This Startup Just Built Japan’s Most Powerful Supercomputer

Preferred Networks is making changes in Japan.

Over the past few years, this AI startup has raised more than $130M in venture funding and grown to more than 130 people.

If you live outside of Japan, you might not have heard of this team, but they are working with Toyota to create the next generation of driverless cars. They are working with Japan’s most advanced industrial robot manufacturers to improve efficiency. They are also working with many financial institutions on fraud detection.

Oh yes, and they also built Japan’s most powerful commercial supercomputer.

Today we sit down and talk with Daisuke Okanohara, the technical co-founder of Preferred Networks. Daisuke and I talk about the story behind Preferred Networks, he also shares his challenges and current strategies for maintaining the company’s experimental and engineering culture as it grows larger and more structured.

Daisuke also talks about his time at Google, how Japanese AI stacks up to China and the US, and why he’s convinced that their biggest competition is going to come from somewhere you would never expect.

It’s a great discussion, and I think you’ll enjoy it.

Podcast: Play in new window | Download | Embed

Subscribe: Apple Podcasts | RSS | More

Show Notes

• What edge-heavy computing is and why it’s important
• How a Google Internship changed Daisuke’s outlook on AI
• The future of driverless cars at Toyota
• Why the team decided to build Japan’s most powerful supercomputer
• Why you can’t sell disruptive products to large companies
• How to keep a curious spirit even as your company grows
• Where the real competition in AI will come from

 Leave a comment

Transcript

Welcome to Disrupting Japan, straight talk from Japan’s most successful entrepreneurs. I’m Tim Romero and thanks for joining me.

Preferred Networks is without question the brightest star in the constellation of Japanese AI startups. It attracted about 130 million in venture funding and have grown to more than 130 people over the past few years.

Of course, if you don’t follow AI, you might not have heard about them at all but they are the technology behind Toyota’s driverless cars, some of FANUC’s industrial robots, many cutting-edge applications in other verticals, and as a side project, they also built Japan’s most powerful commercial supercomputer.

It’s an interesting team to say the least and today, we sit down and talk with Daisuke Okanohara, Preferred Networks’ technical cofounder.

We talk about how Preferred Networks got started and got to scale and he also shares his challenges and strategies of trying to maintain the company’s experimental and engineering culture as it grows larger and monthly revenue pressures increase. Daisuke also talks about his time at Google, how Japanese AI stacks up to China and the US, and why he’s convinced that their biggest competition is going to come from somewhere you would never expect it.

But you know, Daisuke tells that story much better than I can, so let’s gets right to the interview.

Tim: So I’m sitting here with Daisuke Okanohara, the cofounder and Executive Vice President of Preferred Networks, Japan’s leading and probably most innovative AI startup.

So thanks for sitting down with me today.

Daisuke: Thank you very much.

Tim: So Preferred Networks talks a lot about the importance of edge -heavy computing. So can you explain exactly what edge-heavy computing is and why it’s important?

Daisuke: Cloud computing is one of the most important trends in the IT area and most people believe that most computations or operations should be done at a data center or across site, and it’s okay if we deal with fragile information but when it comes to solving real-world problems like operating robots or autonomous driving, we need to process data at the edge site or near to the device.

Tim: So just so make sure I understand it correctly, edge-heavy computing is important because of the latency of how quickly?

Daisuke: Yeah, and reliability.

Tim: Ah, right, right, because you want to always have connectivity?

Daisuke: Yes, current internet, it’s not reliable to use for the mission-critical tasks.

Tim: Okay, so it’s really a trade-off between the latency versus the amount of computing power you have? So if you can wait for the results, it’s great to compete in the cloud but the closer to real-time, the more important edge computing is?

Daisuke: Yes.

Tim: Alright, that makes a lot of sense. But isn’t the promise of big data that you need like, huge data sets? So when you are using edge-heavy computing, do you still send all the information to the cloud for analyzing later or do you generally try to make self-contained systems?

