Who Will Own AI-Built Designs?
Generative AI could transform product design, but raises questions about creative ownership.
The recent ousting and subsequent rehabilitation of OpenAI’s CEO added some theater to the debate and buzz around artificial intelligence. AI, it seems, is everywhere and in everything from our smartwatches and phones to automobiles, data centers and factories.
With the explosion in generative AI like OpenAI’s ChatGPT, it’s also taking on creative roles that we might have assumed would remain the preserve of human intellect. For a while now, it has been possible to generate realistic images of human faces – not copies but unique individuals that never existed except inside a computer. Also, in 2023, the fashion brand Levi’s became one of the first companies to suggest it would use AI-generated clothing models. These are expected to improve the shopping experience for customers by helping them assess clothes on likenesses that have a similar body shape and size to their own. Of course, it’s also likely to help brands cut marketing and merchandising costs.
Let’s set aside the prospects for the first AI catwalk model, or photographer, or bestselling novelist, and consider activities at the border between engineering creativity and design automation.
AI-based design tools are entering a variety of industries and could transform the way companies create new products. Tools like ChatGPT and Microsoft’s Copilot can learn from literally thousands of worked examples to help engineers with relatively minimal coding skills quickly generate software like an embedded application or automated test routine.
There are further roles for AI in reliability engineering and analysis. I was speaking recently at an event hosted by the European Space Agency about circuit boards for space applications. Of course, repairing failed units in space is practically impossible, so reliability is critical. The accumulated experience of space engineers in designing high-reliability systems is essential for commercial operations to be economically viable. Bringing that expertise into AI-based tools, which can continue to learn from ongoing experience and so design progressively better and more reliable systems in the future, will be important for developing the sector.
Here on earth, some forward-thinking PCB design houses are already experimenting with generative AI to assist with placement and routing of components on PCBs, using libraries of existing design examples. We can expect that generative AI will enter the toolchain in specific features and functions and subsequently become more prevalent as users understand its ability to make their lives easier and get new designs to market more quickly.
Top designers acquire their skills through years of learning and experience. As human experts age, and inevitably retire, their industries can lose the benefit of this knowhow. Traditional mechanisms for sharing knowledge have included mentoring, conference proceedings and literature, which others can study and in turn become the next generations of experts. Now, we have the opportunity to train AIs on the collective expertise of all the world’s subject specialists. These can then continue to learn, improve and update, becoming ageless, always current, and accessible to a broad user base.
On the other hand, there are important questions regarding intellectual property. Some vendors of AI-based tools have sought to claim ownership of material generated by users. For its part, Microsoft has faced challenges over Copilot’s reliance on open-source code for training. A judgment against the company could seriously slow development and adoption of generative AI in many areas. Other legal action is building as groups such as writers and artists claim that AI is training on their created material without consent. The dispute between Microsoft and its challengers is over the terms of the open-source license agreement, which places restrictions and obligations when code is taken from repositories such as GitHub. The arguments are far from clear cut as many current licensing and copyright laws originate from eras before generative AI. A judgment could set an important precedent. Both sides know the stakes are high.
Back in the more straightforward world of high-tech product design, bringing AI into EDA tools could help engineers create products that are easier to build, test, approve, and maintain. For many years, designers have been encouraged to consider multiple aspects as early as possible in the design process: design for test, design for manufacture, consider antenna placement, power consumption, electromagnetic compatibility. With so many aspects to prioritize, engineers could be forgiven for being unable to consider everything at the same time. AI-based tools, on the other hand, can deal with numerous issues and variables simultaneously to work out the best possible compromise, also considering supply chain issues like the availability of components, PCB materials and fabrication services, as well as optimizing for recycling and disposal at end-of-life.
In previous columns, I’ve referred to humans’ tendency to overestimate the impact of new technologies in the short term and underestimate their effects in the long term. It’s probably not contentious to suggest that everyone knows AI is going to have a huge influence on many aspects of our lives and is likely to be more transformational than the Internet. Protection is certainly needed, so we should take interest in the development of new regulations like the EU AI Act, which could become the first law of this type in the world, and international forums like the AI Safety Summit. First held at Bletchley Park in the UK last November, it could offer the prospect of a global accord to minimize the risks and maximize the opportunities for all of us.