generate predictions across all these different categories and the fact that we fundamentally empower developers to kind of build products on top of our stack. So super excited to be here. We’ve been building this a long time. We just recently had an article that came out that was the first one that sort of name-checked our age and said how a 12-year-old AI company did this and that. So yeah, we’re thrilled, and we’re in it for the long haul.
Beau HD (03:31.624)
Right on. Okay, yeah, 12 years old. Yeah, you guys are the old group in the, I feel like there’s a little bit of a lot of AI startups, right? So you guys are kind of like the old guys in the room.
Alex Elias (03:42.186)
Right. Yeah, it’s not long in human years, but certainly in software and AI years. It’s amazing.
Beau HD (03:50.952)
Okay, so you guys are essentially taking large data sets and sort of connecting the dots to find meaningful connections and solutions. That’s really interesting. Now, I understand you have a legal background, having completed your JD at NYU Law School, and you mentioned you had to focus on internet privacy. Could you kind of walk us through your background a little bit and how it ultimately led to Qloo?
Alex Elias (04:00.386)
Yep.
Alex Elias (04:17.05)
Yeah, so in many ways, Qloo began even before law school, the sort of the Netflix prize competition, a lot of excitement around building recommendation engines and the late aughts. That’s something that caught my attention. I’m a huge kind of culture file. I play the saxophone and the piano and I love mid-century Italian cinema. We lived in New York City for 17 years, obviously loved indulging in the culinary culture there.
So there was just this kind of intuitive sense that all these different areas were not in fact kind of disparate, but there was deep inner relationships between different cultural domains that was worth exploring and mapping. And much of what was being done was incredibly siloed and niche. You had companies out of MIT Media Lab like Echo Nest that were purely focused on music without any concern for…
how that might interact with film and TV and literature. So those efforts were obviously noble, but began thinking about this problem space really around that time, was in law school, sort of found myself studying and going really deep on internet privacy. So this was kind of 2009 and 10, a decade before GDPR even did some substantial kind of writing around sort of the consumer privacy bill of rights at that time.
I worked on empirical studies on the efficacy of click wrap agreements. You know, whenever you implicitly consent to software T’s and C’s and, you know, no surprise the efficacy is incredibly low in terms of actually garnering consumer consent. So these were things that were percolating in the background. That’s largely what I focused on in law school. And actually we started Qloo towards the end of my law school career. So we raised capital while I was finishing my JD.
ended up completing it, but really that was part of, you know, in 2011, 12, that became part of the background of how we structured our service. And, you know, we basically wanted to find a way to be able to convey meaningful information, meaningful inference about taste without sort of having to basically deal with, you know, privacy issues and deal with identity-based data. So there was kind of a fundamentally different paradigm where you could…
Alex Elias (06:40.418)
focus on the entities themselves and how films and how the characteristics of films and other cultural categories intersect. And then you could actually create services that ultimately have an understanding of taste without an understanding of identity, which was how that all kind of came about. So yeah, sort of played on different strengths.
Beau HD (07:00.1)
Interesting. Yeah. I love that you have a background in privacy, because I feel like that’s an issue that’s so neglected, you know, nowadays, especially in the tech industry. Yeah. And I want to circle, we’re going to circle back around to the privacy. I’m, I’m curious. So you mentioned you, your company was founded in 2012. Obviously a lot has changed in the AI industry since then. I know consumers, especially in businesses as well are certainly much more aware of AI.
Alex Elias (07:09.066)
Right. Totally.
Beau HD (07:29.34)
than they were 10 years ago, let alone 12 years ago. Could you talk a little bit more about how Qloo has changed over the years since when it was founded?
Alex Elias (07:38.734)
Yeah, absolutely. We’ve had some dramatic changes over the years, never in our core focus. So we’ve always been focused on this very specific problem space. Our initial hypothesis was that this would be super interesting for consumers. So the way we initially conceived of the service, it was, let’s move away from the kind of legacy UGC, user generated content properties like…
the time and they still persist, you know, IMDB, Rotten Tomatoes, Goodreads, Yelp, Trip Advisor. You had a lot of these siloed UGC properties, many of which have been rolled up by Amazon. Many people don’t realize that. But Goodreads, IMDB, now Holyon by Amazon. But you know, our hypothesis was that it would be interesting to have a service that was kind of the…
cover all these cultural domains, empowered people to sort of understand the implications of their taste, and didn’t rely on an ad centric model. So actually was ultimately only achieved success if there was a valid recommendation that led to affiliate revenue. And there was many other companies attacking this premise from various angles. I think we were one of the more kind of brash companies in that regard because we just took on all categories at once and said like,
breath is fundamental to this premise. So that turned out to not be a great business model. Affiliate revenue is incredibly slim. And even when you have success in converting, there’s no guarantee that click path and so on is what leads to it. So.
We basically ran with that model for a while. We accumulated a lot of useful, kind of anonymize, a very interesting first-party corpus of data, not only in structuring the entity intelligence themselves. So we sort of developed a rigor around mapping the genome of all the different entities, and so be it a fashion brand, a music artist, a movie. So we started developing really interesting IP just around the structured entity data, but also this kind of cross-sectional preference data that we were mapping.