Daisuke: Our goal is to extract the information at the edge site and only send the essential information to the data center or other place, so we do not even know the data center or cloud but when there are many devices, we cannot send all the data to the cloud and we need to process most of the computation at the edge site.

Tim: Okay, so it’s just a more clear separation of training data and execution data?

Daisuke: Yeah. In the current status, training requires much more computation part so we need to train the model at the cloud site and the operation execution or inference at the edge site.

Tim: Okay. Now, Preferred Networks is involved in AI applications ranging from like, automotive and factory automation, life sciences, network security, and I want to talk about all of those in a little bit but before that, I want to talk a little bit about you.

So before you got your PhD, you were an intern at Google in San Francisco? How did that happen?

Daisuke: Yes. At the time, I started Natural Language Processing and many of my friends started working at Google. I was also interested in how Google solves a problem and how the people in Google are working, so that they could produce many excellent products.

Tim: So did you apply to Google in San Francisco from Japan or did you apply to Google Japan and said, “You need to work with this group in San Francisco”?

Daisuke: Actually, I applied to Google Japan but at the time, Google Japan did not have enough resources to accept intern members, so internship students went to San Francisco.

Tim: Okay, so it sounds like you had a real passion for AI before you started working with Google. So what did you take away from that internship?

Daisuke: Before working at Google, I did not imagine how AI can be used for solving many problems. For example, the search engine, motion transition, image recognition, speech recognition, and so on. So many products and services use machine learning as an essential tool.

Tim: That’s interesting because I think in a lot of technologies, Japan is very strong in academic research but tends to be weaker in creating new products and bringing new products to market, not just AI.

Daisuke: Yeah, I think Japanese is a bit conservative and they hesitate to do different things. In my opinion, it comes from Japan is a monoculture. All people speak Japanese and spend the same expenses wear the same clothes and maybe in the companies, so it is very tough to start new things. I think that diversity is very important to bring new products to our futures.

Tim: Yeah. It makes sense working with Google, you understood the importance of practical applications of AI but then when you came back to Japan, you went back into university.

Daisuke: Actually, I started my original company Preferred Infrastructure. I did research at school at daytime and I did business maybe at night or at the morning, so I did not spend –

Tim: Okay. Well, that’s right because Preferred Networks was spun out of Preferred Infrastructure. So you started Preferred Infrastructure while you were in college. Why did you decide to spin out a new company out of Preferred Infrastructure? Because it focuses on the same kind of technologies, right?

Daisuke: Yes. Preferred researcher focus is not using AI in the real world and we have very good business and the business was growing, so it was difficult to focus on two things, the current business and a very different business on AI and IOT. Therefore, we decided that it is better to separate the company into groups so that each group can focus on the one thing.

Tim: Okay, so if I understand, Preferred Infrastructure is more general AI and Preferred Networks is more applications and IOT?

Daisuke: Yeah. Preferred Infrastructure, the main business is to sell search engines or recommendation engine, and the main customer, the media companies and those main data was text.

Tim: Okay, so Preferred Networks is the one that has a more general mission to bring AI to different sorts of industries?

Daisuke: Yes, and the difference is Preferred Infrastructure, the original company, did not raise the money from the partners or venture capitals at all.

Tim: Okay, let’s talk applications and some of the things you’re working on. So one of the most exciting projects is your collaboration with Toyota and their focus on autonomous vehicles, and you’ve been working with them for three, four years now and they just recently invested $95 million into Preferred Networks to accelerate that research. So tell me about what you’re doing with Toyota and the self-driving Prius.

Daisuke: We try to apply the AI technologies, especially the deep learning and the related technologies, and there are so many programs in the autonomous driving as you know. In our company, we have several chains to solve the problems in the autonomous driving.

Tim: One of the things that I thought was most interesting about your approach is that the cars are actually collaborating with each other in the demonstration I saw, and I haven’t seen that approach taken by Google or Uber. Do you think that that sort of collaboration between autonomous vehicles is going to be something that becomes more important to the future?

