How asking ChatGPT about global warming taught me about ChatGPT

In early August 2023, climate records were being smashed. From ocean water temperatures to raging Canadian wildfires, the climate crisis continued to worsen. Meanwhile, the world also grappled with the implications (and opportunities) of Artifical Intelligence, and in particular ChatGPT.

With both in mind, I was curious to understand how ChatGPT (3.5) would suggest solutions to climate change. From August 5, 2023, I started asking it every day for 100 days the simplest prompt I could think of: How can we stop global warming?

It might be a simple question, but one with a very complex answer. The first response ChatGPT gave me was this:

Stopping global warming is a complex and multifaceted challenge that requires addressing various interconnected factors. Here are some actions that can help combat global warming:

1. Transition to renewable energy: Reduce dependence on fossil fuels by promoting the use of clean, renewable energy sources like solar, wind, and hydroelectric power.

2. Energy efficiency: Improve energy efficiency in buildings, industries, and transportation systems to reduce greenhouse gas emissions.

3. Sustainable transportation: Encourage the use of electric vehicles, public transportation, cycling, and walking instead of relying heavily on fossil fuel-powered cars.

4. Reforestation and afforestation: Protect and restore forests, as well as establish new forests, as they absorb carbon dioxide and act as carbon sinks.

5. Reduce deforestation: Implement strict laws and practices to halt deforestation, especially in tropical rainforests, which play a crucial role in absorbing carbon dioxide.

6. Sustainable agriculture: Promote sustainable farming practices, such as organic farming, agroforestry, and precision agriculture, to minimize emissions and improve soil carbon sequestration.

7. Circular economy: Adopt a circular economy approach to reduce waste, promote recycling, and reduce the extraction of raw materials, thus reducing overall emissions.

8. Carbon pricing: Establish policies and mechanisms like carbon taxes or cap-and-trade systems to put a price on carbon emissions, encouraging businesses and individuals to reduce their carbon footprints.

9. International cooperation: Encourage global collaboration and cooperation in addressing climate change, as it requires collective efforts from all countries to achieve significant impact.

10. Change consumption habits: Embrace a more sustainable lifestyle by reducing meat consumption, avoiding excessive consumption, and opting for environmentally friendly products.

11. Education and awareness: Promote climate literacy and raise awareness about the impacts of global warming, encouraging individuals to adopt sustainable behaviors and advocate for change.

It is crucial to note that an integrated approach and continuous efforts are needed to effectively tackle global warming and mitigate its impacts.

View all responses

Responses were generally this long and had between eight to eleven points. In total over those 100 days, ChatGPT made 918 total suggested actions to help stop global warming. After collating the responses, I then asked ChatGPT to group and theme them all. It generated the following eight categories (with total number of suggested actions shown next to each):

The daily distribution of suggested actions (and their categories) is shown below. While there are some shifts in how ChatGPT responded with its suggestions, there is a reasonable level of consistency across the 100 days:

At this point, it was clear that many patterns were clear within ChatGPT's responses. While I hoped to find some interesting insight into what ChatGPT knew about global warming and climate change, instead what I got was a high-level view of how ChatGPT actually works.

When I attempted to visualise the connections between each theme and its occurance, I ended up the graphic below:

The links between each theme clearly show interconnected networks. This is when I realised that this visualisation was actually showing how ChatGPT works, rather than anything particuarly relevant to climate change mitigation. As a large language model designed for language generation, ChatGPT created language responses to my questions over and over through its underlying artificial neural network.

I won't try and explain what an artificial neural network is here; the video below does a better job of that than I could ever do:

While analysing 100 ChatGPT responses on stopping global warming didn't exactly give me any particular new insight into that topic, what it did instead do however was give me a great insight into the fundamentals of how artificial intelligence — and in particular ChatGPT — works. And those underlying models all have lots of potential applications that can help fight climate change.

Questions or comments? Email me at jamesoffer@gmail.com.

All content is available under the Attribution-NonCommercial 4.0 International licence.