Alex Elias (09:49.398)
And it wasn’t until 2015. So we ran with that model for years, you know, got into the seven figures of registered users, but never really hockey sticked as far as the UGC proposition. And then around 2015, and I think this is declassified now, but we had a very senior product manager, a kind of leader at Twitter at the time, reach out and say, Hey, I’m a huge fan of the Qloo app.
And she basically was like, we’re building something internally that’s focused on commerce. And the idea was if I could tweet about, you know, obviously now the parlance is different, but if I could, if I could tweet about a book or tweet about a restaurant or a movie, as many people do, anyone consuming that tweet could actually, uh, the entity itself would be recognized. So the, you know, the movie, the book, whatever.
and people can click on it and you have a full-fledged commerce experience with recommendations that have this whole discovery funnel. So it becomes this kind of cultural discovery engine sort of embedded within Twitter. It’s a very bold effort. We were partnering with them. We developed, started building over the course of a year, sort of very high scale. So it was really the first time we realized there’s a big model here in sort of externalizing our service as an API.
One of our early shareholders is the, you know, the loud tech voice, Jason Calacanis, who’s been with us kind of from the start. And he, he always saw it this way, you know, to his credit, he was like, this is, this has got, there’s gotta be an API that could just do this across, across a bunch of categories and empower developers. So, so yeah, so we, we ended up, so then what happened was leadership changed to Twitter.
They ended up shelving the effort in favor of kind of customer support. So like having, you know, if people complaining to United Airlines about a flight and creating SaaS around that. So they pivoted dramatically, which at the time was a pretty devastating blow to us. But the silver, because it was incredible. We were at the 11th hour of deployment. I mean, we really scaled this proposition to the moon. But the silver lining was we now had enterprise grade.
Alex Elias (12:06.274)
highly scaled sort of externalized APIs that could do awesome things. Like people could build on top of this now. So we basically took that premise and then started taking it to market. And the first few years were very challenging because there was a ton of identity data everywhere. So from 2015 to about late 2018.
We were sort of a hammer, whatever that expression is about a hammer in search of a nail or a nail in search of a hammer. I don’t know which way it is, but there was basically a lot of companies that hypothetically could benefit from our kind of probabilistic AI that could generate inference about taste. But they had…
amazing first party deterministic joins about people’s identities. So they could actually predict where Alex or Beau, what kind of auto loan we applied for using a third party service to actually broker that identity based data fundamentally without us knowing anything about it. And so everyone was selling data in that era that was actually identity based. And so it was very hard for us to go in and be like, hey, you should rely on sort of this
informed probabilistic kind of inference from our AI when they’re actually getting like direct identity joins. And then of course, you know, GDPR happened, consumers got a little smarter about consent, CCPA happened, you know, cookies started being deprecated. This is all over the course of kind of 2018 to today. And so all of a sudden there’s been a decimation.
of this kind of freewheeling identity data flying all over the place behind the scenes. And companies suddenly realized we’ve got nothing but our first party data that we have the real kind of consent frameworks around and rights to. And we have to figure out a way to be able to infer interesting things about people without betraying their identity. And so in this new landscape, Qloo’s found enormous success in terms of just enriching
Alex Elias (14:19.458)
in a totally anonymized way, just being able to enrich the kind of probability guesses around people’s taste, given some input kernel, like a small amount of sparse anonymized input data. So that’s been the full journey. And now there’s a whole new landscape ahead, which is super exciting with the AI space generally.
Beau HD (14:40.208)
Yeah, that’s really interesting. It’s, it’s crazy how much things have changed over the last decade. You know, just name dropping like Twitter, for example, is, you know, it’s like, we don’t so much as change. We don’t use that term anymore. I feel like what, when you first started, um, in 2012, were you using the term AI?
Alex Elias (14:53.153)
Yeah.
Alex Elias (14:57.57)
Totally.
Alex Elias (15:04.93)
So that’s a very interesting question because we had a lot of internal discussions around how to actually, because a lot of our job in that era was education. Like clients weren’t, a lot of them weren’t developer kind of centric and ready to onboard our kind of software. So we had to talk to business leaders, often CMOs, CFOs, you know, there was, even if we were lucky, there was a product manager to talk to, and certainly the IT leadership, but…
Um, at the time we actually decided that AI was kind of a dirty word. It was, it was something that didn’t get much traction when we would talk to customers and so we had to kind of. Break that term open and describe it in terms of the end function. So it’d be like a personalization engine or, you know, a decision like a, there would be other terms that resonated more, but in that era AI for, for up until maybe only two and a half, three years ago, um,
You know, AI was kind of a dirty, it wasn’t a dirty word. It was just a very fuzzy kind of, you know, it didn’t.
garner respect in the way that I think it does now for real, for driving real outcomes. And I think people needed a more sort of primitive or fundamental sort of description of, you know, what, what the service did. So we started really pushing on AI, I’d say around 2017, 18, like in conjunction with the regulations, it really started to…
You know, we alternated between calling it a cultural AI and a taste AI. Those are, you know, the two ways in which we’ve kind of described it. I think both are apt. And Qloo definitely has a lot of thought leadership and kind of both, you know, mind share at least in both those, those monikers. So.
Beau HD (16:55.256)
Yeah. What about, have you, um, what are some of the technologies you leverage in, in your API, do you utilize, uh, I imagine back in 2012, you, you weren’t utilizing a lot of the large language models or generative AI services, do you use those services nowadays?
