Discourse Series 2024: AI and engineering: research, resources and regulation
3 October 2024, Royal Irish Academy
Speakers: Dr Tom Coughlin, President of the Institute of Electrical and Electronic Engineers, in conversation with Orla Feely MRIA, President of University College Dublin
Rapporteur: Dr Andrew Hines, RIA Engineering and Computer Sciences Committee and School of Computer Science, University College Dublin
This discourse was the first of a series of six on the theme of Artificial Intelligence (AI), which will look at the topic through different perspectives: technology, political, ethical and regulatory, creative, and applications. For the first discourse, Dr Tom Coughlin was in conversation with Orla Feely MRIA. Dr Coughlin, as well as representing the Institute of Electrical and Electronic Engineers (IEEE) as President and CEO, of which he is also Life Fellow, is an expert with over 40 years of experience and is CEO of Coughlin Associates. He has published papers and a textbook as well as patents in digital storage technology for consumer electronics. Professor Feely is a member of the RIA, a Professor of Electronic Engineering, is President of University College Dublin, President of CESAER and an IEEE Fellow, and is a former President of Engineers Ireland.
Using the IEEE approach to AI for context, the discourse reflected on the impact of AI on the technology sector and, more broadly, the industrial competitiveness and positioning of AI for Europe and Ireland.
To better understand the importance of the IEEE as a global organisation of real significance to the electronic and electrical engineering profession, Professor Feely outlined its scale and scope. IEEE has over 450,000 members in more than a hundred and ninety countries. It publishes more than two hundred transactions with more than six million documents in the IEEE Xplore Digital Library, a portfolio of 1,144 standards, and sponsors more than 2,000 conferences and events in 190 countries.
The talk began by setting the scene: AI has been a concept for decades but recent technological progress in processing power, networking, and data storage have provided the infrastructure for new applications. We are on the brink of a technological explosion, with new architectures and methods emerging that could enhance AI’s efficiency and effectiveness. However, concerns about the energy consumption and Resource Scarcity of AI models prompt sustainability concerns.
Dr Coughlin’s personal area of expertise is data storage. He pointed out that people are now more aware of ‘the cloud’ and the underlying physical components needed for digital storage, like semiconductors and metals, and the significant energy and water resources required to create and maintain these services. The pace of R&D in traditional semiconductor technology is slowing due to physical constraints in silicon manufacturing. However, innovative approaches such as chiplet technology, where multiple specialised semiconductor devices are integrated closely to function as a single unit, can provide a modular approach that allows greater flexibility in manufacturing and the potential for interoperability among different companies’ devices. Graphics Processing Units (GPU), originally manufactured for gaming, are now powering AI modelling. He highlighted the importance of data storage and data centres for AI and computing, likening it to plumbing — essential infrastructure, yet often overlooked. Pointing to the evolution of storage technology he described the technology trade-offs of size, speed and cost through examples of magnetic core memory to modern solid-state options like NAND Flash. New innovations in non-volatile memory, which can retain data without power, were highlighted as particularly promising for consumer devices, especially those reliant on battery power, like smartphones, wearables and connected devices. For example, implanted devices in the body to monitor health metrics, could utilise AI to monitor and detect health issues earlier than traditional methods.
The IEEE is known for its commitment to public service and professional conduct in engineering. The discussion highlighted the importance of collaboration between technical experts, policymakers and legislators.
The IEEE has been actively helping to shape AI public policy and has been involved in addressing the ethical implications of AI since 2015. It has created comprehensive reports and standards focused on the ethical design of intelligent autonomous systems, with contributions from over 700 global experts.
The IEEE has also created standards related to AI applications in various sectors, including business and education, and established a certification process to ensure compliance with these standards. This work has influenced public policy, encouraging corporations and governments to adopt ethical guidelines for AI. For example, work on the online rights of children, where the IEEE aims to protect young users from inappropriate content, ensuring they can form their own opinions in a safe environment. The organisation has had a tangible impact on legislation and is working to influence corporate practices through its certification processes.
