Using generative AI to create hardware prototypes
Cette publication existe aussi en Français
A project in the UK and the Netherlands is using Large Language Models (LLMs) to create hardware designs with a technique it calls Conversational Prototyping.
While LLMs are widely used for developing code, the project is exploring how they can be used for the hardware architecture and even the PCB board layout. Developing hardware can be more complex, however, especially when making use of things like sensors and actuators. Iterating for small or large scale production can require considerable time and resources.
The project is seeking to use LLMs for ‘Conversational Prototyping’ that would allow developers and end users to work in conversation with an AI system to develop the functionality of a prototype, incorporating sensing and actuation.
The project is led by Lars Erik Holmquist, Professor of Design and Innovation at Nottingham Trent University with Dutch media company Poppe and Partners and Electric Circus, a collaboration between a Dutch inventor and a puppeteer. DataLink in Loughborough is providing the electronics hardware support.
A physical testbed that connects an LLM with sensors and actuators is being developed alongside a set of prompts and commands guiding the LLMs interaction with the testbed. It will evaluate this approach to prototyping.
The project is part of the pro² network which seeks infrastructure solutions beyond traditional manufacturing that support the production of digital devices. This brings together the Universities of Bristol, Lancaster, Nottingham and Bath in the UK to look at ways to boost manufacturing over the next five years.
Other partners include universities at Aalborg and Aarhus, Copenhagen, Saarland, Genoble Alpes and Ludwig Maximilian University of Munich as well as Stanford in the US. Commercial partners include ARM, prototype board maker Eurocircuits (UK) and Silicon Press in Cambridge.
www.prosquared.org; datalink-electronics.co.uk
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
