It is a little more than a year ago since ChatGPT was launched in the public domain, write Frank Houghton, Lisa O’Rourke Scott and Jennifer Moran Stritch, Technological University of the Shannon, Co Limerick.

Ongoing debate

In that relatively short period of time ChatGPT and its competitors have ignited an ongoing debate both within academia and the wider world.

Despite resistance and concerns in some quarters artificial intelligence (AI) has been firmly embraced by many fields, most notably in universities¹, including in engineering education².

It is highly probable that numerous applications currently being made to the €95.5bn Horizon Europe Research & Innovation (R&I) funding programme, which runs until 2027. will involve an AI component of some kind³.

The potential benefits appear significant across many domains from medicine development⁴ to manufacturing⁵ which, according to a recent review of the literature, is currently leading the publication field^6,7. However, even putting aside more controversial comments⁸, the increased power required to run AI software, particularly in the short term is a given⁹.

As noted in an earlier Engineers Ireland article, Google’s AI alone may use as much electricity as all of Ireland¹⁰. More alarmingly, estimates suggest that global AI training and inference activities may use as much electricity as all of the Netherlands¹¹.

Improved processors

Even in scenarios in which improved processors are being installed in data centres with limited local power supplies, such upgrades themselves incur an environmental cost in terms of raw material extraction, manufacturing, shipping and installation.

To many people such concerns are negligible, particularly in the context of Horizon Europe R&I grant applications. However, it is important to remember the varying mindsets and belief systems that may frame this issue among those reviewing such applications.

Although inconsequential to many, some reviewers may approach any additional electricity use as exacerbating a climate emergency in which anthropogenic global warming is already leading to crop failures and desertification in some areas, alongside glacial melt and consequent sea level rise and flooding in others. Both scenarios inevitably appear to involve increased civil strife, war, forced migrations, hunger, disease and death.

Such framing is important as Horizon Europe reviewers are routinely asked to evaluate, ‘is this proposal compliant with the "do no significant harm" principle?’.

Ultra-competitive

Interpretations of the key term ‘significant’ may vary dramatically, but any perceived shortcoming may be crucial in the ultra-competitive, and often multi-stage, Horizon Europe review process.

It is important to remember that although Ireland is currently led by a coalition government that includes the Green party, the environmental movement here is extremely modest compared to that evident in some Scandinavian countries and the Netherlands^12,13. From such perspectives any increase in electricity use may be viewed as contributing to climate breakdown and a resulting climate catastrophe playing out in real time.

Therefore it is suggested that those submitting Horizon Europe R&I funding applications, especially those involving AI, strongly consider explicitly including details on reducing the carbon footprint of the proposal through energy and carbon offset mechanisms.

Needless to say the initial proposal stages of such applications are obviously restricted in terms of pages and every sentence is valuable ‘real estate’. However, between two and four lines out of a 10- or more page proposal focused on mitigating the energy issue may be a very wise move in future applications.

This potential requirement may in itself lead to further linkages between companies and entrepreneurial opportunities, be it in retrofitting, carbon sink development or renewable energy sources such as solar, wind or hydro.

Authors: Dr Frank Houghton is Director of Social Sciences ConneXions at TUS and has an interest in the impacts of AI on Geographies of Health & Public Health. He teaches on the BSc in Applied Psychology and the Masters of Engineering (MEng) in Digitalisation of Manufacturing at TUS. Dr Lisa O’Rourke Scott is the Principal Investigator of the EDGE (Exploring Diversity, Gender & Equality  and has a strong interest in the use of AI in Psychology education. Jennifer Moran Stritch is the Principal Investigator of the Loss and Grief Research Group, part of Social Sciences ConneXions Research Institute at TUS.  She has an interest in the potential of AI for social engagement and communication.

References

1) Alajami, Q., Al-Sharafi, M.A. & Abuali, A. (2020) Smart learning gateways for Omani HEIs towards educational technology: Benefits, challenges and solutions. International Journal of Information Technology and Language Studies, 4: 12-17.

2) Shukla, A.K., et al. (2019) Engineering applications of artificial intelligence: A bibliometric analysis (1988-2018). Engineering Applications of Artificial Intelligence, 85: 517-532.

3) European Commission. Horizon Europe Research and Innovation. See: https://research-and-innovation.ec.europa.eu/funding/funding-opportunities/funding-programmes-and-open-calls/horizon-europe_en

4) Gorriz, J.M., et al. (2020) Artificial intelligence within the interplay between natural and artificial computation: Advances in data science, trends and applications. Neurocomputing, 410: 237-270.

5) Marr, B. (2023) Artificial Intelligence In Manufacturing: Four Use Cases You Need To Know In 2023. Forces 7 July 2023. See: https://www.forbes.com/sites/bernardmarr/2023/07/07/artificial-intelligence-in-manufacturing-four-use-cases-you-need-to-know-in-2023/

6) Chu, H., Tu, Y. & Yang, K. (2022) Roles and research trends of artificial intelligence in higher education: A systematic review of the top 50 most-cited articles., Australasian Journal of Educational Technology, 38: 22-42.

7) Crompton, H. & Burke, D. (2023) Artificial intelligence in higher education: the state of the field. International Journal of Educational Technology in Higher Education, 20: 22.

8) Hern, A. (2023) OpenAI leaders call for regulation to prevent AI destroying humanity. The Guardian, 24 May 2023. https://www.theguardian.com/technology/2023/may/24/openai-leaders-call-regulation-prevent-ai-destroying-humanity

9) Engineers Ireland. (2023) Google’s artificial intelligence on its own may use as much energy as Ireland. Engineer’s Ireland Journal, 11 October, 2023. See: Google’s artificial intelligence on its own may use as much energy as Ireland.

10) Patterson, D., Gonzalez, J., Hölzle, U., Le,Q.,  Liang, C., Munguia, L-M., Rothchild, D., So, D.R., Texier, M. & Dean, J. (2022) The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink. arXiv:2204.05149 [cs.LG]

11) DeVries, A. (2023) The growing energy footprint of artificial intelligence. Joule, 7: 2191-4.

12) van der Heijden, H.A. (2005)nRestoration, Environmentalism and the Dutch Politics of 'New Nature'. Environmental Values, 14: 427-446.

13) Larsson Heidenblad, D. (2021) The environmental turn in postwar Sweden. Manchester : Manchester University Press.