Alex Elias (17:13.598)
Yeah, so in 2012, the paradigms are very different. Deep learning was very much in its infancy. A lot of the paradigms then were around kind of matrix factorization and more linear algebra oriented approaches to kind of generating inference about, you know, the connections between entities and then graphs sort of started getting more and more complex.
modeling, and then we sort of got into the embeddings and transformer errors and deep learning and so on. Um, and we’ve benefited from all those waves because every sort of AI paradigm has different offers, different opportunities and benefits, and we tend to actually leverage them all. There’s never, we don’t sort of deprecate old methods. We sort of add, add to our, uh, different model, our kind of, you know, all the different models we run. Uh, and in terms of the gen AI space.
and so on, we are avid fans of it. And we have, there’s implications, Qloo has implications for it, both on the input side and the output side. So in terms of our property space, for instance, we have thousands of properties, you know, covering that we’ve normalized and structured and developed a lot of AI around. So everything from, you know, for any book and print, knowing the setting, the plot, the characters.
a movie, we understand the music in it, the kind of this, the micro genres of music. And if there’s a jazz club, that’s a merchant ID, an actual physical location, we conceptually understand that music genre links to X, Y, and Z music and so on. So there’s a lot that language models have helped with in terms of bolstering that.
foundational corpus we have, so elaborating on it. Like we’ve been able to discover new microgenres, running kind of LLMs against all our existing corpus and learnings, and been able to create directed graphs that show how there might be interesting relationships between entities. But the real opportunity for us is that the language era
Alex Elias (19:23.678)
we’re in right now where we kind of see language as the greatest interface over data. So it’s kind of like, you know, mobile was presented new opportunities. It was a new interface essentially over data. But I think LLMs now are essentially the new, one of the greatest interfaces for interacting with data. And
That interface still needs to be imbued with data and an oracle of truth and so on. And so a lot of the exciting commercial traction we’re seeing now, both with existing customers and prospective customers, is a lot of them are super excited about generative AI. You know, there’s multiple customers now at various stages of production that are heavily investing in the idea of creating generative itinerary planners, for instance, for travel.
And they’ve come to rely on Qloo to provide kind of an oracle of taste. Like what is it if we feed in a couple, a couple anonymized sets of context, what are the actual entity, the actual places someone should travel to the restaurants they should go to, where they should shop, et cetera, that we infuse the LLM with and then create a narrative around it. So Qloo’s tooled up to generate that inference within five milliseconds and it’s highly structured. And then it’s perfectly suited to.
being elaborated by a language model. And one of my favorite examples is late in 2022 in the early days of LLMs being deployed in customer-centric settings. So it was GPT-3 at the time, and we were supporting a hackathon where people were leveraging Qloo in conjunction with GPT-3. And one of the winning concepts had, you could just put a single taste in, so any one of our half billion plus.
tastes that we have as named entities. So it could be a music artist, it could be your favorite local brunch spot. So you just put in one entity and then it loops through Qloo, instantly generates this whole tapestry of probable tastes in other areas in metadata, and then layers that payload within milliseconds into the LLM, which then takes a little bit longer, but generates a long form essay that insults you.
Alex Elias (21:32.722)
insults your taste and in great detail. And it was pretty amazing. It was extraordinary to see. And it was kind of a silly demo that was like people were just, it was hysterical. But, you know, it sort of portended this kind of more profound thing that, oh, if you have.
really solid structured inference and clean data and kind of a knowledge graph that’s globe spanning. This could be tied into LLMs in such unique, interesting ways. And we’re starting to see that it’s only really in the past two quarters that we’re starting to see that meaningfully commercialized. Like companies are finally, you know, and we have Fortune 50s that are at the CEO level. They’re kind of like worried about putting LLMs that are unconstrained in front of
And one of the big places now, one of our big buying factors now is the fact that Qloo can help constrain, imbue LLMs with truthfulness, at least as it pertains to kind of culture and taste, and guide that. So yeah, it’s an exciting new era and we sort of have been benefiting on both ends of the stack.
Beau HD (22:42.044)
Wow. Yeah, that’s really interesting. I would love to play around with that. Um, you know, uh, just to have some fun. Yeah. Hey, you know, I’ve done some YouTube work. I’ve, I’ve developed that, you know, over the years. Yeah. No, so it reminds me. So you, you know, it’s not like you’re all in on artificial intelligence. You use it as a tool amongst your other arsenal of tools, essentially kind of, I imagine it’s helped you become more efficient. Um,
Alex Elias (22:46.254)
Yeah. You have to have thick skin. Yeah.
Alex Elias (22:54.371)
Yeah, you can too.
Beau HD (23:11.808)
extracting meaning. Yeah. It kind of reminds me of like, I took a few data science courses in school and it’s like we, you know, they taught us with Python, Matplotlib, the Pandas library, but it was also at the time where artificial or generative AI was taking off. So it was kind of like they weren’t explicitly like teaching me, you know, they’re kind of like, they’ll stay weary of it, right?
Alex Elias (23:12.302)
Yeah, absolutely, yeah. And it’s, yeah, absolutely.
Alex Elias (23:34.24)
Right.
Alex Elias (23:41.237)
Right.
Beau HD (23:41.5)
But I was using it to kind of see what worked, what didn’t. And I was able to become more efficient in certain areas. Yeah, so I imagine you.