Daisuke: I want to clarify the demonstrations. So in the demonstrations, each car doesn’t share the information; they only share the knowledge or model, doesn’t communicate to each other in real-time. We hope collaboration will reduce accidents but the countries are – it is difficult to implement in the actual cars.

Tim: And so the challenge again is just the amount of time it takes to collaborate with other cars, it takes too long?

Daisuke: Yeah, and also the reliability. Most cars are equipped with – there is communication devices, that it is okay to use the communication tool but it will take more time.

Tim: Earlier this month in America, we saw the first fatal accident from an autonomous vehicle and Uber has announced they are scaling back on their testing. A lot of other companies have announced they’re going to scale back their testing. Has that accident changed Preferred Networks’ plans for development to rollout or has that accident changed Toyota’s plans on how autonomous vehicles will be developed and rolled out in Japan?

Daisuke: We’re also concerned that these accidents are very important but on the contrary, we think that our current plan do not need to change but we need to avoid such accident. I think that especially in the cityside, it is very difficult to make the model accurate enough to avoid any such fatal accidents. It will take more years than the autonomous driving system in the highway.

Tim: For the past couple of years, American companies have been getting a lot of press attention about the advanced self-driving functions and I’ve noticed in Japan, the tests for self-driving vehicles have been very, very controlled. They tend to be on a very specific route; they are safe, and American companies are more likely to do real-world road testing. Is that because the American companies’ research is further advanced or is Japan just more conservative about how they want to test autonomous vehicles?

Daisuke: I think both. Current autonomous driving technologies largely depends on the success of the DARPA Challenge, but when it comes to today’s autonomous driving system which heavily use deep learning technologies, especially image recognition, I think that many new startups compete with each other not only in Japan but also in China and Europe, but I think that some of the companies in the US have great technologies.

Tim: How do you think autonomous vehicles will roll out? Do you think that there will be more of a focus taken away from city driving and place more in rural environments or long-haul trucking on highways? How do you think this technology will progress in the next five or 10 years?

Daisuke: I think the first autonomous driving technology will appear in the highway. It is much more a simpler case than a city or rural case. I think that many companies will provide the highway first. In the case of the city, we maybe need some new regulations or some new environment equipment to make sure that the autonomous driving system is safe.

Tim: Alright. Let’s talk about the MN1, Preferred Networks’ private supercomputer. So this is the fastest industrial supercomputer in Japan, it’s 12th fastest in the world. This is really cool but why? What do you use this for?

Daisuke: To develop new AI applications, we require talented researchers, engineers plus data, plus computing resources.

Tim: Is the supercomputer itself used for the AI modeling or is it mainly a research tool that Preferred Networks is using internally?

Daisuke: We currently use this computer resource for the research and development, and several researchers use this resource to try new ideas or develop new models. Also, many developers use this to fine-tune the model or validate the model.

Tim: Okay, let’s talk a bit about Chainer, your open-source and distributed deep learning application. What kind of applications are other people building with Chainer?

Daisuke: By using Chainer, many researchers and developers develop several models and since many Japanese companies and researchers must use a Chainer, so there are many products in Japan made by Chainer.

Tim: Oh, okay.

Daisuke: Yeah, like speech recognition system or image recognition. So many products made by Chainer in Japan.

Tim: Okay and are there international users as well?

Daisuke: Yes, several researchers use Chainer, Microsoft Cambridge Research Group use Chainer because by using a Chainer, the developers can try a very different models . It enables developers to implement new models in different way.

Tim: So people are using it both for research and experimentation and to create products that are being brought onto the market?

Daisuke: Yes.

Tim: That’s fantastic. And I noticed, you also have Paint Chainer which is the course in line drawings. Is that an application or is that just kind of a fun thing you guys built?

