Larry Ellison, Chairman & CTO at Oracle, announced the launch of a new generative AI cloud service during the company’s Q4 2023 quarterly earnings call.
In partnership with Cohere, Oracle is adding an AI cloud service to protect the privacy of enterprise customers’ training data, allowing these customers to securely utilize internal data to enhance their own large language models (LLMs).
Such specialized LLMs will likely play a big role in the future of generative AI.
Already the healthcare industry has already reportedly seen the benefits of specialized LLMs through the speed of uncovering life-saving drugs, improving patient care, and increasing access to healthcare with reduced costs.
After noting this, Ellison added: "Specialized large language models will be instrumental in helping highly trained professionals use their precious time more efficiently.
Cohere and Oracle are working together to make it very, very easy for enterprise customers to train their own specialized large language models while protecting the privacy of their training data.
"Over the next few years, lots of companies are going to train their own specialized large language models."
Large Language Models
An LLM is a language model comprising a neural network with multiple parameters that are trained through unlabelled text using self- or semi-supervised learning.
Such models are the underlying technology behind recent AI breakthroughs like ChatGPT.
Oracle’s latest generative AI cloud service allows customers to make their LLMs secure enough to add their own data, which is timely following high-profile data breaches and compliance infringements resulting from ChatGPT.
Rebecca Wetteman, Principal at Valoir, predicts that such cloud services are the future for LLMs in the enterprise.
"Maybe the internet is not the best place to train your model," she said during CX Today’s latest BIG News Show
Vendors will use a combination of large data sets, smaller private ones, specific use cases, industry data models, and all those sorts of things to be very clear about where the parameters are and where the data lives.
"That is going to be an important part of security, training, and making the most effective models possible."
