"Conversing with Artificial Intelligence Models Can Aid in the Creation of Microprocessor Chips"

2023-06-07

Researchers at NYU Tandon School of Engineering have created a microprocessor chip through simple English "conversations" with an artificial intelligence model. This unprecedented achievement could accelerate chip development and enable individuals without specialized technical skills to design chips. In a study published on the arXiv preprint repository, the research team demonstrated how two hardware engineers interacted with ChatGPT-4 in standard English to design a novel microprocessor architecture. The researchers then sent the design to manufacturers for production. Typically, the development of any type of hardware, including microelectronic component chips that serve as the brains of electronic devices, begins with describing the tasks the hardware should perform in a conventional language. Trained engineers then translate this description into a hardware description language (such as Verilog) to create actual circuit components that enable the hardware to perform its tasks. In this study, the Language Model Microarchitecture (LMM) was able to generate usable Verilog through bidirectional dialogue. The subsequent chip manufacturing involved benchmark testing and fabrication of the processor using a process called Tapeout, conducted in the Skywater 130nm shuttle. Tapeout is a specific type of semiconductor manufacturing service accessible through Tiny Tapeout. "We believe that the results of this research enable us to manufacture physical chips for the first time using AI-generated HDL," said Hammond Pierce, Assistant Professor and member of the research team at NYU Tandon. "Some AI models, such as OpenAI's ChatGPT and Google's Bard, can generate software code in different programming languages, but their application in hardware design has not been widely studied. This research demonstrates that AI can play a beneficial role in hardware manufacturing, especially when you can engage in a back-and-forth dialogue to refine the design." The research team at NYU Tandon also includes Professor Ramesh Karri, Associate Research Professor Siddharth Garg, and PhD student Jason Blocklove. They used LMM to handle eight hardware design examples, specifically generating Verilog code for functional and verification purposes, and then focused on in-depth case studies for chip manufacturing. Previously, researchers had tested using LLM to convert English into Verilog, but they found that interacting with field engineers produced the best results. According to the researchers, implementing LLM conversations in chip manufacturing can reduce human errors in the HDL conversion process, promote productivity, shorten design and market time, and allow for more creative designs. This process they developed can also eliminate the need for chip designers to have a proficient understanding of HDL, which is a relatively rare skill and a significant barrier for individuals interested in pursuing such work. The researchers note that further testing is needed to identify and address security considerations involved in using AI for chip design. Source: https://techxplore.com/news/2023-06-conversations-ai-microprocessing-chips.html

RECENT AI NEWS

RECENT AI TOOLS