Turing Company Deck Nobember 2024ver

Transcript

1. © 2024 TURING, INC. Autonomous Driving with E2E and Generative AI September 2024 COMPANY DECK
2. 2 CONTENTS TURING COMPANY DECK | CONTENTS 01 Company
 02 Mission
 03 Projects
 04 Business
 05 Careers

3. TURING COMPANY DECK | COMPANY 3 Turing at a Glance
 ※2024年8月時点
 Company
 Turing Inc.
 Founded
 2021/8/20
 What We Do
 Development of fully autonomous driving 
 (level 5) technology
 Headquarters
 East Tower 4F, Gate City Osaki, 1-11-2 Osaki, Shinagawa-ku, Tokyo 141-0032
 CEO
 Issei Yamamoto
 Total Funding
 $40MM
 # Employees
 48 employees
 COMPANY 3
4. The Shogi AI “Ponanza,” which he developed, achieved its first victory against a Shogi Grandmaster.
 
 At the AI company HEROZ, he successfully led the company through IPO as the lead engineer.
 
 To build a generational company as his next challenge, Yamamoto founded Turing in 2021.
 4 TURING COMPANY DECK　 | CEO Co-Founder, CEO
 Issei Yamamoto
 4 *shogi ≒ Japanese Chess

5. 5 Mission Achieve Fully Autonomous Driving TURING COMPANY DECK | MISSION 

6. Turing aims to achieve level 5 “fully autonomous driving.”
 The goal is to develop a system that completely replaces human driving.
 6 TURING COMPANY DECK | MISSION Driver Driving Automation Definition Autonomous Driving System Level 5 Under all conditions and areas, system drives（fully autonomous driving） Need technological breakthrough Level 4 Under limited conditions and areas, system drives (Human driver is NOT required) Early experimentation started in the U.S. and China Human Level 3 Under limited conditions and areas, system drives (Human driver is required) Already being sold and starting to become widespread in the world Level 2 Support brake/acceleration AND steering Level 1 Support brake/acceleration OR steering Level 0 No automation - Levels of autonomous driving

7. 7 Point cloud data obtained from LiDAR HD maps (High-definition 3D maps) Waymo robotaxi equipped with LiDAR TURING COMPANY DECK | MISSION As exemplified by Google subsidiary Waymo, 　　　　　　　　　　　　　　　the mainstream approach in conventional autonomous driving combines LiDAR sensors and pre-acquired HD map data with a rule-based system.
 The mainstream approach combines LiDAR and HD maps

8. 8 Existing technology cannot handle the long tail situations
 Many / Easy How do you define slight situational differences with rules? Difficulty of Driving Frequency Few Difficult Easy Many Few / Difficult Left：Proceed according to the instructions of the traffic worker Right：Need to be cautious of the pedestrian but no need to follow his instructions TURING COMPANY DECK | MISSION In traffic environments, there are situations that are infrequent, diverse and challenging, i.e., long tail situations.
 It is impossible for humans to comprehensively define all traffic situations using rules, and thus with the currently mainstream rule-based approach, achieving “fully autonomous driving” is impossible.

9. 9 PROJECT 01 End-to-End Autonomous Driving TURING COMPANY DECK | PROJECT 

10. “AI without human intervention will surpass rule-based system.”
 
 
 
 10 The solution is an AI model without human intervention
 The Shogi AI “Ponanza,” developed by CEO Yamamoto, improved at a pace surpassing rule-based system through machine learning. In 2017, it became the first in Japan to defeat a reigning Shogi Grandmaster.
 CEO Yamamoto and the Shogi AI “Ponanza” Performance Technological Progress Today Exponential growth AI Model TURING COMPANY DECK | PROJECT Linear growth Rule-based Model
11. Autonomous driving is rapidly shifting to End-to-end (E2E)
 Moving away from human-defined rule-based system to an AI-controlled E2E system.
 
