Human-Content-to-Machine-Data_Final - Flipbook - Page 12
There is also the signiûcant burden large AI models place on the environment, such as via
carbon emissions and electricity costs. The carbon footprint of training a single large
language model is approximately the equivalent of 125 round-trip üights between New York
and Beijing.39 A 100-word email generated by ChatGPT requires the equivalent of one water
bottle to cool the servers for the underlying GPT-4 model to function.40 For some, including
those whose works may have been used to train such models, the beneûts of AI may not be
worth the environmental impact and extraction from our natural commons.
It is difûcult to overstate the wider disruption that advances in AI could cause to the
knowledge ecosystem, from the future of livelihoods in some industries to the ability to
separate fact from ûction and the value of human authorship. In creative industries such as
art, music, and creative writing, there is concern that generative tools will interfere with
human creatives9 ability to create, share, and earn compensation.41 Voice actors have
discovered their likenesses have been used in inappropriate commercial applications or
political messages.42 Journalists are concerned about the effects of their reporting being
presented to users without its original context or editorial standards.43
Evidence of the Broken Social Contract Is All Around Us
The backlash against advances in AI is sweeping, and demonstrates the extent to which the
current relationship between AI developers and the commons is broken.
We9ve observed signiûcant unrest among some web communities, including among those
who have expressly created (and in some cases, openly licensed) their content for wide use.
In 2019, many Flickr users were dismayed to learn their openly-licensed images had been
used to train facial recognition models.44 More recently, many of Reddit9s biggest forums
39
Dhar, P. (2020, August 12). The carbon impact of artificial intelligence. Nature Machine
Intelligence. https://www.nature.com/articles/s42256-020-0219-9
40
Verma, P., & Tan, S. (2024, September 18). A bottle of water per email: the hidden environmental
costs of using AI chatbots. Washington Post; The Washington Post.
https://www.washingtonpost.com/technology/2024/09/18/energy-ai-use-electricity-water-data-centers/
41
Zhao, B. (2024, March 28). Replacement of human artists by AI systems in creative industries.
UNCTAD. https://unctad.org/news/replacement-human-artists-ai-systems-creative-industries
42
Pistilli, G. (2025, March 26). I Clicked