The issue of AI systems perpetuating biases as an important ethical consideration was discussed. Systems that are trained on data that may inherently contain biases necessitating effective guardrails to evaluate and address the problem. Despite awareness of bias, it remains a significant challenge: technologically and legislatively. Data quality and innovative technical approaches can mitigate bias but simply legislating technical solutions may overlook the nuances involved.
As a publisher, the IEEE publishes academic journals, conferences proceeding, books and other publications and has seen evidence of AI-created content in submissions. It has mechanisms in place to identify AI-generated submissions, rejecting those that are entirely written by AI. Tools exist to help determine if a piece of writing is AI-generated, but there is a clear need for better understanding of acceptable AI usage in publishing.
The ethical considerations surrounding the rights of authors whose publications and works may be used in training models were discussed and the current lack of common established rules governing how AI can be trained on existing content, for example through protection, payment or compensation.
The question of whether AI will fundamentally alter research methodologies or transform the research landscape was explored. Positive potentials of AI were presented, for example in tools that could provide concise summaries of technical topics or help to identify gaps in research. While AI can produce some irrelevant suggestions, it may also offer valuable insights that researchers might otherwise overlook. In the fields of chemistry and biology where there have already been both beneficial and harmful outcomes. While AI was presented as a tool that can enhance research and development, significant ethical and practical challenges were also raised.
The concern that traditional research institutions may struggle to compete with large corporations was discussed, highlighting the vast resources and manpower industry is dedicating to AI research. How can universities maintain their relevance and contribute to cutting-edge research when faced with the financial and technological advantages of corporate giants? Possible solutions were suggested including fostering partnerships with these corporations to leverage their resources, advocating for legislative support to provide smaller institutions with necessary tools, and exploring innovative approaches to research that could reduce costs and energy consumption. The need for collaboration among researchers, industry, and government to ensure that traditional research institutions can thrive was emphasised.
The discussion also touched on the challenges of scaling AI development, including the significant funding required, the availability of quality data for training AI systems, and how to sustainably meet the energy demands of AI technologies. The optimistic conclusion was that while there are obstacles, there are also opportunities for more efficient and innovative approaches to research and AI development.
Professor Feely concluded the discourse by reflecting on the role of engineers and the technologies they create in generating wealth and creating new industries and jobs, shaping visions for the future and solving major global challenges, such as climate change. Technology is essential not only in addressing these issues but also in fostering a community of individuals who are knowledgeable and passionate about engineering. The methodologies engineers use to apply physical laws and logic to create functional solutions and the collaborative spirit within the engineering community generally aim to improve lives and inspire others. The engagement of engineers in professional organisations helping to shape the future of AI technology and policy was acknowledged.
In the question-and-answer session the approaches to AI regulation and innovation in the United States and Europe was contrasted. The EU AI Act promotes a more proactive approach to governance that prioritises societal benefits over fostering unchecked but rapid technology innovation. The discussion suggested that instead of letting technology dictate terms, society should define what it wants from technology and govern accordingly.
The issues of global inequality in access to AI resources, how technology can be harnessed to improve lives and AI in an educational context were also raised.
Overall, an accessible and lively introduction to AI was delivered through the opening RIA 2024 discourse series conversation between Professor Feely and Dr Coughlin. The conversation framed the discussion with context on the underlying hardware and software technologies. This was followed by providing a social context for technologies researchers and professionals as well as policy makers. Examples illustrated the suggested practices and policies that might address the challenges. A proactive approach to AI oversight was encouraged: developing metrics that can guide the responsible use of AI, ensuring that it serves to enhance human life rather than diminish it. A vision was presented to navigate the future of AI towards positive outcomes, avoiding pitfalls that could arise from unchecked technological advancement. Individuals need to be prepared for these changes through continuous education and training. While there are fears that AI could displace jobs, history shows that technological advancements often lead to the creation of new types of work. The potential of AI to foster creativity and solve complex problems was highlighted by Dr Coughlin by recalling the IEEE tagline, ‘Advancing Technology for Humanity’, as he advocated for ‘augmented intelligence’, where AI technology enhances human capabilities rather than replacing them.
