2024.01.26
Oracle unveils 'safer version of ChatGPT'
BY LEE JAE-LIM, Korea
JoongAng Daily - Texas-based software giant Oracle has rolled out a new global
cloud computing offering with built-in generative AI, the company announced on
Thursday.
Dubbed Oracle Cloud
Infrastructure (OCI) Generative AI, the system designed for enterprises is
equipped with two large language models (LLMs) — Meta’s Llama 2 and Cohere,
named after Toronto-based AI company Cohere — to address a wide range of
businesses in customer service, marketing, product development and
cybersecurity.
Just like everyday
users utilize ChatGPT or other LLMs to define, analyze and code data,
enterprises can turn to OCI Generative AI for help. For instance, in customer
service, the models can automatically record and transcribe the phone
conversation or collect data on customer feedback for later improvement.
Oracle explains the
service as a full-stack tool kit that can help enterprises solve business
problems smarter, faster and in a cheaper way.
“When OpenAI’s
ChatGPT came out, it was phenomenal, but companies were hesitant to use it
because they feared their private data be exposed,” said Oracle Korea’s
Managing Director Tom Song, at a press event to introduce the company’s latest
services on Thursday in southern Seoul. “Oracle saw the demand right there and
dove into development. Our clients utilizing our service can now address
business issues such as text generation, summarization and semantic similarity
tasks within their own Oracle system without having to reach out to exterior
LLMs.
“And of course,
Oracle is bound to confidentiality to not to share enterprise data with anyone
else, including our own clients,” Song added.
The service, in
time, will be able to support some 100 languages and can be utilized in the
Oracle Cloud and on-premises in regions where Oracle data centers are
established, including Korea.
Oracle currently
operates two data centers in Korea, in Seoul and Chuncheon, Gangwon.
The company also
released the beta version of OCI Generative AI Agent, which goes a step further
than traditional LLMs. This advanced iteration enables agents to autonomously
access and train themselves using enterprise data for more sophisticated applications.
“This is possible
through Oracle’s Retrieval Augmented Generation [RAG] technology,” Song said.
“Basically, think of generative AIs as a really smart university graduate, but
when they are rookie employees, they might be clumsy or uncertain of what to do
because they are new to the company. Oracle’s RAG technology equipped in this
agent will allow it to quickly learn enterprise data and consistently improve
its functions as it continuously engages in machine learning.”