#098a Now Generative AI Is Being Used to Design Chips That Make Generative AIs Work
Fresh & Hot curated AI happenings in one snack. Never miss a byte 🍔
This snack byte will take approx 4minutes to consume.
AI BYTE # 📢: Now Generative AI Is Being Used to Design Chips That Make Generative AIs Work
⭐Chip design is one of the most complex and time-consuming engineering tasks, especially as the demand for advanced chips grows in applications such as AI, high-performance computing, and autonomous vehicles.
Traditional methods of chip design can take months or even years to complete, and require a large team of highly skilled engineers.
But what if there was a way to speed up and simplify the chip design process, using the same AI systems that power ChatGPT, the popular chatbot that can generate realistic and engaging conversations?
That is the vision of some of the leading players in the semiconductor industry, who are using generative AI to automate and optimize various aspects of chip design.
Generative AI is a branch of AI that can learn from data and create new content, such as text, images, audio, or video. By using Large Language Models (LLMs), Gen AI can understand the structure and patterns of natural language, and generate coherent and relevant responses to queries or prompts.
In chip design, Gen AI can be used to write hardware code, verify design correctness, explore design alternatives, and provide guidance and assistance to engineers.
For example, Synopsys, a chip design software firm, has developed Synopsys.ai Copilot, a generative AI tool that can answer questions, create scripts, and generate RTL code, a form of chip design language, just by having a conversation in plain English. Synopsys.ai Copilot is integrated into the full Synopsys EDA stack, and is being used by Microsoft’s in-house silicon team.
Nvidia, the chip maker behind the popular GPUs for AI, has also developed its own Gen AI tool, called ChipNeMo, to help its engineers design and improve its chips.
The company trained its system on top of models including Meta Platforms’ open-source Llama 2, and the system is designed to be used with existing design automation tools like those from Synopsys.
ChipNeMo can answer questions about GPU architecture, generate chip design language code, and summarize notes and status updates for hundreds of different teams. Nvidia claims that ChipNeMo has helped its engineers save time and increase productivity.
Owing to their ability to process thousands of tasks at the same time, chips like GPUs generally require nearly 1,000 people to build, and each must understand how pieces of the design work together as they work to continuously improve them.
But with Gen AI, you don’t 1000 people anymore.
Google’s research lab, DeepMind, has also appliedGen AI to improve logic synthesis, a chip design phase that involves turning a description of a circuit’s behavior into an actual circuit. DeepMind’s AI system can find better solutions than existing methods, and may be used to enhance Google’s own custom AI chips, called Tensor Processing Units (TPUs).
Gen AI is not only being used by large companies, but also by researchers and academics who are exploring new ways to design chips. A team at New York University’s Tandon School of Engineering designed a chip in about a month by conversing with ChatGPT, using a technique called “Chip Chat”. The chatbot was able to automatically write Verilog, another chip design language, based on the researchers’ inputs.
Gen AI is still in its early stages, and there are many challenges and limitations to overcome. For instance, generative AI can sometimes produce errors or “hallucinations”, which need to be carefully validated by human engineers.
Moreover, Gen AI cannot yet automate the entire chip design process, from design to verification, implementation, and testing. Experts estimate that the ability to autonomously create a functional chip using Gen AI is about five years away.
Nevertheless, Gen AI is a promising technology that has the potential to transform the semiconductor industry and enable faster and better chip design. By using the power of ChatGPT and other AI systems, engineers can collaborate, innovate, and create the next generation of chips that will power the future of computing.