(Reuters) – Nvidia has released new research on the use of chatbots in the semiconductor design process that can generate human-like responses.
Designing modern chips, which are made up of billions of transistors, is a complex task that can take up to two years to complete with the involvement of thousands of engineers. Nvidia, known for its complex chips, has now applied its expertise to chatbot technology.
In their research, Nvidia used a large language model typically used in chatbots and augmented it with 30 years of chip design data from their archives. The aim of this application was to utilize the company’s extensive experience in answering questions.
According to Nvidia’s chief scientist, Bill Dally, the company found that the inclusion of specific data from their design experience improved the accuracy of the chatbot. This approach can also help control costs.
Additionally, Nvidia demonstrated the use of artificial intelligence (AI) to generate code. Dally explained that engineers often spend a significant amount of time identifying and troubleshooting non-functional components within a chip. By using AI to generate scripts for testing tools, engineers can save time and increase productivity.
Dally clarified that the goal of these advancements is not to replace human designers, but to enhance their abilities and make them more productive.
“It turns out a lot of our senior designers spend a fair amount of their time answering questions from junior designers,” Dally told Reuters. “So the protocol was a junior designer asks the chatbot. This can save senior designers a huge amount of time.”
The research findings indicate that even a basic chatbot can surpass an advanced one by incorporating specific data from the company’s experience. This discovery aligns with Nvidia’s objective of efficiently utilizing its resources.
Nvidia’s developments in using chatbots and AI for chip design mark significant progress in the technology industry. These innovations have the potential to improve efficiency, reduce costs, and empower engineers to accomplish more in less time.
Stephen Nellis reporting from San Francisco; Editing by Marguerita Choy