 
 Camera Lidar HD Map Lane Recognition Sign Recognition Object Recognition Self Position Rule-based driving instructions Car End-to-end AI＆Data Centric Camera Image Car Neural Network Effective Scalable Easy to Maintain Rule-based Sensor & Map Centric （No HD Map） TURING COMPANY DECK | PROJECT 11
12. In just 1.5 years since founding, we developed the ADAS※ whose performance is comparable with the ADAS installed in production cars sold in Japan.
 Next, we will develop fully autonomous driving system.
 12 TURING COMPANY DECK | PROJECT Turing development progress to date
 ※ADAS＝Advanced driver-assistance systems. Driving in private area Autonomous driving on a designated route within the test course 2022.01 2023.01 2025.12 Tokyo30
 Sold one car equipped with our ADAS (level 2) Our ADAS on public roads Turing’s own ADAS comparable with ADAS sold in the market Autonomous driving on public roads Driving in Tokyo for over 30 minutes without human intervention
13. Reference
 Significant Breakthrough of “Tesla FSD v12” Wayve in the U.K. raised $1Bn in 2024/5 Europe Even in complex situations, such as turning at intersections or coordinating with pedestrians, it can make natural, human-like driving decisions. It demonstrated a high level of flexibility not found in conventional autonomous driving systems that follow predefined rules. Human-like driving decisions The fact that it was achieved without using HD maps, which were considered essential for autonomous driving, and relying solely on cameras for sensors, was a significant breakthrough. Furthermore, 300,000 lines of C++ code were scrapped, and the system was implemented using neural networks. Achieved without HD maps or LiDAR China The world is rapidly shifting from rule-based system to E2E system With large scale investments, rapidly catching up to Tesla 13 TURING COMPANY DECK | PROJECT
14. By the end of 2025, we aim to develop an end-to-end (E2E) autonomous driving model capable of driving on Tokyo's complex roads for over 30 minutes without any human intervention.
 TURING COMPANY DECK | PROJECT Tokyo30
 2025/12 Demonstration Experiment 14
15. TURING COMPANY DECK | PROJECT MLOps platform is the key to success
 公道での自動運転 東京都内を30分以上 人間の介入なしで走行する ”Tokyo30” MLOps Platform Data Collection Model Development Creation of Dataset Just as quality cars are produced from quality factories, a quality autonomous driving model is created from a quality MLOps platform.
 Check out our tech blog for more info ▶ https://medium.com/@turingmotors/e2e-series-part-1-a-new-paradigm-in-end-to-end-autonomous-driving-db4d901851e8 15
16. Structure of E2E autonomous driving model
 16 TURING COMPANY DECK | PROJECT Neural Network 3D object recognition & motion prediction Map recognition Vector representing the world Path planning Control Multi-Camera Images Various subtasks, such as map recognition and 3D object recognition, are performed based on multi-camera input. The future path is then generated as output, enabling autonomous driving.

17. 17 PROJECT 02 Generative AI TURING COMPANY DECK | PROJECT 

18. GenAI will enhance E2E autonomous driving
 18 TURING COMPANY DECK | PROJECT 公道での自動運転 東京都内を30分以上 人間の介入なしで走行する By equipping AI that has driving capabilities with human common sense, it leads to autonomous driving that is closer to human-like behavior.
 
 
 
 
 
 
 Neural Network Common Sense / Understanding of the World
 VLM(Vision Language Model) A multimodal GenAI that interprets image data from cameras
 World Model
 A model that simulates how a specific state will change 
 in the future and generates it as a video
 Driving Ability
19. To achieve fully autonomous driving, we have developed “Terra,” a video- generating AI that understands and predicts complex real-world situations.
 
 
 
 
 
 TURING COMPANY DECK | PROJECT A video generated by Terra. Terra is specialized for driving environments and can generate driver perspective videos from in-vehicle cameras. (Left) Generated video following the green trajectory (Right) Generated video following the red trajectory 
 The model can also accept prompts and generate different situations Generates the continuation of a short video using it as an input World Model ”Terra”
 19
20. Gaggle-Cluster-1 scheduled to go live in 2024/9 TURING COMPANY DECK | PROJECT Organizing the training environment for AI development
 In future GenAI development, securing and utilizing GPU resources will be key.
 
 
 
 
 
 
 
 
 GENIAC Project (2024.2-8) ▪ Japanese gov’t supports leading AI companies in Japan ▪ A large number of H100 are available for recipients https://www.meti.go.jp/policy/mono_info_service/geniac/index.html Turing’s own GPU cluster “Gaggle-Cluster-1” (2024.9-) ▪ H100 96 notes ▪ Maximizing inter-node comm. and storage I/O ▪ An HPC-style job management system 20
21. 21 BUSINESS Business Model & Organization TURING COMPANY DECK | BUSINESS 

22. Vehicle Driving Control E2E AI Model Functional Safety Mechanism ADAS ECU AD/ADAS Starting in 2025, we will provide an MLOps platform to auto OEMs and support the integration of E2E models into mass-produced vehicles.
 TURING COMPANY DECK | BUSINESS Providing MLOps platform to auto OEMs (2025 ~)
 MLOps Annotation, Curation Model Deployment on the Road Data Collection E2E MLOps Platform A platform that provides end-to-end support from preparing training data, model design, training, to validation ML Model Development Camera Car ADAS Development by OEMs Support 22
23. 　
 TURING COMPANY DECK | BUSINESS Why Will Turing Win?
 
