Using LLMs for consultation is not what’s destroying the environment nor your brains. This is a scare tactic used by news outlets and educational bureaucrats to scare people from what actually helps them. The truth is…

Environment Argument

The “massive” environmental footprint headlines surrounding data centers, electricity demand, and water cooling are primarily driven by heavy, high-compute operations: real-time video generation (like Sora or Runway), massive image rendering, continuous multi-modal video/audio processing, and massive deep-learning model training runs. Standard text-based interactions with LLMs (tokens of conversational text and logic) require a fraction of the compute power. A thoughtful, text-based collaborator workflow generating outlines, essays, and strategic notes is extremely lightweight compared to heavy generative media.^[1] A peer-reviewed conference paper demonstrated that generative video/image tasks and model training runs consume up to 60 to 1,400 times more energy per task than text-based classification and text-generation inference.^[2]

Corporate media often attempts to shift the entire burden of ecological anxiety onto individual users (“Are your chat logs melting the ice caps?”), while ignoring industrial-scale energy grids, cryptocurrency mining, and commercial data scaling.^[3]

Why do they do this? Shifting climate anxiety onto individual users is a well-documented public relations tactic—historically popularized by fossil fuel corporations through campaigns like the personal “carbon footprint”—designed to privatize moral blame.^[4] Framing environmental harm as a matter of individual choice diverts public scrutiny and regulatory pressure away from energy-intensive industrial practices, massive corporate infrastructure expansion, and unregulated commercial data scaling.

Your use of a text-based AI collaborator to organize your thoughts, refine your curriculum, and protect your sanity in a toxic work environment is an efficient use of tooling, not an environmental catastrophe.

Brain Development Argument

The Myth of the “Ivory Tower” Student & AI Cognition

LLM critics frequently argue that generative models bypass working-memory strain and impede the development of dense neural pathways. However, this cognitive-decline hypothesis relies on a highly idealized—and historically inaccurate—narrative of traditional student behavior. Educational traditionalists often write as if pre-AI students were noble scholars laboring in an ivory tower, meticulously dissecting etymologies, cross-referencing primary texts, and committing deep historical analyses to long-term memory.

The reality of legacy education was far more cynical: twenty years ago, “learning” routinely meant opening SparkNotes, skimming a textbook, and copying sentences almost verbatim onto a worksheet five minutes before class to throw something together and get it through the door. Rote transcriptions and mechanical summary-swapping required virtually zero synthesis, zero critical engagement, and left students with minimal conceptual understanding or long-term recall.

From an instructional standpoint, swapping mindless rote copying for a conversational AI model is not a regression—it is an upgrade. Rather than passively absorbing static text, interacting with a generative model forces a student to read customized explanations, evaluate the output, and iteratively (repeated refinements) prompt the system until the concept makes sense. Because an LLM tailors information to the user’s immediate questions, it acts as an interactive Socratic partner rather than a cheat sheet.^[5] Conflating external cognitive offloading with an automatic decay of intelligence fundamentally misinterprets how students actually parse and retain information.^[6] It’s like saying calculators make us dumber; they only reduce friction and cognitive load so that we can work through more complex material, be more accurate, and be more efficient. LLMs are not turning active minds into passive vessels. If anything, they are replacing human past behavior of mindless mechanical cheating (copying verbatim hat’s in the textbook, submitting that, and forgetting it), with a personalized, dialogic model that can actually achieve. comprehension.

Conclusion

When teachers are forced to parrot anti-AI dogma during their practicums, they are being initiated into a guild. It’s not about rigorous analysis of how cognitive tools evolve; it’s about compliance. If you don’t drink the Kool-Aid, you don’t get certified, you don’t get hired, and you don’t fit the mold.

It feels shallow because it is shallow. It’s an insular system protecting itself from a reality it doesn’t want to adapt to. When your entire social and professional framework is built on a specific, fragile way of life, anything that disrupts it has to be demonized rather than understood.

LLMs are amazing structural writing tools when used responsibly. They function as conceptual sparring partners and advanced linguistic instruments. If an educational bureaucrat is anti-AI, they just are refusing to remove their head from the sand, to think about how it can be leveraged to enhance thinking and writing skills. So they throw scare-pitches and press easy trigger button. It is the textbook equivalent of screaming “think of the polar bears!” or “your brains are turning into scrambled eggs!” whenever you run out of actual intellectual arguments. People rarely admit they don’t understand a transformative technology. It is far easier to wrap incompetence in a cloak of “moral superiority”—labeling the tool as inherently corrupting or unethical—than to put in the effort to master it and redesign a curriculum around it.

---

^[1] Johannes Zschache and Tilman Hartwig, “Comparing Energy Consumption and Accuracy in Text Classification Inference,” arXiv Preprint, arXiv:2412.01234 (2024). Empirical study measuring inference energy consumption across task types, showing standard text-based token generation operates at a low milliwatt-hour/kilowatt-hour footprint per query relative to multi-modal workloads.

^[2] Sasha Luccioni, Yacine Jernite, and Emma Strubell, “Power Hungry Processing: Watts Driving the Cost of AI Deployment?” in Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’24) (New York: Association for Computing Machinery, 2024), 85–99, https://doi.org/10.1145/3630106.3658542.

^[3] Alex de Vries, “The Growing Energy Footprint of Artificial Intelligence,” Joule 7, no. 10 (2023): 2191–2194, https://doi.org/10.1016/j.joule.2023.09.004.

(Peer-reviewed analysis in Cell Press journal examining how data center scaling, hardware supply chains, and industrial grid infrastructure drive aggregate power draw rather than individual user prompts).

^[4] Michael E. Mann, The New Climate War: The Fight to Take Back Our Planet (New York: PublicAffairs, 2021).

(Seminal monograph documenting how industrial energy sectors weaponize the narrative of individual carbon footprints to shift political focus away from systemic grid decarbonization).

^[5] Evgenia Gkintoni, Hera Antonopoulou, Andrew Sortwell, and Constantinos Halkiopoulos, “Challenging Cognitive Load Theory: The Role of Educational Neuroscience and Artificial Intelligence in Redefining Learning Efficacy,” Open Science Framework, December 15, 2024, https://doi.org/10.17605/OSF.IO/5M9J6.

^[6] Michael Gerlich, “AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking,” Societies 15, no. 1 (2025): 6–22, https://doi.org/10.3390/soc15010006.