Riskgaming

Why Tokyo’s Sakana AI is pioneering a new vanguard of national AI foundation models

Lux announced big news today: we are leading a $30 million founding seed round into Sakana AI, a Tokyo, Japan-based AI research laboratory that uses evolutionary methods, collective intelligence and character-level training to radically accelerate the training and development of nature-inspired AI foundation models. It’s a marquee check for Lux into the Asia-Pacific region, and represents the continuing democratization of the frontiers of computer science to all regions of the world, a trend we’ve championed for years now.

I, your host Danny Crichton, wanted to spend some time on the economic and technological milieu that is changing the face of startups and entrepreneurship globally in 2024. In this episode, I walk through three themes that are driving Lux’s interest in Sakana AI and other companies, including the rise of national and indigenous AI foundation models, the return of Japan’s dynamic economy after decades of stagnation, and the broader ambitions of the Asia-Pacific region as China recedes from the minds of international investors in the midst of President Xi Jinping’s crackdown on private entrepreneurship.

Produced by ⁠⁠⁠⁠⁠⁠Christopher Gates⁠⁠⁠⁠⁠⁠

Music by ⁠⁠⁠⁠⁠⁠George Ko

Transcript

This is a human-generated transcript, however, it has not been verified for accuracy.

Today is a big deal for Lux, as we're coming out with one of our largest investments ever in the Asia Pacific. We announced that we are leading a $30 million founding seed round into Sakana AI, a research lab based in Tokyo, Japan that's developing nature-inspired foundation AI models. We'll have a lot more to say on Sakana in the weeks and years to come on the podcast, but I wanted in this short episode to highlight a few trends that are converging that made this investment possible.

First, we're seeing the rise of indigenous and national AI foundation models that take advantage of the unique cultures and talents of countries all over the world. Google and OpenAI are fundamentally intertwined with the region of Silicon Valley and Stanford and Berkeley that form its heart. In much the same way, DeepMind was formed in London when its founders met at University College London. Eventually, DeepMind would become one of the United Kingdom's most critical AI champions. Mistral in France has raised buckets of euros to build upon the deep bench of expertise at France's top polytechnic universities.

We're seeing similar startups begin to emerge all over the world, and now we have Sakana, which is taking advantage of the prodigious computer science experience at the University of Tokyo dozens of other Japanese technical institutions. Japan's digital app culture is a mix of local and international companies. It's one of the world's largest users of Twitter and Google services, but it also maintains its own popular payment companies like PayPay, social networks like Line, and news aggregators like SmartNews. That's a healthy balance, one that maintains Japan's entrepreneurial edge while also being open to the world.

As the internet and software have come to dominate more and more of our lives, countries are increasingly looking to own a part of that wealth. A decade ago in TechCrunch, I talked about how governments, and specifically the United States and China, were splitting the singular internet into the plural internets, focusing on data sovereignty and building legal and extralegal barriers to the free-flow of information that was the internet's origin dream. The digital world has never stopped splitting. The Eurorpean Union's GDPR law helped to usher in a distinct regulatory structure for data of EU citizens. South Korea requires that all mapping data about the country be held by local businesses, preventing companies like Google and Apple from offering competing products. India has outright banned TikTok and many other Chinese apps in the name of digital sovereignty. Similar laws that have splintered the internet have been passed in places like Australia, Canada, Brazil, Russia and throughout the Global South.

I'm an advocate for free and global competition, but I understand why these barriers are getting erected. Artificial intelligence is so important and so powerful that it is hard to argue that any one company or country should have a durable monopoly on such a crucial technology. Furthermore, these barriers actually create competition, ensuring that artificial intelligence technologies don't languish from the lethargy of the have-already-wons.

How many cutting-edge AI companies can the world sustain? In my view and in Lux's view, the number is higher than most people predict. There are clearly network effects across AI products, whether that's from developing unique training datasets or from the popularity of consumer brands and word of mouth. But AI is still in its infancy, which means that brand new ways of building AI models can quickly dominate the market if they are unlocked. That offers a unique edge to companies that are outside the so-called Cerebral Valley of San Francisco which take advantage of local knowledge and research and talent to do something different.

We expect to see a a new global axis of national AI hubs, each with their own strengths and expertises, which together offer the best AI services on the frontiers of what's possible.

Okay, so the first trend we are seeing is this rise of indigenous AI models. Second, our investment in Sakana AI is a major commitment to Japan and its startup ecosystem. Japan has been dominated by groups of keiretsu, networks of businesses and financial institutions that together dominate major Japanese industries like automotive and construction. That concentration of corporate power in a handful of companies has the benefit of scale and efficiency, a feature particularly needed by Japan as it rebuilt from the ashes of World War II. But that extreme integration also precludes the kind of frenetic innovation that's necessary to compete in the digital economy today.