Alex Elias (23:48.29)
That’s awesome. Yeah. I mean, co-pilot and all the implications for just, you know, productivity or just enormous. Um, it’s yeah, it’s been, it’s, it’s a really exciting era to be building things because a lot of, a lot of, you know, a lot of road things that are not.
particularly interesting in terms of, you know, knowledge work or being automated. And I think that actually liberates a lot of it allows for certainly on our team, there’s more thinking happening, ironically, which is, which is great.
Beau HD (24:18.28)
Hmm. I love that. Yeah. Yeah, cause it’s not just like, automated automation and job replacement. It’s, it’s shifting the way the business works and things with no problems.
Alex Elias (24:27.362)
Totally.
Alex Elias (24:31.998)
Yeah, I think at its best, it’s liberating, you know, knowledge workers to sort of think on to be able to explore in a higher plane and that, you know, it’s sort of like a.
It’s a highly scaled version of how Google in the past used to enforce some portion of time for sort of free thinking and so on. This is kind of a more organic mechanism for enabling that. So I’m very optimistic about the impact. And I think it’s like a fractal where, you know, it seems like…
Beau HD (25:03.014)
Yeah.
Alex Elias (25:09.426)
AGI or generalized LLMs and so on are going to solve every problem. But the reality is that it’s just creating new surface areas for problems to be solved. And so.
You know, I have a close friend who’s, uh, she, she’s a radiologist and sort of doing research on the impact on AI and sort of mammogram interpretations. And, you know, there’s, there’s a lot of opportunities, even in, you know, you think it would replace a physician interpreting it, but it actually kind of augments their, their intuition and, you know, can create a lot of, a lot of new surface area.
Beau HD (25:43.304)
Interesting. Yeah, well, keeping with that trend of where things are going and the changing landscape. So I understand when your company was first founded, it was predominantly a business to consumer company. Working directly with consumers, I know you guys acquired a taste dive in 2019, which is cultural recommendation engine, which has a social community, according to my notes, of some 4 and 1 half million users.
Very impressive. Can you just talk more about that transition from a B2C company into more of a B2B solution?
Alex Elias (26:19.058)
Yeah, so when we started, we had that hypothesis that we were aiming this recommendation capability. We were trying to arm consumers with that superpower of understanding their taste and exploring it and being able to really get meaningful discovery, like have something they could depend on to interpret their taste and…
connect them with things that would be meaningful to them in a much more efficient way that wasn’t adulterated by ads or by sort of this reversion to the mainstream that we see in the world. So it was kind of an optimistic hypothesis. Again, didn’t end up being a great business, but we were already consuming our own APIs. So we were already effectively an API business because we kind of containerized our logic.
and created endpoints. And so the front end application, so we had an iOS app at the time, we had a website, et cetera. All that was essentially a client of our own API. Such that one Twitter, the reason Twitter was interested in that era was that we were already structured in such a way that we could now just externalize, using some of the API kind of paradigms at the time and OAuth and so on. We could then externalize that. So what happened is over the years,
You know, we firmly pivoted to being supporting enterprises and developers. Um, and you know, we never lost our passion for the consumer problem here. Like that, that’s really what that was kind of the foundational, like motivator of myself personally, of our founding team, kind of starting the company was trying to solve that problem for individuals. Um, and of course there’s the argument that, you know, by empowering enterprises and other developers are kind of implicitly solving it in more of
of a B2B2C way. But that passion never left. And there was a company founded by former IBM engineers based in Amsterdam called Tastedive, which started before us. It actually was founded in 2008. And I’d become a user over the years. So part of what led to my passion is like, I kind of needed, you know, the Qloo app was…
Alex Elias (28:38.526)
no longer consumer facing. And so I wanted a place to be able to catalog. And so I became enamored with the community in it. The sort of, it was really capturing a long tail of preferences. Like it had a very niche and very global. So 30% only is US based. So you had, but it would show you how your taste overlapped with other members of the community. It continues to do that. And I was just shocked that there was, you know, there, there’d be a woman in Turkey who had
very specifically, you know, shared my taste in music and cinema, you know, and there would just be, it was kind of this beautiful sort of optimistic thing. They were running on Google ads at the time. It was basically a, so, and they also had an API. So they started building an API that they were externalizing and they had a very different customer profile. They were making it mostly available for free to fairly high rates.
And mostly kind of servicing individual developers who were sandboxing. You know, there was all sorts of things like the biggest movie site in Romania was entirely powered by the taste dive API. So it had, and then there was some large enterprises that were kind of tinkering with it. Some of our big customers now started off tinkering with taste dive. So we decided to, you know, we had discussions with our founding team and some of their investors and. You know, ultimately decided to acquire them and.
with really the intent of continuing and expanding. So we’ve been steadily expanding Taste Dive. We have some really exciting things coming to it. We expanded the category coverage and now covers places globally in addition to the media categories, expanded it to podcasts, which it wasn’t covering before. And we removed all advertising. So it’s completely ad-free, completely free to use.
And yeah, so it’s super exciting. It’s one of the most symmetrical cross-domain panels. So the tastes inputs that are being generated are very explicitly in a highly structured way connecting the worlds of music, film, TV, dining, podcasts, literature, et cetera. It’s like the average contribution on that platform. There was a stat we pulled a little while back. It was like 42.7 signals per capita.