 02 Fully autonomous driving cannot be achieved by simply developing AI software. The integration with hardware is essential, and it is precisely because Japan possesses advanced technical capabilities and a robust supply chain that we can take on the challenge of developing fully autonomous driving. To attract top AI engineers, it is essential to have compelling goals, the right approach to achieve them, and strong leadership. Under the leadership of CEO Yamamoto, who has a proven track record in AI development, Turing is taking on the grand challenge of fully autonomous driving. Strong auto industry in Japan 01 Strong Leadership in AI 23
24. TURING COMPANY DECK | BUSINESS Currently, the majority of sales in the automotive market come from hardware, but the software market, particularly around autonomous driving, is expected to expand.
 Reference｜Software in the automotive market is increasingly becoming more important
 Source: "Mobility DX Strategy (Draft)" <Summary Version> p.44 ※May 20, 2024 METI, Manufacturing Industries Bureau, Automobile Division, Mobility DX Office MLIT, Logistics and Automotive Bureau, Technology and Environment Policy Division https://www.meti.go.jp/policy/mon o_info_service/mono/automobile/ji do_soko/r6dxjimukyokushiryou2.p df 24 Sales forecast for OEM’s hardware 
 and software
 Sales forecast for OEM’s software 
 and electronics
 % of SDV (software defined vehicles) 
 sales over total sales
 
 
 Software sales
 Hardware sales
 Sales forecast for OEM’s software and electronics
 % of respondents
 USD MM
 USD Bn
 5% and <10%
 10% and <15%
 15% and <20%
 20% and <25%
 25% and <30%
 0-5%

25. Organization Chart
 TURING COMPANY DECK | ORGANIZATION 25 R&D Department CXOs E2E Development of E2E autonomous driving model Edge Computer Development of autonomous driving-related OS/applications GenAI Development of multi-modal AI, and construction and operation of large-scale GPU clusters Corporate Department Internal Control Legal and Risk Management CEO Office Recruiting Department Finance Fundraising, Financial Management Corporate / Backoffice Legal, Labor, Admin PR Business Development HR
26. Employees by Function
 Organization Structure
 TURING COMPANY DECK | ORGANIZATION 26 2021/8 2022/8 2023/8 2024/8 Finance BizDev PR Corporate Recruiting Semiconductor Edge Computer CXOs GenAI E2E 35 people 35
27. Management
 TURING COMPANY DECK | ORGANIZATION CFO Masato Morishima Co-Founder Shunsuke Aoki CTO Yu Yamaguchi CEO Issei Yamamoto COO Daisuke Tanaka CBO Yugo Tokuchi 27
28. Our Employees’ Background 
 TURING COMPANY DECK | ORGANIZATION 28 AI Talent Software Engineer Business / Corporate Kaggle Grandmaster Kaggle Master
29. TURING COMPANY DECK | WORK ENVIRONMENT Work Environment and Benefits
 29 You can work both at the office and from home. The work environment is set up so you can adjust your schedule according to family needs. Flex Time PC/Equipment Support Maternity, childcare, and caregiving leave system Monthly All Hands Regardless of your year of joining, you can take maternity leave, childcare leave, or caregiving leave at any time. We have a flex-time system with core hours from 10:00 to 15:00. By communicating with your team during core hours, outside of core hours, employees can work flexibly at their own discretion. Once a month, we hold a meeting to share the company's overall strategy and review each team's efforts and achievements together. (Light snacks and drinks provided.) Twice a month (biweekly), we host a happy hour. The cost of snacks and drinks is, of course, covered by the company, and employees' friends or acquaintances are also welcome to join. We allow you to choose your PC upon joining. All office desktops are equipped with ultra-wide displays, and we will continue to support an environment where you can focus on your work. Hybrid Work Happy Hour Night
30. TURING COMPANY DECK | WORK ENVIRONMENT Work Environment to Boost Your Engineering Skills
 30 Grand Challenge for Mankind Overseas Conference, Speaking Engagements You will have the opportunity to travel overseas to experience the latest autonomous driving systems and to attend top conferences such as CVPR. Many new graduates have also had overseas business trips, allowing them to gain a global perspective in their R&D work. We also actively participate in domestic conferences such as AWS Summit, providing opportunities to present in these conferences. 30 Turing is the only company in Japan developing an “End-to-End” autonomous driving system (Tesla in the U.S., Wayve in the U.K.). To achieve the grand challenge of “fully autonomous driving,” we all must take on bold challenges. The development of autonomous driving with GenAI is an area that is rapidly gaining attention worldwide. The knowledge and experience you gain in this field will become invaluable in your career. Top Engineering Talent and Vast Compute Resource Turing has a diverse engineering organization that brings together top engineers, including: ▪ Management team with strong expertise in AI and software ▪ Engineers from leading tech companies / Japanese manufacturers Additionally, the amount of compute resources available to each engineer is among the highest in Japan, as we have been selected for the METI’s GENIAC and have built our own GPU cluster (96 H100 nodes).
31. 31 TURING COMPANY DECK | CAREERS 
 CAREERS Hiring Process
32. TURING COMPANY DECK | VALUES 32 GO GO Turing is an organization that values positivity, energy, and kindness. We will never hire or recognize individuals who negatively impact those around them, no matter their skills or experience.
 Turing takes on big challenges and aims to make a significant impact in the world. 
 Remember that you are here at Turing to take on these big challenges, and let's embrace the spirit of 'If it's not a NO, then it's a GO.
 Turing values the safety and health of its employees. Please follow safety rules and always be mindful of preventing accidents and disasters. 
 As we take on long-term challenges, we want you to play the long game by taking care of your health as well.
 Be a Good Person Safety First OUR VALUES Communication as a Shared Responsibility Turing is a place where people from diverse backgrounds work together. To build great products as a united team, we must strive for clear communication, both as speakers and listeners.

