NVIDIA has launched a new suite of NVIDIA NIM microservices designed to enhance the deployment of generative AI applications with advanced regional and cultural relevance. These innovations are set to bolster the development of sovereign AI systems across Japan and Taiwan, catering specifically to the unique linguistic and cultural contexts of these regions.
The newly introduced microservices are tailored to support regional language models, addressing the growing global trend of sovereign AI—a movement focused on leveraging domestic infrastructure, data, and expertise to align AI systems with local values and regulations.
In the Asia-Pacific region, generative AI software revenue is projected to soar to $48 billion by 2030, a dramatic increase from the $5 billion forecasted for this year, according to ABI Research.
The NIM microservices include:
- Llama-3-Swallow-70B: Trained with Japanese data to enhance understanding of local legal and cultural nuances.
- Llama-3-Taiwan-70B: Developed with Mandarin data to improve performance in regional language tasks.
- RakutenAI 7B: Built on the Mistral-7B model, offering Chat and Instruct services in English and Japanese. These models have recently achieved top scores in benchmarks for Japanese large language models.
These models enhance communication by accurately reflecting regional language and cultural subtleties, thus improving performance in legal tasks, question-answering, and translation.
Countries like Singapore, the UAE, South Korea, and Sweden are heavily investing in sovereign AI infrastructure, reflecting a broader global trend.
The new NVIDIA NIM microservices are optimized for deployment with NVIDIA AI Enterprise and the TensorRT-LLM library, providing up to five times higher throughput and reduced latency. These microservices are now available as hosted APIs, facilitating the development of advanced chatbots, copilots, and AI assistants.
NVIDIA’s new offerings support various sectors, including healthcare, finance, and education. For example, the Tokyo Institute of Technology has fine-tuned Llama-3-Swallow-70B for Japanese applications. Preferred Networks is leveraging the model to create a healthcare AI, while Chang Gung Memorial Hospital in Taiwan is utilizing Llama-3-Taiwan-70B to enhance medical communication and patient care.
Taiwanese electronics manufacturer Pegatron and other global firms like Chang Chun Group and Unimicron are adopting the new models to improve operational efficiency and develop custom AI applications.
NVIDIA AI Foundry is also available, providing a comprehensive platform for customizing foundation models and creating regional NIM microservices. This service offers tools and support for enterprises to tailor AI solutions to their specific needs, ensuring culturally and linguistically appropriate outcomes.
With these advancements, NVIDIA is solidifying its commitment to supporting sovereign AI initiatives and fostering the development of regionally relevant AI technologies.
Related topics:
What Is Emotion Classification NLP?