In our roles, hopefully we can encourage others to be mindful of their use of GenAI and to develop these other vital aspects of AI literacy beyond just writing effective prompts (and combining those two - writing effective prompts that reduce environmental impact, like in the blogpost you shared)

pollution from mineral mining etc. Then there's the social issues like child labour (including deaths and chronic illness from mining and kids classifying extreme images with lasting psychological harm) + lack of consent to use data in training and how data are classified... [2/2] #LTHEchat

There are so many harmful environmental impacts, GenAI is something I wouldn't want to become a routine part of what we do, I'd rather its use were reserved for exceptional circumstances. Harmful air pollution from generators, enormous carbon emissions and water use for cooling, [1/2] #LTHEchat

A5 Key parts of AI literacy are an in-depth understanding of the ethical, environmental and social implications, and inherent bias in training data. With colleagues in HE, also how encouraging use of GenAI conflicts with initiatives relating to things like sustainability and decolonising. #LTHEchat

I recommend Kate Crawford's Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence to anyone who wants a good intro to some of these issues including how GenAI reinforces colonial narratives and the potential harm within data classification #LTHEchat

katecrawford.net/atlas

A4 This blogpost by @dustinhosseini.bsky.social contains a good example of this sort of thing. Raising students' awareness of these issues is a critical part of AI literacy. #LTHEchat
www.dustinhosseini.com/blog/2023/08...

I've heard the same from every colleague I've spoken to, which also reflects what I've read in the literature - it certainly does seem to treat 'white' as default despite only around 15-20% of the global population being white. #LTHEchat

It's certainly worth being very mindful of use e.g. not using it where an existing image would have done the job, but if the careful crafting of an effective prompt formed part of an activity for students to reflect on how something could be visualised, could have potential learning value? #LTHEchat

Apparently CogVideoX uses 3.4 million joules to produce 5 seconds of video, 700x the amount required to produce an image. #LTHEchat

www.technologyreview.com/2025/05/20/1...