33. ▪ ML Researcher ▪ ML Engineer ▪ Software Engineer ▪ Data Engineer ▪ GPU Cluster Engineer ▪ Hardware Engineer ▪ QA Engineer ▪ Procurement ▪ Engineer (New Graduate) 　　Engineering Open Positions ▪ Business Development ▪ Public Affairs ▪ PR ▪ Executive Assistant ▪ Recruiter ▪ Recruiting Operation Leader ▪ Recruiting Marketer ▪ Recruiting PR Lead 　　Business/Corporate Open Positions Open Positions
 TURING COMPANY DECK | CAREERS 33
34. TURING COMPANY DECK | CAREERS 34 Recruiting Process | Full Time
 
 
 Flexible Selection Process The selection process outlined on the left is just the standard process. We can adjust for interviews, office tours, or Open Office events based on your individual circumstances. Of course, we are also open to discussions if you would like to expedite our selection process to align with other companies' selection timelines. Additionally, for some positions, we may include technical tests or work samples as part of the process. The typical selection period is around 2 to 3 weeks. 1. Casual / Informational Discussion 2. Resume Screening 3. Interviews (1~2) + tech tests 4. Trial Employment (0.5~1 day) 5. Final Interview 6. Job Offer
35. Recruiting Process | Internship
 
 
 TURING COMPANY DECK | CAREERS 35 About Trial Employment In the trial employment program, you will be granted the same level of access to internal information as regular employees, including tools like Slack and Notion, and work alongside the team you are assigned to for up to two days. This process allows you to experience Turing's culture first hand and ensures that there is no mismatch between both parties. 1. Casual / Informational Discussion 2. Resume Screening 3. Interviews (1~2) 4. Trial Employment (Max. 2 days) 5. Final Interview *Could happen before/after trial employment 6. Job Offer
36. TURING COMPANY DECK | CAREERS 36 Stock Option Plan (Currently in Planning)
 ▪ Turing goes public ▪ You must be a full-time employee at the time of SOs grant ▪ In general, SOs granted are tax-qualified SOs ▪ 1 year cliff, 5 year vesting (20% vest each year) ▪ Upon resignation, you can retain the vested SOs Stock options are a type of incentive that grants employees the right to purchase company shares at a set price. At Mercari, one of Japan’s leading startups, over 30 employees are said to have gained assets worth over 600 million yen. Turing was also founded with capital acquired by CEO Yamamoto from the IPO of his previous company, HEROZ, and the utilization of SOs. At Turing, we aim to distribute SOs to as many members as possible, so we can produce many next-generation of entrepreneurs who in turn create new industries and new economic opportunities. Since the SO plan is currently in planning phase, those who will join Turing are highly likely to be granted these rights. We are considering the conditions for SOs as outlined below. Additionally, detailed information about the stock option plan is documented and made available to employees. Please note that as the number of employees increases or the business phase changes, there is a possibility that stock option plan may change.
37. © 2024 TURING, INC. https://tur.ing/en
 WE ARE HIRING!