Skip to Main Content

Generative Artificial Intelligence

This guide provides documentation about generative artificial intelligence.

Environmental Harms

There is currently a lack of consensus around just how much impact generative artificial intelligence has on the environment. Some writers claim that initial estimations of GenAI's energy consumption were either exaggerated or as yet remain unclear, while others assert that the impact will certainly grow as tools become more widely adopted and powerful.

However, we DO know for certain that future GenAI developments will continue to exert pressure on fossil fuel extraction, require vast amounts of water for cooling data centres, and contribute to a global e-waste crisis.

The following list of articles is not exhaustive, but will provide some perspective on the issue.

 

 

What can you do on an individual level to minimize the environmental harms of GenAI use?

Consider asking yourself the following:

  • Do I need an AI, or can I do this myself?
  • Can I use a smaller or locally hosted LLM instead? or even just a search engine?
  • Will I be able to re-use what I've created with an AI?
  • Who benefits? Who might bear the costs?

Adapted from Climate Conscious AI Use - Wrestling with Environmental Impacts, Emily Simpson/BC Campus.

For some easy-to-adopt ideas on how to mitigate the impact of your AI use, see the 5 Practical API Tips to Lower your Environmental Footprint, from Tilberg University.

Social Harms

Ethical issues surrounding the creation and adoption of generative AI present challenges to users. Be aware of these social and political harms associated with this developing technology, and then decide how you will engage with it.

  • Misinformation - it is likely that generative AI will contribute to growing misinformation because of its inability to fact-check itself and its tendency to create fabrications or "hallucinations"
  • Gender, racial, other biases - because LLMs are trained on data that contain inherent biases, the output will perpetuate these same biases, further harming marginalized communities
  • Ghost labour - as social media platforms did before them, AI companies have outsourced the labour of manually removing illegal and offensive content, usually to low-paid contractors in the Global South
  • Copyright - it is widely known that most of the large AI companies have trained their LLMs on copyrighted works of artists, authors, journalists and other creatives, in ways that are less than transparent and possibly illegal
  • Privacy - companies may require users to submit personal data and accept terms of use that put personal information at risk

 

Here are a few articles about some potentially harmful social impacts of Gen AI, as well as some tips on how you can protect yourself.

H5P interactive created by Rebecca Sweetman, Queen's University. Creative Commons license CC: BY-NC-SA

TED talk by Sasha Luccioni, AI ethics researcher