Name: Brainstorm AI Singapore 2024: The Future Of E-Commerce
Uploaded: 2024-07-30
Duration: 11 min 54 s
Description: Ewa SZYMANSKA, Global Head, Rakuten Institute of Technology (RIT); Executive Officer, Rakuten Group Moderator: Clay CHANDLER, FORTUNE

Ewa SZYMANSKA, Global Head, Rakuten Institute of Technology (RIT); Executive Officer, Rakuten Group Moderator: Clay CHANDLER, FORTUNE

Brainstorm AI Singapore 2024: The Future Of E-Commerce

Conferences

Lia. Thank you so much for joining us today. We've had a lot of discussion about Southeast Asia, Southeast Asian firms and policies. So it's refreshing to be able to get a chance to have a little bit of a discussion about a Japanese company. Uh show of hands, how many know who, what rocket is, everyone pretty good. All right. So why don't we just start there? And you can tell us a little bit about Rocket 10 and maybe about the, the Rocket and Institute of Technology. But this is a, this is a company that's into all kinds of businesses. Right. Right. Right. Right. Right. Um So starting from the beginning, Rockton was founded in 1997. So it's a time of economic downturn um in Japan, um many businesses are struggling uh shutting down um small businesses uh need help. And this is when um r and online marketplace um opens up new markets to small and medium merchants, to rice sellers and uh you know, second brewers before could only sell um to in their cities in their rich regions. Now, the whole Japan uh opens up to them, um you know, fast forward to today. Uh We have 57,000 merchants uh on our platform and uh e-commerce still at the very heart of Ragon. Uh But since then we, we expanded and we are now in fintech in uh digital content in Telco. So, uh we have over 70 businesses. So overall, um can think of, you know, if you want to do travel, booking or uh no um hair salon or uh uh horseback racing. Uh uh there's just a wide variety of payments, car, credit card insurance, you know, I can go on overall, we have over users in Japan. Um So most of people in Japan are involved with at least one of our businesses with a cross usage of, of over 76%. Um and uh 40 million uh users a month engaging, engaging with, with R and which is uh basically a third of, of Japan population. So it is very much a household name in Japan. So you, you here in Singapore, tell us about the Rin Technology Institute and what that does. Right. Right. Right. So rec of technology is Rin's Research and Development Center and our mission is to bridge the science with business applications. And I'm sure that Rin benefits from this tremendous advance in uh in A I. Um so we worked uh with uh computer vision language or customer behavior from our 70 businesses and uh we do both work traditional A I as well as now uh generative A I and your background is that you're a, you're an expert in, in the science of human decision making. Right. Right. Right. Right. So, before artificial intelligence, I was studying human intelligence, it's a good, a good combination. You make an interesting point about Japan, which is that, although, you know, I mean, we, we rank its giant companies in the Fortune 500 list and they're probably globally well known that Japanese Japan's economy overall is very heavily dependent on small and medium sized enterprises who are often not as technically savvy or sophisticated as Sony or Toyota or, or something like that. Um And these are, these are kind of the bread and butter customers for, for rocket. Tell us a little bit about their level of kind of technological sophistication and what, how we should think about how, how tech savvy they are. Sure. So we're uh serving all our, our partners. So regardless of their technological proficiency. So we make sure that they have the know how um and, and, and skills to make the most out of our platform. Um You know, when, when we started um no, uh most of these sellers actually didn't know much about internet at all. I mean, we're talking about late nineties. Um So very early on um we established this function of the Ecommerce consultant. Um So that's uh uh uh uh a person that works uh very close mostly with the merchant and assist them with you know, whatever is that, that they need, uh it might be help with setting up a storefront, uh running a marketing campaign, educating them about, you know, search engine optimization. What are the social media out there? What is Facebook, what's uh Instagram and how to take advantage of those? Um So, uh so, so very much uh this partnership with, between the consultants and the merchants is at the, you know, the heart of uh R and marketplace. So a lot of these folks would maybe be, you know, small merchants, maybe live in a kind of rural area, uh might be older, kind of mom and pop type businesses. So that's um that's, that's very intense to reach out directly to those people. Um How big is the, the kind of force of these, these consultants? Right? So we have our regional offices all over Japan and we have ecommerce consultants in all of the three offices, making visits to merchants sitting at the table with