« Back to List High-Performance Computing Infrastructures and Technologies for Wide Adoption of AI in Taiwan
Time / Place:
⏱️ 09/27 (Fri.) 14:30-15:00 at R0 - International Conference Hall
Abstract:
With the booming demands for artificial intelligence applications on a wide range of enterprise work loads and products development, software vendors all over the world are rushing to develop many domain-specific applications that have great potential for customers in boosting either their internal productivities or external product sells. Taiwan is of course no exception. Even with a wide availability of open-source AI models in the past few years, Taiwan still needs AI models that are tailored to local language and societal elements. For this reason, advanced technologies such as continuous pre-training, fine-tuning, and domain specific RAGs play an important role in all the large-language models related development. Infrastructures capable of handling large models and large amount of training data are in urgent need for Taiwan in the next few years. Furthermore, technologies related do domain-specific model development need to be continuously refined and developed within HPC enabled data centers. For the obvious reasons in AI governance, many training data will be stored not only domestically, but also under a secure enclave such that sensitive information won’t be leaked to the third parties. National Center for High-Performance Computing (NCHC)’s recent endeavors in the above issues will be discussed in the talk. Opportunities and challenges are also laid out.
Biography: