There’s a justified horror many people feel about the environmental impact entailed every time someone shares a prompt with a language model. While I’d be lying if I said I fully share that feeling, I experience something similar when I see pointless uses of image and video models. I’ve largely stopped making AI-generated images for this reason. I almost never have a real need to do it, unlike the other ways I use LLMs. I can understand that what I feel about these uses mirrors what others feel about all, or nearly all, uses of LLMs.
Why, then, do we so rarely see a platform like TikTok discussed in these terms? As a commuter whose brain is wired in a way that public-transport cacophony can be mildly agonising, I’ve spent a lot of time thinking about this. The soundscape I find most difficult is when multiple people are absently flicking through short-form videos without ever watching one in full: each playing a couple of seconds from a whole sequence of clips, creating a disordered, overlapping noise on all sides. At this point, I find it almost impossible to tune out.
One estimate suggests that TikTok’s annual carbon footprint may already exceed that of Greece. Mobile video is intrinsically energy-intensive, and the platform’s design encourages distracted browsing. For most users, it’s also a profoundly passive space: they consume rather than create, often in ways that feel distracted and subjectively unsatisfying. There’s something increasingly grotesque about this: a sense of people hypnotised by an object that simultaneously erodes their capacity to focus. It feels almost like a cruel joke: a self-fragmenting obsession.
What strikes me as odd is that we feel so little moral unease about TikTok on these grounds, while generative AI attracts so much: and, for the most part, rightly so. For the avoidance of doubt, I’m not suggesting either/or, but both/and. Our analysis of the environmental impact of digital activity needs to encompass the whole platform landscape, and it needs to recognise the link between distracted engagement and (environmentally) wasteful use.
Mark Carrigan 