Alex Elias (30:59.57)
spanning at least 3.8 data domains. So it was kind of like a, it’s an unparalleled, and it’s wholly owned by Qloo. It’s something that we train on those joins between entities. So nothing to do with consumer identity. You could use taste dive totally anonymously. We’re purely interested in kind of how these nodes, these different cultural entities interrelate. So I think that
Beau HD (31:03.365)
Hmm.
Alex Elias (31:27.254)
That’s given Qloo a big competitive advantage from a modeling standpoint because we have this very proprietary corpus in addition to all our other sources that helps us continue to refine particularly the long tail of taste and the cross domain associations. So.
Beau HD (31:42.056)
Huh, yeah, that’s so interesting. Yeah. Have you used, like, I’m curious more about like your clients. You know, I noticed you work with PepsiCo, you’ve worked with Universal Music Group, Netflix. I saw Samsung, a lot of really recognizable global brands. Is there an industry or industries that you could, you think could benefit most from Qloo?
Alex Elias (32:11.23)
Yeah, I think there’s quite a few. One of them, I think financial services is already leveraging it. There’s some big firms using it. And I think they have a huge amount to benefit from because they have this really interesting kind of nexus of transactional data. They’re heavily regulated. They’re very privacy focused. There’s very little they can do with their data. And so Qloo can uniquely, given very sparse inputs,
help them enrich at a merchant level to like understand, go from just understanding where people spend money to what their tastes are across all categories. And when you look at your average large, very large kind of global financial institution that’s spending billions in marketing, just the marketing efficiencies and actually understanding at a granular level what tastes look like across their categories are enormous, let alone driving rewards and spend for travel and things like that.
I think travel generally has been a big beneficiary of Qloo. You think about a global hotel group that has, you know, they’re really relegated now to their first party data. And so all they really know is like where people have traveled in the past and what hotels they might have spent money at and so on. And to be able to enrich that anonymously to this kind of 360 degree view of where people might want to travel and transact and what events in a particular geolocation might be most appealing to.
You know, send offers out for that hotel, you know, there’s just huge bottom line impacts when you do very basic things like tying Qloo into a CDP or CRM and helping decision, you know, what offer goes to whom, et cetera. So, uh, also, uh, the automotive space has been increasingly an area that’s, that’s been interested in our offering. So a lot of cars have connectivity now they have.
a lot at stake in terms of just understanding drivers and passengers. And in this new era, there’s kind of more of that demanded, and they want a privacy centric solution. So I think those are some areas that are really benefiting. There’s a very large consumer service we’re doing something with that’ll be announced soon that’s kind of integrating Qloo in a very novel way.
Alex Elias (34:30.554)
to make the consumer experience tremendously better, their core value proposition. So there’s not much I could say there, but it will be announced soon. And then we’ve had exciting, things like the ticketing space, event space, there’s an ephemerality to that good, events come and go, they’re highly taste-focused. So even though they might ultimately touch on a compound,
areas of our knowledge graph like you know a concert might have multiple music artists or a you know a comedy show.
might have a lineup and so on, but these are like fundamentally taste-based goods. And there’s very limited taste-based personalization going on in a lot of these platforms. So, you know, any platform that’s doing hundreds of millions in GMB, you know, selling a taste-based good can benefit from taste AI in a very kind of, you know, I don’t want to call it low hanging fruit, but it is. Yeah. So those are some of the areas. Like really, when we zoom out.
I think the core value proposition is any company that has something at stake in understanding consumer taste, either to increase marketing efficiency, sell more goods, et cetera, and also in recognition of the fact that tastes are getting more and more fragmented and faster moving and more difficult to predict. So sort of the traditional methods of kind of focus group or demographic-based analysis
Beau HD (35:42.204)
All right.
Alex Elias (36:01.674)
you know, the basic panels like a Nielsen and so on are sort of, you know, not going to cut it in a very kind of dynamic, fragmented landscape of taste. So I think those are some key areas where people can really, really benefit. And the other one is really like companies that are siloed in one area of knowledge and they have something to gain. So like Netflix knows about the content it distributes.
And they’re interested in expanding their knowledge to other areas. Like what music is most relevant to various audiences and various entities they cover. Even what film and TV shows out of network. So like, you know, content acquisition decisions, like licensing, what to, what to bring into their network. But, but moreover, even things like fashion tastes based on TV, you know, TV viewing habits for their merchandising team to better, to do better endorsement deals and bring on there’s just.
There’s a lot at stake in kind of understanding consumer taste. And I think that’s really where we end up playing a role.
Beau HD (37:09.816)
Interesting. Yeah, that’s, those are some great examples. Cause when I first was trying to visualize what, you know, an, an area, an industry that could be perfect for, I was thinking of maybe a coffee chain, you know, helping them curate, um, a music playlist for a cafe or even help curate maybe featured beverages based on the location.
Alex Elias (37:29.238)
So we’ve done that exact, the Qloo literally powers for a major QSR company what music plays where. And I’d love to take you through, we should sit down with the API one day and it’s just so fun to punch in random coffee shops in Japan versus parts of Paris first and just see how the music tastes. And even within New York City comparing the Bronx to the Midtown to Tribeca.
Beau HD (37:42.417)
Yeah.