Daisuke: Paint Chainer and actually the Chainer itself was not actual products. Our company has 20% project. We also use this tool and several engineers and researchers try new ideas and the developer creates new projects and develops new tools.

Tim: It’s the same idea that Google popularized where employees can spend 20% of their time working on projects that they are excited about and if they might be able to bring some new product to market.

Daisuke: Yes.

Tim: Okay, and Paint Chainer was one of those?

Daisuke: Yes, so the Paint Chainer was first developed by one of our engineers, Yonetsuji. At the time, it was a robotics engineer, he has also interest in such a painting and artwork, so he just applied a neural network model to colorize line art and it worked very well from the first attempt. So as members helps him to release service, so our company supports this activity and tries to self or marketize this prototype of product to the market.

Tim: Okay, we will have to make sure to put links to all of that on the website when the podcast goes live so that our listeners can check this out because it’s something you really need to see. It’s hard to explain on an audio podcast but it’s really interesting.

Now, most AI companies tend to try to focus narrowly on a couple of key products, whether it’s search or discovery, or factory automation but Preferred Networks has gone the other route. You guys have gone really wide and are trying to work with as many different industries as possible. So why did you decide to go wide instead of going narrow?

Daisuke: This is one example of our unique culture, we want to mix different culture or different markets so that we want to make an environment where the new innovations emerge. When we visited Silicon Valley and discussed with several venture capitalists and the venture capitalists said that, “Please focus on the one product or one service,” but we just denied and yeah, maybe this is not the standard way to do business but I think that trying to solve very different problems, then there may be some new connections found between the different products or markets, and I think that today, we can try new things with a lower cost. Before, when we, for example, build the new software products, it requires several months or maybe one year with several engineers, but today, we can make new models within one week.

Tim: Well, I think this is something that – it’s maybe not unique to Japan but this is something that is very Japan-specific. A lot of Japanese startups do seem to have the sort of integrator model. It seems that a lot of large Japanese companies don’t have in-house expertise on new technologies and whether we’re talking about drones or AI, the Japanese companies, they want to reach out and work with new startups rather than trying to develop things internally but American companies seem to want to do it themselves. Do you think this will hurt you in the long run? So in the short run, it seems like a huge advantage because you have access to all of this expertise and all of this data that Toyota has on driving and that FANUC has on industrial robots that would take you years and years to try to develop on your own, but as time goes on, do you see Preferred Networks staying as kind of an AI specialist and an integrator or do you think that some of those 20% projects will become something that you guys will decide, wait, this is great, this is a project we’re going to sell and bring to market?

Daisuke: I think we first need to show that this is a product made by AI so is it a more vertical model. We need to develop library software and solutions for each industry but when we understand that yeah, this is a market, then we can focus on the most strongest part.

Tim: So I think one of the challenges is truly disruptive innovation, something that is really new. If you take something that’s really new to Toyota or Hitachi, they would want to do it because it’s too new and too disruptive. So do you think there might be a point where Preferred Networks says, “Wait, this can change the world and we’re going to do it ourselves”?

Daisuke: There are several difficulties in providing such disruptive technologies. First of all, it is difficult to develop such technologies so we need enough environment in our company to develop such disruptive technologies. This maybe require very different thinking style, but I think that now we have such an environment, and the second difficulty is how to make these technologies accepted by the companies or consumers? But anyway, we first need to provide these products to the early adopters. When they understand that this is very useful, they can create the attitude towards such technology and we saw such changes in the previous 20 years. When the smartphone appears, the past two or three years, many people thought that this is very unique but I don’t need to use it, but within one year, it will become the usage percentage surplus 50%. So when people accepted, the change is very lobbied.

Tim: There are so many Japanese companies. Japan as a country was really surprised by that disruption of the smartphone. So do you think that because of that, are big Japanese companies more concerned about disruption, are they more anxious to do new projects with new technologies than they used to be?

