This is an AI translated post.
Turing's pursuit of fully autonomous AI technology development in Japan
- Writing language: Korean
- •
- Base country: Japan
- •
- Information Technology
Select Language
Summarized by durumis AI
- Turing, a Japanese AI company, is developing its own AI semiconductor and software, aiming for commercialization of fully autonomous vehicles by 2028. To overcome the limitations of traditional rule-based systems, they are introducing a new approach utilizing AI generative models.
- Turing is developing the 'Hummingbird' semiconductor, which minimizes semiconductor capacity and power consumption for deploying high-performance AI models in vehicles, and the 'Heron' AI software, capable of handling various input modes such as images, videos, text, voice, and actions.
- Turing is conducting the 'Tokyo30' project, aiming to achieve 30 minutes of human-free driving in the Tokyo area by 2025. They are also dedicated to securing high-quality autonomous driving data and developing new data processing technologies.
Turing, a Japanese artificial intelligence (AI) company, is developing AI semiconductors and software with the goal of commercializing fully autonomous vehicles by 2028. Turing is seeking a new AI-based approach to overcome the limitations of existing autonomous driving technology.
Existing autonomous driving technologies have been rule-based systems that operate according to predetermined rules depending on the situation. However, this method has limitations in responding to unpredictable and complex situations. Turing is exploring the use of AI generative models to overcome this. The idea is that by recognizing the surrounding situation with a camera and inputting it into the AI model, the vehicle can be controlled based on the results.
To this end, Turing is developing an AI semiconductor called 'Hummingbird'. Hummingbird minimizes semiconductor capacity and power consumption to allow high-performance AI models to be embedded in vehicles. It features optimization for quantization and integer operations rather than precise numerical calculations. Currently, development has begun with experimental semiconductors based on FPGAs, with mass production targeted for 2028.
In the AI software field, Turing is developing a multi-modal generative AI model called 'Heron'. Heron has the ability to handle various input modes, including images and videos, as well as text, sound, and behavior. It has a 'world model' that understands and predicts the world, providing the advantage of being able to respond to unexpected situations.
Meanwhile, Turing is also undertaking the 'Tokyo30' autonomous driving project, which aims to achieve 30 minutes of autonomous driving in the Tokyo area by 2025, using only cameras and AI without human intervention. To achieve this, Turing is also working hard to acquire high-quality autonomous driving data, and is developing new data processing technology that combines driving video data with map data.
Amidst the fierce global competition in AI technology, Turing is leading AI technology innovation for the commercialization of autonomous vehicles. Verification of actual driving performance is expected to be the key.