NVIDIA will give its employees superpowers using AI to help
Hardware

NVIDIA will give its employees superpowers using AI to help develop chips

NVIDIA not only actively promotes generative artificial intelligence systems among customers, but also uses them to optimize its own business processes. Using the combined experience of the company’s developers over 30 years, the AI ​​chatbot helps novice engineers get answers to frequently asked questions without distracting more experienced colleagues from their work. In addition, it is proposed to use AI to generate program code and work with errors in chips.

    Image source: NVIDIA

Image source: NVIDIA

NVIDIA senior scientist Bill Dally spoke about this application of AI this week. In published Article details how NVIDIA engineers created their own large language model for internal use called ChipNeMo, trained on the company’s internal data to generate and optimize software and help people design chips.

The company has loaded its own 30-year archive of semiconductor component development documentation into the language model. As NVIDIA’s research director explained, in practice, senior developers spend a lot of time answering questions from their younger colleagues. If this function is entrusted to artificial intelligence, the most valuable employees will have more time to develop chips.

According to an NVIDIA representative, such a chatbot can achieve quite high efficiency with moderate development costs if more targeted information is loaded into the system, taking into account the company’s previous experience. By using system resources wisely, you can reduce the costs of implementing relevant projects. The chatbot helps engineers find the necessary documentation in the archive without distracting their colleagues.

Another promising area of ​​application of generative artificial intelligence in the development of NVIDIA chips is the writing of program code fragments. An AI code generator is already in development and will be integrated into existing chip design tools. AI can also help document defects in developed chips. The artificial intelligence system will quickly cope with this task, freeing developer resources for other operations.

“Our goal is not to automate a process or replace people, but to equip our existing employees with superpowers to increase their productivity.” – Bill Dally explains. And Mark Ren, director of research at NVIDIA and lead author of the paper, noted: “I believe that over time, larger language models will help all processes [разработки чипов]

In this example, NVIDIA demonstrated how the NeMo ecosystem can be used to optimize large language models used in semiconductors and other industries. NVIDIA customers and partners can use these tools to improve the efficiency of their own business processes. Fine-tuned, specialized language models can perform much better than more resource-intensive general-purpose models.

About the author

Dylan Harris

Dylan Harris is fascinated by tests and reviews of computer hardware.

Add Comment

Click here to post a comment