How Japan’s largest AI unicorn is shaping the future of deep learning

Preferred Networks: Pioneering AI Innovation in Japan

Key Concepts:

• Deep Learning
• Artificial Intelligence (AI)
• Machine Learning
• Autonomous Driving
• Robotics
• Healthcare
• Supercomputers
• AI Processors (MN-Core)
• Energy Efficiency
• Large Language Models (LLMs)
• Open-Source AI
• Edge Computing
• Generative AI
• Unicorn (startup)
• IPO (Initial Public Offering)

1. Company Overview and Vision:

• Preferred Networks (PFN) is a Japanese tech startup founded in 2014, specializing in AI, deep learning, and machine learning solutions.
• Vision: To make the world a better place by making the latest technology available in the shortest possible time.
• Focus: Creating new value through computer science and addressing real-world issues across various sectors.

2. Addressing Real-World Challenges with AI:

• PFN is working on a wide range of issues including cars, robots, and medicine.
• Challenges include data problems, underdeveloped AI technology, and lack of societal demand.
• The company has been solving these problems one by one over the years.

3. Strategic Collaboration with Toyota:

• Toyota invested 11.5 billion yen into Preferred Networks.
• Focus: Applying AI to the field of automated driving.
• Specific application: Addressing Japan's trucking industry challenges, including an aging workforce and long working hours.
• Investment in T2: A joint venture with Mitsui, focusing on automated driving of trucks.

4. IPO Ambitions and Funding:

• Goal: To list on the stock market (IPO) within 3-5 years.
• Reason: To raise large-scale funding, especially for hardware development, which requires significantly more investment than software.
• Investment is also necessary for software development.

5. MN-Core AI Processor Series:

• Developed in collaboration with Kobe University.
• Achieved top ranking on Green500's list for energy efficiency.
• Hardware focus: Maximizing computing power within a limited number of transistors using a unique architecture.
• Software focus: Tuning and optimization to create a high-performance chip.
• Competitive advantage: By only installing what is necessary for the use of AI.

6. Competition with NVIDIA:

• PFN aims to gain a competitive advantage by focusing on deep learning and AI-specific processors.
• Emphasis on performance and efficiency per unit of power.
• Addressing the challenge of increasing power consumption and heat generation in next-generation semiconductors.
• Developing chips that handle data transfer and require a lot of power.
• PFN aims to break into the field of AI chips, currently dominated by NVIDIA.

7. Response to China's DeepSeek:

• PFN recognized DeepSeek's potential even before it gained widespread recognition.
• Belief that even without the most advanced hardware, cutting-edge models can be created by improving software.
• Expectation that various players will enter the world of chips and continue to create new products.

8. Japan's Competitiveness in AI:

• PFN aims to make Japanese industry more competitive through its work in semiconductors and AI.
• Goal: To provide products and services to the world.

9. Plammo: Open-Source Large Language Model:

• Built from scratch by Preferred Networks.
• Excels at Japanese-specific tasks.
• Supports multiple languages, including English (supports 30 languages).
• Focus: Edge computing applications, particularly in robotics and automotive industries.
• Vision: To leverage generative AI technology to transform these industries.

10. Leadership and Management Style:

• Similarities in management styles between the founders.
• Evolution of management style: From a flat organizational structure in the early stages to a more structured organization as the company grew.
• Okanohara's approach: Trusts and delegates to others, promotes talented individuals suitable for specific projects.
• Nishikawa's approach: Gets into the details and manages things personally, especially regarding important matters like semiconductors or AI.

11. Attracting Talent:

• Company motto: "Learn or die," emphasizing the importance of continuous learning and flexibility.
• Focus on hiring individuals with a willingness to learn.

12. Lessons Learned:

• Challenge: Maintaining a unified direction as the company grew rapidly after receiving investment.
• Solution: Making painful decisions to realign the company's focus.

13. Future Vision:

• Global success: Aiming for widespread adoption of PFN's processes and applications in the global market.
• International expansion: Establishing bases in other countries and understanding local cultures to compete on the world stage.
• Addressing major global problems: Contributing to solving significant societal challenges through AI.

14. Notable Quotes:

• Nishikawa: "We are trying to create a lot of new value in the world by focusing on computers and computer science which we love."
• Nishikawa: "Our vision is to make the world a better place by making the latest technology available in the shortest possible time."
• Okanohara: "One of our company MOS is learn or die which expresses our stance on learning being a specialist is good but when it comes to hiring we look at whether the person has a willingness to learn and is flexible."

Synthesis/Conclusion:

Preferred Networks is a leading Japanese AI company with a clear vision to solve real-world problems through deep learning and AI. Their strategic partnerships, particularly with Toyota, and their development of innovative technologies like the MN-Core AI processor and the Plammo LLM, position them as a significant player in the global AI landscape. While facing competition from established giants like NVIDIA and emerging forces like China's DeepSeek, PFN's focus on energy efficiency, edge computing, and open-source models, combined with a strong emphasis on continuous learning and adaptation, provides a pathway for continued growth and global impact.