Alex Elias (37:55.286)
and just seeing how music tastes, the inference around it are totally, totally different. And you’re never gonna get it 100% right with AI, but it’s a highly informed guest that just gives them a huge leg up and store experience. So, yeah.
Beau HD (38:10.488)
Interesting. Yeah. Because it’s, yeah. When I think of coffee shop, I’m like, I, there’s something to be said about a coffee shop. That’s like the same everywhere you go, but I, I myself gravitating towards, you know, uh, a coffee shop with his own persona and personality. Also, you know, I’m in Oregon, so the, my coffee shop is going to be different, have a different feel than maybe like a shop in Florida or Hawaii, right?
Alex Elias (38:19.083)
Right.
Alex Elias (38:26.207)
Ready.
Alex Elias (38:31.809)
Oh gosh.
Alex Elias (38:36.906)
Right. And Oregon has great coffee. I know we were talking about coffee earlier, and I have a large pour over here that I sort of chug on all day, speaking of which.
Beau HD (38:46.924)
Yeah. Right. Yeah. It’s tough to go anywhere else. If you like for coffee. Now there’s, there’s a lot of great coffee shops all down, but can’t complain.
Alex Elias (38:53.087)
Yeah.
What’s your favorite one in Oregon? What’s the, if you were to name one.
Beau HD (39:00.832)
Um, my, one of my favorite chains, um, is a company is called black rock coffee. It’s great. Quick. Yeah. Shout out to black rock coffee. Um, but there’s no shortage of local coffee shops around. Now I want to, I’m curious, what, what are some of the, the biggest or most impactful problems that you, you saw for your clients? I know you touched on some of them, but do you have any, um, anything that comes up?
Alex Elias (39:07.293)
Oh cool.
Alex Elias (39:11.484)
Nice.
Alex Elias (39:14.934)
That’s amazing.
Beau HD (39:29.616)
frequently where it’s kind of some friction or just issues.
Alex Elias (39:34.234)
Yeah, so some of the biggest challenges that we solve are really where a company is hitting a wall in understanding the consumer. And they have a big decision to make, something to deploy, or just a lot at stake in terms of… So we had recently an energy company that…
was basically looking at installing electric vehicle chargers at all their gas stations. And they just kind of hit a wall. They had major consultants brought in. They had major research firms, consumer firms, and hit a wall in terms of understanding where demand, where consumer tastes would be such that you have the Tesla owners, the Rivian owners, et cetera, that would, and we did this very successful kind of project where we basically ran
our geospatial kind of taste AIs and sort of generated predictions for where those types of vehicle owners would likely gravitate to and pass through. It was a very novel extension of our AI, but it had a big, big impact in that decisioning. And those are the things that get us really excited. And there’s a lot of examples of that in a very systematic way of kind of helping to really decision around consumer taste and helping break walls open.
in a way that’s totally ethical, that’s totally privacy centric, but gives these organizations kind of almost like a superpower to now understand at a very niche degree what the podcast tastes might be and where they should spend media dollars and what item or travel or anything in their inventory they should present when. So yeah, there’s been a lot of that kind of impact.
And, you know, we work with JC Deco, for instance, which is the largest out of home billboard company in the world. They have over 1.5 million billboards around the world. They have an amazing data apparatus internally. They bring in so many data sources, even from IBM with weather and all these different sources and.
Alex Elias (41:43.286)
You know, it was, it was really amazing to see them kind of since a few years now. They’ve been deploying Qloo to basically, you know, so despite all the data that they’ve collected and gathered and, you know, there was just something that was, uh, that was kind of unique to the taste AI where translating specific locations of billboards into granular tastes and everything from
TV shows that forthcoming television programs that people be most likely to want to consume in those areas to, you know, this very specific fashion taste that very block by block in Manhattan and so on. And, you know, so the taste AI was able, Qloo was able to kind of imbue it with kind of that understanding that had a big bottom line impact for their, you know, their business, particularly because more than 13% of their signage is digital.
And so they could do real-time bidding and personal, you know, there’s really a real-time decision as to what is shown. So those are some cool examples. We’ve been kind of extending the AI. There’s a really exciting new kind of capability around generating joint. So being able to pass in two sets of context and explore the overlap between tastes. And there’s huge implications in certain industries, as you could probably imagine.
Beau HD (43:02.312)
Hmm.
Alex Elias (43:07.242)
So yeah, we’re excited about that as well.
Beau HD (43:10.608)
Hmm. Do you, when you sit down with a, with a client and start discussing your API, do you find that there’s one particular feature or concept that’s difficult to communicate or do you kind of just let your API do the talking? Um, pretty self-explanatory.
Alex Elias (43:26.738)
Yeah, I think that’s a great question. I think something that comes up or something that often needs to be clarified is how much I think clients have a misconception that they have to bring an enormous amount of data to the table for this to work because there’s a lot of solutions out there where it’s like kind of a generalized recommendation as a service type platform where you basically bring all your data to the table and then it helps companies make sense of that and Qloo is fundamentally not that what Qloo’s actually proposing and what we offer is the ability to bring very little to the table. So if all you know is one little kernel of taste given whatever category you’re in or just a basic demographic parameter or some generalized kernel of first party Qloo can then probabilistically enrich that across all categories. So even with a sparse input you get.
the rich tapestry of output. Obviously, the more compound the input is, the more data points there are, the more accurate the inference is ultimately. But even with very, very little, you’re sort of kind of jumping to a different plane in terms of understanding the consumer. So that’s something that often needs to be clarified because…
You know, we were talking with a sort of additional hotel group we’re now bringing on that’s very, very large and global. And that was a misunderstanding. They felt like, Oh, they need, you know, they need to have all the data in the world for this to do anything for them. And the reality is that they, they need very little. And that’s why Qloo comes in to help them kind of enrich and expand their, their understanding.
Beau HD (45:04.424)
Hmm. That makes a lot of sense. I’m hoping some of your future clients listen to this podcast because you do a great job explaining it. Now.
Alex Elias (45:11.358)
Oh, thank you. I hope they do too.
Beau HD (45:14.432)
I’m curious, do you, I imagine you do at this point, but do you have any direct competitors? Cause I know the market has changed so much. And if you do, what’s something that you really do well? I know you definitely, I feel like you’ve probably answered this, but what’s something you do really well your competitors don’t or alternatively, are there alternatives you find your customers evaluating against Qloo?
Alex Elias (45:40.382)
Yeah, great question. So we’ve seen competitors within silos, so companies like Echoness, which were acquired by Spotify, they had a very similar thoughtful approach and offered a wide range of endpoints pertaining to music based on knowledge, content, you know, co-occurrences between music artists. So certainly anyone looking to specifically consume our service.
for the purposes of music might find some alternative APIs out there and similarly in the world of film and some other categories. But what we most find is that the thing a lot of our customers are most interested in is that enrichment across categories. And that’s something that’s been fairly unique to Qloo and Taste Dive, which we acquired and internalized. But there’s what we often find, I think is the most
Often suggested alternative is sort of a role your own approach. So we have a lot of companies just given the scale of the organizations and the amount of technical resources they have. There’s just very often this kind of sense of like, maybe we should try building this ourselves and try mapping tastes and entities. And, uh, and we’ve seen that with some of the biggest, one of the biggest retailers in the world. Um, we entered into a protracted kind of contract with, you know, three years ago.
And they ultimately at the board level decided that this was a capability they needed to own and they were going to build it themselves. And we checked in with engineers. So we ultimately, you know, didn’t engage and we checked in with some of the engineering leadership there, you know, a few years later and they hadn’t yet built it. You know, even though they have all the huge balance sheet.
you know, huge headcount. It’s a very niche problem. It takes time. The procurement of the data is extremely complicated. It’s very difficult to do it with regulatory compliance in mind as well. So, yeah, we often find that, you know, a company that we kind of help a lot of companies avoid the sort of headaches of attempting to build a global understanding programmatically delivered of consumer tastes. So.
Alex Elias (47:50.518)
Yeah, it’s been exciting from that perspective.
Beau HD (47:54.596)
Now, do you have, I know you teased some new features or announcements, do you have anything that you would like to share in the podcast or do you have anything you’d like to tease further about Qloo? I have to ask.
Alex Elias (48:10.73)
Yeah, so I do think so that idea of kind of comparing tastes and bringing multiple tastes into account has huge implications for consumer facing applications. So there’s even like a large sports ranking for us to think like the handicapping these types of things, they have a highly scaled app, they’re actually using it for sort of a matchmaking application.
Definitely implications there where, you know, we’re, we’re going from single player mode of here’s a set of context and generate inference to multiplayer mode where it’s like, here’s multiple sets of contacts and show us the overlap and, and then the other major product feature really is that we realized in many of our conversations that many customers are not yet at a point where they need this at programmatic scale programmatic scale. What is what I would consider to be more than.
million requests per month, for instance. And some clients have provisioned five billion requests per month, but anything sort of north of a million is no longer ad hoc research. It’s like a programmatic use of our system. But there’s a lot of companies, so in everything from independent film producers to a restaurateur with three restaurants to who desperately want to kind of have that same understanding of taste and aren’t really interested in consuming APIs at a programmatic scale and so on.
So we’re launching very eminently and anyone who’s interested in this should reach out. We’re launching kind of essentially a fully self-serve research tool that requires no technical knowledge whatsoever, but can fully interact with the expanse of our kind of AI and generate, you know, so if you were a restaurateur and you wanted to know what…
The music tastes of your three biggest competitors, within a certain locations are for whatever reason, or if you wanted to understand, if you’re a hotel owner, we have a shareholder who has a large hotel group, probably easy enough to look up. And he’s been leveraging, he’s been very interested in just ad hoc research about like what shampoo brands to put where.
Alex Elias (50:29.142)
you know, through, throughout their hotel, they have very niche hotels with kind of interesting personalities. And so, or if it’s a film producer who’s looking at casting decisions or scoring, or if it’s a fashion brand that’s looking at who to do an endorsement deal with, these are all kind of ad hoc research use cases. And so we basically are rolling out a product that’s gonna be a much lower cost that allows for queries in the range of thousands of requests per month rather than millions.
allows people to fully interact with the AI and create queries that are very intuitive and visualize the output in cool ways. So that’s coming online very soon. It’s pretty late in development, but it’s a super exciting expansion for us as a business because it allows us to take on a huge kind of area and pockets of demand that we just haven’t been able to address in the past. So that’s some of the hope there.
Beau HD (51:26.916)
Wow, yeah, that’s exciting. Thanks for sharing that.
