Artificial Intelligence & Machine Learning , Next-Generation Technologies & Secure Development , Video
Training LLMs on Private Data for Higher Accuracy
Rodrigo Liang of SambaNova Systems on Large Language Models, Industrial Revolution
Large language models have revolutionized various industries by automating language-related tasks, enhancing user experiences and enabling machines to communicate more naturally with human beings, according to Rodrigo Liang, CEO of SambaNova Systems.
See Also: The SIEM Selection Roadmap: Five Features That Define Next-Gen Cybersecurity
Rapid adoption of LLMs is contributing to significant industrial revolution. These models such as ChatGPT have been integrated by numerous individuals and businesses, leading to the creation of new processes, products and services, he said.
"The most private data in your company is your most valuable data. Training these models on your private data gives you the highest accuracy," Liang said. "Businesses have a lot of private data, some of which they can't disclose due to GDPR or clients' privacy. Yet, they want to train [LLMs] in a place that allows them to be able to gain the knowledge and insight that the data provides."
In this video interview with Information Security Media Group at Black Hat USA 2023, Liang also discussed:
- Data privacy challenges when it comes to training LLMs;
- Customizing LLMs for specific business needs;
- Defensive measures for organizations to ensure the safety and control of LLMs.
Liang has nearly three decades of experience in the cybersecurity industry. His areas of expertise includes cloud computing, multiprocessors, enterprise software, SaaS, SOC, among others. Prior to joining SambaNova, he was the senior vice president at Oracle Corp.