Over the last decade, former prime minister Shinzo Abe and his successors including current prime minister Fumio Kishida have pursued a policy of what's sometimes dubbed Abenomics — a policy to juice Japan's economy through expansive supply-side business loans, deregulation and increased immigration. The goal is to take the general lethargy out of the economy that has existed since the Japanese real estate bubbled popped in the early 1990s and bring back the momentum of growth that the country experienced in the postwar period.

These policies have moved the needle, even if much more work has to be done. Japan's inflation rate has risen, spurring further spending and investment after decades of deflationary stagnation. A more liberal immigration policy has opened Japan's borders wider to talent of all stripes, from clerks at local conbini to top technology executives. And venture capital has expanded tremendously, with local firms like Geodesic and many others encouraging a whole new cohort of Japanese entrepreneurs to get underway.

Japan is an extraordinarily advanced and developed economy, and that means change can be hard. Japan has a lot to lose with policy mistakes, and given its wealth, arguably not nearly as much to gain. But stagnation has a draining effect on any culture, and we're seeing the greenshoots of entrepreneurial excellence that make all of us at Lux excited for Japan's long-term economic outlook.

So we've seen the rise of indigenous AI models and we're seeing Japan get its moment in the spotlight. The third and final trend I want to highlight is that we're seeing a growing appreciation in Silicon Valley for Asia-Pacific's wider economic dynamism. China's economic luster in the press has been thwarted by President Xi Jinping's centralization policies the past decade, which has increasingly seen whole swaths of the Chinese economy decimated in the pursuit of common prosperity and other central Beijing policies. That leaves open the door to many other countries in the region to breathe and capture the attention of international entrepreneurs and investors.

Japan is obviously one candidate as I was just describing, but there are many others. South Korea has been an innovation powerhouse for years, particularly in online gaming, where a number of its companies dominate the global industry. Just this past week, the country announced that it will offer a digital nomad visa, which will allow itinerant workers the ability to stay legally in the country for two years while they work for international companies. Heading south, Taiwan has been discussed ad nauseam in the press in the context of TSMC's chip fabs, but there is also a large bench of talent on the island for entrepreneurship.

Singapore, Malaysia, the Philippines, Indonesia and the rest of Southeast Asia have been fervent adoptees of digital technologies, and they have become some of the most entrepreneurially dynamic countries in the world.

And then there's India. Coverage of India tends to skew political in the United States, where coverage of Narendra Modi maintains a consistently negative valence. The upshot though is that India's economy has never been stronger or growing faster than it has in the past few years. Infrastructure spending is booming, construction is seen in every major and minor city across the country, and Indian entrepreneurs want to own the next-generation of international startup juggernauts just as American founders started Google, Microsoft and Apple and Chinese founders started Alibaba, Tencent and ByteDance. One should never underestimate what intense energy and ambition can accomplish when it comes to innovation.

When we connect the dots here from national champions of AI, to Japan emerging from its economic doldrums, to the larger rise of Asia-Pacific, we are witnessing a shift in the world order of innovation. It used to be that most innovation took place within a stone's throw distance of Stanford University or perhaps Zhōngguāncūn in Beijing. But now we are seeing the rise of a world where any country has a shot at being a technological leader on the frontiers of science and innovation.

As VCs, that means we need to adapt to the world as it exists, open to more places and ideas and models for startup development and growth. Our rigor certainly hasn't changed, but widening our perspective means we can find the company that maybe couldn't have made it ten years ago, but is now taking advantage of a changing global macro landscape to capture winds in its sails.

All of which brings us back to Sakana AI. Sakana's two co-founders are both based in Tokyo, and enjoy not only the extreme quality of life that the city offers, but its incredible bench of talent. I've spent six weeks in the past year in Japan, and I'll just say that the country is so primed for startup growth in the coming years that no one should miss the signals. Sakana has pulled in early employees from some of the most ambitious AI labs the country offers, even as it seeks out global talent that wants to live in Japan as a choice. Most importantly, Sakana is pioneering new AI models and techniques from the ground up, unlocking more efficient ways of training and developing AI models that take their cue from evolutionary methods, collective intelligence and character-level training. This is the real deal.

It's not always easy to be the single fish swimming the opposite direction of everyone else. But that's exactly what Sakana wants to be, and it's built that mission into its name. Sakana means fish in Japanese, and its logo shows a solitary red fish swimming the opposite direction of a school of blue fish. That's the contrarian impulse that defines the best startups and the best leaders, and increasingly, more and more of the world.

continue
reading