2024.05.17
Amazon Web Services promises energy efficiency through cloud computing, generative AI
BY
LEE JAE-LIM, Korea JoongAng Daily - Korean clients using cloud computing
solutions and AI chips from Amazon Web Services (AWS) can prioritize
sustainability and cost optimization through generative AI, AWS said on
Thursday.
Korean
clients are particularly interested in reducing their carbon footprint, which
can be achieved using the AWS Graviton processor, designed to offer a
high-performance computing environment with an emphasis on energy efficiency.
“Korea has the second-largest user base within
the Asia-Pacific region that utilizes Graviton,” AWS Korea Director Ham Kee-ho
noted at the AWS Summit Seoul 2024 hosted in southern Seoul’s Coex. The annual
event gathers IT experts and business leaders to showcase instances of clients
utilizing AWS services.
Another
realm of interest for Korean companies is minimizing costs when utilizing AI.
Local demand for two of AWS’s homegrown AI chips, AWS Trainium and AWS
Inferentia — both designed to optimize graphics processing unit (GPU)
performance and reduce costs — is swiftly on the rise,
according to Ham.
“Domestic virtual human startup Klleon is a
representative example of using Inferentia,” he said.
SK
Telecom, a major Korean mobile carrier, is another AWS client currently
utilizing Amazon Bedrock, a fully managed cloud service that offers a range of
high-performing foundation models available for use through a unified
application programming interface (API).
Using
Bedrock, clients can easily develop AI applications by fine-tuning existing
large language models (LLMs) with their own data and adding functions like
retrieval-augmented generation (RAG) technology.
“We are developing telco-specific LLMs through
fine-tuning LLMs and using RAG technology on Bedrock,” said SK Telecom’s Chung
Suk-geun, head of global AI tech division. “We plan to launch the telco LLM and
personal AI assistant service based on this product within the second half of
this year.”