them and uh working out the strategy, you know, for, for sales for next year, for, for, you know, Valentine's campaign, et cetera. And so do you see that A I will make it easier or, or actually more complicated for those kinds of merchants to, to operate? Yeah. No, that's, that's a great question. I mean, we're uh extremely uh positive about this technology about A I I mean, Rin was founded on, on this belief that technology brings people closer together, you know, even internet, our online, online uh service people can think no online means uh impersonal. It's convenient like a vending machine where Drax and have a very different approach, right? And from the very beginning, it was all about allowing merchants to express their personality of their business, tell their story uh of the product of the family. So with generative A I, um we see a lot of opportunity to help merchants uh you know, do this even better. So, you know, a lot of them, as you pointed out are, are small merchants, they may not have marketing teams or designers on their team and getting in their hands, tools that allow them to, you know, create more beautiful storefronts, tell their story more visually compelling way. Um It's something that we are really excited about. Um Another area is just to give them ways to connect with their uh their, their customers better allow for personalization, uh help them to craft responses to the customers, which of course they have full of control over the, the final output. Um We believe that this will help them be, you know, even closer to their uh and customer. Um which kind of plays also on this um idea of Amo Teni. And we discussed this at clay earlier uh which is very central to uh Japanese culture. And it, it um uh uh goes back to the tea ceremony and this is the mindfulness and the principle of um really uh care and hospitality, understanding the needs um of, of, of your guest, of your customer and then going above and beyond to, you know, meet and exceed those expectations. Uh So we believe those tools will now allow our merchants to practice that. Oh, very Japanese kind of idea. Let me ask you just a quick question about, about language and culture and how it combines with A I I mean, obviously, you both develop in House Rogen's own large language models in Gen A I. And you're also working with companies like open A I and others to um to collaborate in rolling things out. So you, you kind of pick and choose as to when and where to make use of, of which one. How can you give us a Japanese perspective on this question that we discussed earlier this morning about whether or not uh large language models that are trained mostly on Eng in English can actually be very responsive and useful in different cultures. You know, like, like Japan, Japanese language where the culture is so specific, right? Um Very good question. Um a few months ago, I would give a different answer. Um Today. Uh My answer is that these models are, are pretty good. Uh They are not as good as they are in English. Um There is still certain um cultural differences and uh I think the panel discussed this before as well for other Asian languages, the politeness levels. Um, there's some, you know, some, some country might, might not be correct and, but, but it is close, um, having said that it's still a little bit more effort for the Japanese user to, um, to utilize that and the reasoning skills are not as good as, as, as they are in English. But, uh, but yeah, it's actually pretty impressive, an interesting perspective. I think it's a question about which there's a lot of uh different opinions and points of view, especially the more advanced the technology, right? But I would like to know something however, and that um there are certain, so, so tokenization for Japanese um was not working. Um so, so well. Um and basically tokenization is that, you know, when you go from, from text to uh uh numeric encoding, you need to uh kind of parse these, these uh text input into, into tokens. It can be as long as a word or a sub word or a character. And basically these um you know, commercial large language models that they have been available, they were very inefficient in a way that they tokenize Japanese, specifically Japanese language. And they would use multiple tokens for uh a single country, for example, which was very inefficient, very costly. So what we put effort into when we uh developed and, and open source, our uh our act and LLM um 7 billion parameter model, we put effort into developing a tonier that is much more efficient and uh you know, is able to take in more context uh into uh uh and uh and be just uh more uh less computer computationally intense. Um So that was kind of the, the contribution that, that, that we made and that was something that had to, that was very language specific, right? So, you know, my uh my view is that uh you know, we are very close, but certain things have to be solved in order to close the gap. So that's all the time we've got, but a fascinating Japanese perspective. Thank you so much. Thank you. Pleasure.