Daisuke: They understand the importance of such disruptive technologies but it is not specific to the Japanese company but big organizations, it’s very difficult to start such new things because such a new idea is a very fragile and very weak. Let’s say that in the main business, several hundred million dollar scale but when it comes to the new markets, market size is very small. So they cannot focus on this new market and this is very large, so therefore, they try to find such a new idea or solution outside of the company.

Tim: So you think that the big Japanese companies are more willing to try new things, to look at new technologies?

Daisuke: Yeah, I think so, but it would take time to accept such technology.

Tim: Okay, well, progress is always step-by-step, right?

Daisuke: Yeah.

Tim: So what is Preferred Networks’ core business model? Are you selling software licenses, are you selling consulting services, are you selling project implementation?

Daisuke: Our main core business is providing licenses to our customers and also the ongoing research projects. This business model depends on each market. For example, the autonomous driving or industrial robots, or machine tuning, and life science or healthcare. In each market, it requires a different business model. So we have a different business model for each market.

Tim: Oh, wow, okay.

Daisuke: Maybe it’s difficult to understand our business and our business model but –

Tim: It seems like you have a core engineering and technical competence, and you’re willing to be flexible in addressing whatever market you can with that competence.

Daisuke: Yeah. As I said, to develop an AI product, the important thing is how to retain talented people and enough data and enough computing resources. These are three important partners and I think that since AI development requires maybe several hundred different ideas or products, we cannot simplify such technology to one technology.

Tim: So is your vision for Preferred Networks to have a lot of independent semi-autonomous groups that this is the autonomous driving group and this is the factory automation group, and they might have different business models?

Daisuke: Yes, different business but the important thing is that technology is shared. Program code and data and model are shared among the chains.

Tim: Now, that is sort of like the engineering dream situation. So you’re coming from a technical background, I also come from a technical background, but how do you keep that spirit and philosophy as the company grows larger? So you guys have been scaling up really fast. How many staff do you have now?

Daisuke: 130.

Tim: So how do you maintain that independence and that philosophy as the company gets bigger?

Daisuke: Yeah, we face the problem of how to scale our chains and we tried several things. For example, we use chain management tools, also, the communication tools like Slack and other tools. We also made sure that each chain have enough support from the outside, so we are now going to be able to develop a system so that we can scale more.

Tim: Okay, but so far, so good?

Daisuke: Yes, so far, so good.

Tim: Maintaining that philosophy requires a different structure as the company grows larger.

Daisuke: Yes. This is a good thing and a bad thing but the cofounders, I and CEO Nishikawa did not have experience of working at ordinary companies at all. We just –

Tim: Yeah, that’s definitely a good and bad thing.

Daisuke: Yeah, good and bad thing. So we think that this should work but our experienced engineer people, they are from Sony or Walmart, so we discussed about what is the best organizations and the best way to keep our companies.

Tim: Right, so try to get the best of both sides?

Daisuke: Yeah.

Tim: Alright. Well, who do you see as your main competition?

Daisuke: In each market, there are different competitors, especially in the technology side. I benchmarked several research institutions companies Deep Mind, Open AI, and many Chinese startups. Actually, we discussed with those people in such groups but when it comes to the business, I think in the long run, competitors have not emerged yet. When new technologies appear, there are new companies, soon companies.

Tim: So you think that the real AI breakthroughs and your real competitors are not necessarily Google or IBM but some new startup that doesn’t exist yet?

Daisuke: Yeah, yeah, I think. These startups will emerge from new world.

Tim: You know, artificial intelligence is a really strange topic. So over the last 40 years, there’s been like two opposing forces where we have marketers who want to say that everything is artificial intelligence, right? And any simple algorithm or heuristic, they’ll say that’s AI, but on the other hand, we have like, academics who keep moving things in the other direction, who when a computer learns to paint or write music, they’ll say, well, that’s not really intelligence or when the computer becomes the best chess player in the world, they’ll say, that really isn’t intelligence. So what applications do you think that AI as it exists today is really good for and what applications do you think that AI is not really ready for?

