The LLMs were also tested on their ability to parse datasheets and design circuits. They were able to accurately extract data from datasheets and convert it into useful code. However, when tasked with designing a circuit, the LLMs made slightly strange decisions, suggesting that while they can assist in the design process, they are not yet capable of completely replacing human expertise.
Key takeaways:
- Large Language Models (LLMs) can be used to assist in circuit board design, with the study focusing on the utility of frontier models such as GPT4o, Claude 3 Opus, and Gemini 1.5.
- LLMs can be used to extract data from complex sources like PDF datasheets and convert them into useful code, which can be a helpful automation technique.
- While LLMs can understand a schematic image well enough to turn it into a netlist, they struggle with designing the circuit itself due to the complexity and precision required.
- The study found that LLMs can be useful in certain areas of circuit board design, but they still have limitations and are not yet capable of replacing human expertise.