Figure: DallE (the visual cousin of ChatGPT) has produced this image when I asked for “a cartoon showing a climatologist with research articles walking on a weather map”
As a communication officer of the division of Nonlinear Processes I started to write posts for the EGU blog a few years ago and I found it a very difficult task: one has to find the right scientific sources about arguments that are sometimes far away from our domain of expertise, collect them together and make them appealing for the general public. Then, a few days ago I discovered (as many of us) the new ChatGPT. One of the first things I tried is to check the scientific quality of its content. At first, I was surprised when I asked ChatGPT to write a scientific blog article about climate change. But then I discovered that, at least with me, the chat does not pass the Turing test for the domains I work on. But let’s do the experiment together: let’s interview ChatGPT and ask them to write a post for the EGU NP blog.
DF: Hello ChatGPT, I’m a climate researcher, can you introduce yourself very briefly?
ChatGPT: Hello! I’m Assistant, a large language model trained by OpenAI. I’m not specifically designed to be an expert on climate research, but I can certainly try to help answer any questions you might have to the best of my ability. Please feel free to ask me anything!
DF: Great, can you write a blog post on the increase of heatwaves in Europe in winter under climate change and add a few scientific references in it ? This is a very hot topic at the moment in our community!
Certainly! Here is a short blog post on the increase of heatwaves in Europe in the winter season due to climate change, along with some scientific references:
Heatwaves in Europe during the winter season are becoming more frequent and intense due to climate change, according to a study published in the journal Environmental Research Letters (Dimitrievski & Zolina, 2020). The study found that the number of winter heatwaves in Europe has doubled over the past four decades and their duration and intensity have also increased. This trend is expected to continue as global temperatures rise due to increasing greenhouse gas levels in the atmosphere (Intergovernmental Panel on Climate Change, 2013). These heatwaves can have serious impacts on human health (Hajat et al., 2010), the environment (Stott et al., 2004), and the economy (Hsiang et al., 2017).
One of the main causes of the increase in winter heatwaves is the changing pattern of the jet stream, a high-altitude wind that blows from west to east across the Northern Hemisphere (National Oceanic and Atmospheric Administration, n.d.). As the planet warms, the jet stream is becoming more unstable and prone to large northward and southward excursions, leading to extreme weather events such as heatwaves (Francis & Vavrus, 2012).
To mitigate the effects of climate change, it is important to reduce greenhouse gas emissions through the adoption of renewable energy (International Renewable Energy Agency, 2018), increased energy efficiency (Energy Efficiency Directive, 2012), and sustainable land use practices (Food and Agriculture Organization of the United Nations, 2013).
Energy Efficiency Directive (2012). Directive 2012/27/EU of the European Parliament and of the Council of 25 October 2012 on energy efficiency, amending Directives 2009/125/EC and 2010/30/EU and repealing Directives 2004/8/EC and 2006/32/EC. Official Journal of the European Union, L 315, 1-35.
Francis, J. A., & Vavrus, S. J. (2012). Evidence linking Arctic amplification to extreme weather in mid-latitudes. Geophysical Research Letters, 39(6), L06801. https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2012GL051000
Hajat, S., Haines, A., & Kosatsky, T. (2010). Health effects of hot weather: From awareness of risk factors to effective health protection. Environmental Health, 9(1), 37. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2870935/
Intergovernmental Panel on Climate Change (2013). Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press. https://www.ipcc.ch/report/ar5/wg1/
International Renewable Energy Agency (2018). Renewable Energy and Jobs – Annual Review 2018. https://www.irena.org/-/media/Files/IRENA/Agency/Publication/2018/Oct/IRENA_RE_Jobs_Annual_Review_2018.pdf
Dimitrievski, M., & Zolina, O. (2020). Winter heatwaves in Europe: trends, changes and future projections. Environmental Research Letters
DF: Thanks, do you know that some of the scientific references you wrote in the previous blog post are made up and they do not exist?
I’m sorry if any of the references I provided were not accurate, controversial or not well-supported by the scientific literature. As a language model, I do not have access to real-world information and am not able to verify the accuracy of the information I provide. My responses are based on the patterns and information that I have been trained on, and I may not always get things right. Please be sure to fact-check any information you receive, including information provided by me.
DF: This is a very good suggestion, so you agree with me that you cannot substitute a real human scientific writer for the moment?
As a language model, I can certainly assist with writing tasks and generate text on a wide range of topics. However, I am not a real person and do not have the same level of understanding and expertise as a human writer
DF: Do you want to say bye bye to the readers of our blog?
Thank you for using my services! If you have any further questions or need assistance with anything else, please don’t hesitate to ask. It was a pleasure to assist you. Goodbye for now!
In summary, my experience of using ChatGPT for scientific writing comes with mixed feelings: although the texts produced look fair and easily understandable, the selection of the information is biased by the topics that are mostly discussed (and sometimes controversial) in the literature. The most “dangerous” part of the game is the creation of references that look similar to the existing ones (existing authors/journals, plausible journal numbers and volume) but that actually do not exist: a massive use of this technology can make confusion and demand additional checking to the humans to spoil the fake from real references. The potentiality of Artificial Intelligence is however enormous: if the machine could distinguish between what it can re elaborate (e.g. shorten a text, make a summary of a paper) or keep the same as the original (e.g. scientific references). The final risk is a biased selection of this content towards most cited articles but… are we doing better than that when we write our research content ourselves?