Zidong Taichu

Zidong Taichu is a cross-modal general artificial intelligence platform developed by the Institute of Automation of the Chinese Academy of Sciences (IAAS), the core of which is the world's first graphic-text-audio (visual-text-speech) tri-modal pre-training model (OPT-Omni-Perception pre-Trainer).

Background and Significance::

• The development of Zidong Taichu marks a breakthrough in the field of artificial intelligence, especially in pre-trained models.
• The platform is based on multimodalLarge ModelAs the core, based on full-stack localized basic software and hardware platform, it can support the whole scene.AIApplications.

Core technology features::

• Cross-modal understanding and generative capabilities: Zidong Taichu is equipped with cross-modal understanding and cross-modal generation capabilities, and is able to perform multi-task joint learning without supervision and quickly migrate to data from different domains.
• Unified representation of the three modes: Through the introduction of speech modality, Zidong Taichu realizes the common graphic-text-phonetic-semantic spatial representation and utilization, and breaks through to directly realize the unified representation of the three modalities.
• Unique Application ScenariosIn particular, Zidong Taichu has made the "sound from picture" and "picture from sound" a reality for the first time, providing model-based support for more diversified scenarios, such as video dubbing, voice broadcasting, headline summarization, poster creation, and so on.

Milestones::

• On July 9, 2021, Zidon Taichu officially reported at the World Artificial Intelligence Conference (WAIC) 2021 Rise AI Summit.
• On June 16, 2023, the Institute of Automation of the Chinese Academy of Sciences (IAAS) released Zidong Taichu 2.0 in Shanghai, with significant improvements in decision-making and judgment capabilities compared to the first generation.
• March 5, 2024 - Wuhan Institute of Artificial Intelligence (WIAI) and Institute of Automation of Chinese Academy of Sciences (IACS) independently developed "Zidong Taichu" large model has been iterated to version 2.0, and it is expected that "Zidong Taichu 3.0" will be released in the first half of 2024, which is a new version of "Zidong Taichu". It is expected that "Zidong Taichu 3.0" will be released in the first half of 2024.

Markets & Applications::

• Zidong Taichu Big Model has passed the Interim Measures for the Administration of Generative Artificial Intelligence Services for the record, and can be officially online to provide services to the public.
• The platform has a wide range of application prospects in the fields of medical care, transportation and industrial production, and will play a greater role in these fields in the future.

Partners and impact::

• Newland, one of the founding partners of Zidon Taichu, is ranked number one in terms of algorithm quality in the relevant field.
• ZiDong Taichu 2.0 Omnimodal Large Model won the "Excellence in Artificial Intelligence Leadership Award", the highest award at the World Conference on Artificial Intelligence (WCAI) 2022, proving its leadership and influence in the field of Artificial Intelligence (AI).

To summarize, Zidong Taichu, as a masterpiece of Institute of Automation, Chinese Academy of Sciences, has not only made remarkable breakthroughs in technology, but also demonstrated a wide range of application prospects and great potential in the market.