Alex Elias (51:29.478)
Yeah, of course. Yeah. So it’s funny because we’re going this direction towards making it intuitive, self-serve, total layman. And then we also have a few applications more on the hardware side where they really want on-device learning. And so we’re also working on containerizing our AI as an appliance that a phone maker, in-flight entertainment…
Beau HD (51:33.008)
Yeah.
Alex Elias (51:56.054)
you know, on an airplane can have taste AI. So if you’re on that flight and it’s, you know, needs to, you want film recommendations, you want restaurant recommendations from when you land, et cetera. It’s all containerized in a, in an on-device, totally privacy centric kind of solution as well. So we’re sort of expanding in both directions out from our core API service.
Beau HD (52:17.724)
Do you see that, do you kind of see that that’s where it’s going is more on device processing or I’m curious about your thoughts on the industry as a whole, like where you see it’s going in the next five to 10 years from now.
Alex Elias (52:32.318)
Yeah, I do think differential privacy is something that’s a huge theme at companies that we all know, like Apple and so on. And on device learning is a huge theme. And, you know, we were initially skeptical that we’re like, is this something we’re building where there’s just one customer out there or three? You know, but what we’re slowly starting to realize, like we had a conversation with a major airplane company lately, and that was they’re one of many airlines, you know, and.
They’re interested in it. And then we had a conversation with, you know, uh, an automaker. They’re interested in it. And so suddenly we started realizing there is more of a, and there’s something that’s very compelling about the fact that it doesn’t need to take a round trip to an external server. Um, so you sort of inherently, you are delivering an appliance that is capable of just living on-prem within their environment.
Um, and I think there’s something very compelling about that. You know, there’s still the need for entity hydration and updating the model and so on, but it does give them, um, you know, a very feature complete kind of version, uh, that. So I do think that’ll be a bigger and bigger trend, particularly given where the regulatory environments had it and everything else. So yeah, definitely.
Beau HD (53:45.124)
Yeah. And I imagine you also see looking distant in the future. Every company is going to be utilizing, um, their data, right. To, to target more customers as more data becomes more prevalent. So I imagine every, like, there’s going to be a lot of customers for you specifically for Qloo, uh, in the, in the coming years, right.
Alex Elias (53:56.294)
Right. Exactly.
Alex Elias (54:07.262)
Right. Yeah, yeah, absolutely. And there’s more, you know, I’m sure that coffee company you mentioned that you love in Oregon, like they could probably, they could probably benefit from it. We’d love to pull some data for them. So.
Beau HD (54:14.597)
Yeah.
Beau HD (54:21.048)
Well, I’m coming up towards the end of my questions for you, but I want to do something a little fun that we can pull a little snapshot from. And I want to circle back to Qloo’s solutions. So there’s a lot of sort of useless marketing jargon out there, especially when it comes to artificial intelligence. I feel like every business is guilty of it to some degree. So if you were to give a short audio snippet of your company solutions right now, what might you say?
Alex Elias (54:48.542)
A short audio snippet. So Qloo allows you to generate informed guesses about taste in music, film, travel, all the major categories using very little input data as contacts. So you can pass one little thing that’s known and then Qloo will generate an AI informed guess about all the other tastes and preferences.
And it’s been proven to have a big impact and to be shockingly accurate. Uh, and it doesn’t require anyone’s identity or anything of that nature. That might be my best. Yeah.
Beau HD (55:26.34)
Very privacy centric. Love it. And then is there one big thing that you would like to tell potential customers? Like if you were standing on a rooftop, shouting out to your customers, what might you say is the one big thing?
Alex Elias (55:42.254)
The big thing would be that Qloo is easy to use. You need barely any data to tap into it, and you probably have a huge amount to gain if you get this highly-scaled global understanding of taste. It could make you more efficient marketers. It could help you sell more goods. It could help you create better content. So definitely come check it out, because it’s a big thing.
requires very little resources and very little data to get the flywheel spinning.
Beau HD (56:12.848)
That’s awesome. That’s great. Well, thank you so much. I’m curious, where can people find out more about you and get in touch with you?
Alex Elias (56:20.83)
Yeah, so feel free to reach out to us info at Qloo info at Qloo.com We check all the inboxes if you want to reach out personally alex at Qloo.com pretty simple Would love to hear from you all and super excited for what’s ahead and particularly if you’re Individual developer who’s building something interesting or within a large company and you have some cool ideas That you want to sandbox around with we’d love to support you. So definitely reach out
Beau HD (56:50.544)
Wonderful, that’s great. All right, Qloo.com, Q-L-O-O.com. All right, well, thank you so much for your time, Alex. I really appreciate it. I think this has been super insightful. I know personally, and I hope a lot of our listeners got something out of this as well. So hopefully we’ll have you back on these days.
Alex Elias (57:04.33)
Awesome. I would love to come back and thank you. I really appreciate the thoughtful questions and look forward to being in touch.
Beau HD (57:12.348)
Awesome. Well, thank you. Thank you all for listening to the Slashdot Media Sourceforge podcast. I’m your host, Beau HD. Make sure to subscribe to stay up to date with all of the upcoming B2B software-related podcasts we have for you. Hope you guys take care, and I’ll talk to you the next one. Thanks, Alex.
Alex Elias (57:28.39)
Awesome. Thank you, Beau. Really enjoyed it. Cheers.