Daisuke: Currently, the AI, especially the machine learning and the deep learning is good for the interpolation problem. When it sees a problem in a similar way, then they can solve the problem but when it tries to solve the extrapolation problem, it is not good as a human at all. For example, a robot can grab the object if they see the object before but when it tries to kick a totally different object, then its performance is very bad. It’s a generalization program and the machine learning tries to solve this generalization program over the several 10 years but still, it is very far from the human capability.

Tim: Do you think we will ever see general artificial intelligence or do you think that’s strictly science-fiction?

Daisuke: I think that we can achieve this in the future but I’m not sure when.

Tim: Okay. Let’s talk a bit about Japan. I hear a lot that AI research in Japan is falling behind what’s happening in the US and in China. Do you think that’s true?

Daisuke: There are many AI researchers in Japan but they are not focused or not organized. Japanese government tries to organize several new AI forecaster institutes but especially in the US, they have very strong IT companies and they have enough resources to start very challenging projects.

Tim: Is the problem what you mentioned before about American companies like Google being very good at taking technology and making products and Japanese companies falling behind in that area? Or do you think the core research in Japan is a little behind?

Daisuke: We can read papers, we can also share the programming cost, so I think the gap is shrinking every year. Three or five years ago, the only very strong research groups did great research but today, thanks to arXiv and GitHub, and Twitter, and many related social networks, we can share the research real-time and without anybody else. So I think many researchers, especially the young researchers now are publish, releasing new research.

Tim: Okay, so in a sense that from country to country, if the court research gap exists, it won’t be as important because everyone has access to the same research?

Daisuke: Yeah, so it is not specific to the AI but today, such an open research changed the environment of the research.

Tim: What about – so you have a particularly strong background in NLP, Natural Language Processing. How related to human language is that? So for example, research done in the US on English data sets and research done in China on Chinese data sets, is that applicable to NLP in Japanese?

Daisuke: Recently, the neural network tools, machine translation, speech recognitions enable us to develop tools independent of languages. I believe that in the new future, such a language gap will decrease.

Tim: Okay. Well, listen, Daisuke, before we wrap up, I want to ask you what I call my “Magic Wand” question and that is, if I gave you a magic wand and I told you that you could change one thing about Japan – anything at all – the education system, the way people think about risk, the legal system, anything at all to make it better for startups and innovation in Japan, what would you change?

Daisuke: I want to use this magic that people have the ability to learn new things even if they are adult. Today, I think we need to learn new things every day, every year, but people, maybe not specific to Japan, but people tend to not learn new things.

Tim: Do you think that’s because it’s difficult to learn new things as you get older or do people just not want to, they don’t try?

Daisuke: I think that yeah, they just don’t try. Current education system is very not flexible. People studied only during school and after graduating school, they don’t want to learn or they don’t try to learn new things but today, there are many new ideas and technologies, and new programs. So people have to study every day, not only English or such computer science but also need to study new things.

Tim: You know, I think that’s true and you know, I think that’s true worldwide, not just Japan. Our whole idea of how we educate people is just, okay, you go to elementary school so you can get into high school, you graduate to high school, you can go to university, you get good grades in university so you can get into the company, and then you’re done, and it isn’t true anymore – maybe it never was but it certainly doesn’t seem true now.

Daisuke: And I also feel that learning does not mean that remembering some new rules or some new formula, or some new historical event. Human can acquire new skills. This idea comes from the AI research. Remembering something is a very small portion of intelligence and for example, when we study the reinforcement learning, we find that ability of the learning to learn is very important but in the school, we explicitly learn such skills, learning to learn.

Tim: That’s a really good point. I think there’s maybe like, kind of three different steps to it. So as we were saying, just learning new information and historical facts, that’s easy and people will do that, and the second level is kind of acquiring new skills, like learning a new programming language and some people will do that but I think the third step and what’s really hard as people get older is learning new behaviors and changing the way you interact with people in the world.

Daisuke: Yeah, and changing the habits or customs. Yeah, this is very important but it is not taught by school.

Tim: Yeah, but I think you almost – I think you really need a magic wand to make that happen. It’s a deep part of kind of human nature, I think. We don’t like to change.

Daisuke: Yeah. Yeah, I imagine that if, for example, most people can learn new programming languages, and not only the programmers but most people can use the program to develop a new software, new customized tools, then many programs are solved, or if they can learn new languages, English or Chinese or other languages, then they can directly read other countries’ news articles or new opinions. So our ability is very restricted. Maybe this comes from our type of thinking but we need to change.

Tim: Yeah, so be willing to acquire new information and new skills, and new behaviors throughout your life.

Daisuke: Yeah.

Tim: Excellent. Hey, well, listen, Daisuke, thank you so much for sitting down with me.

Daisuke: Thank you very much.

And we are back.

Let’s talk a bit about what I call the integrator strategy for technology startups. The integrator strategy is when a startup focuses on a core technology and then sells services around that technology to companies in a wide variety of different industries. Each engagement results in a completely new product tailored to that client and to that industry. Preferred Networks is clearly using this strategy right now as is our guest from last month, don’t integrator Blue Innovation.

Now, the integrator strategy is very different from just simply selling a product to a wide range of verticals. Salesforce and Oracle for example, sell to companies in almost every market and there is often a lot of integration involved in the rollout of their product. However, these are product companies.

Everything they do is geared to increase the sales and penetration of their platform. Integrator startups are producing completely different products with each engagement and it’s easy to see the appeal of this integrator strategy. Since these projects are long-term and involve large clients, you have fantastic PR exposure and very steady revenues, and best of all perhaps, you are not locked into a specific product.

You have access to the best domain knowledge across a huge variety of industries and you can focus not only on perfecting your core technology, but experimenting with a lot of different business models. In this way, you know that you will have the best business model and the most advanced technology to pivot onto a specific product when the time comes.

When the time comes.

But are integrator startups able to make that pivot? It’s incredibly hard to give up on the steady revenues and the great publicity to bet the company on this one new thing. It’s not impossible by any means but I don’t know of any startup who has successfully pulled it off.

Now, it doesn’t mean it hasn’t happened, of course, and if you know of an example, drop me an email – I’d like to talk to you about it.

The other challenge with the integrator strategy is that it makes it very hard to expand globally. For example, I talk with a steady stream of American AI integrator startups who want to expand into Japan. When I ask what they do, a typical company will give examples from finance, energy, travel, and e-commerce and assure me that we can work with almost any industry.

Well, yeah, sure, if you have someone willing to spoon feed you the domain knowledge and take a chance on you, I’m sure you can, but why should they do that?

The integrator strategy requires deep relationships and deep trust, and when you go into a new market, you don’t have that. You lack not only the commitment to the Japanese market but to a specific product on a specific industry, and you’re asking for someone to risk their career or maybe at least risk their promotion by selecting your company for a long-term collaborative engagement. It’s a hard sell, particularly when there’s local competition pursuing the integrator strategy, and there’s always local competition.

But still, this wave of AI is new. A lot of real research has moved out of the university labs and is now being done by AI startups. This might represent a fundamental shift in how AI research is done and how significant projects are brought to market, and if this is a true sea change, then there’s no question the Preferred Networks is far ahead of everyone else in this race.

If you’ve got an opinion about artificial intelligence or the integrator strategy, Daisuke and I would love to hear from you. So come by disruptingjapan.com/show120 and tell us about it, and when you come by the site, you’ll see all the links and notes that Daisuke and I talked about and much, much more in the resources section of the post.

But most of all, thanks for listening and thank you for telling people interested in Japanese startups and innovation about the show.

I’m Tim Romero and thanks for listening to Disrupting Japan.