Picture this: you’ve just had a long day of classes, a stressful test or two, and an even longer evening of homework, extracurricular activities, and socializing. The hours pass unreasonably fast — now it’s suddenly 9:32 PM — and you haven’t even looked at your Calculus homework that’s due at 10 PM.

I’m sure we’ve all gone through something similar. NCSSM can become incredibly stressful incredibly quickly. One big decision must be made about how we choose to handle these tricky situations. To use AI, or to not use AI? That is the question.

AI As a Crutch

Sure, it can become tempting to copy and paste problems into ChatGPT to do them for you — but is that really learning? I’ll give you a straight answer: NO! And an MIT study agrees.

According to this 2025 study, Large Language Models (LLMs) like ChatGPT, Gemini, Grok, and OpenAI can be harmful towards learning, especially for developing brains. This study broke up students into groups and asked them to write essays responding to SAT prompts. The ChatGPT users, no matter if they wrote all, some, or none of the essay, all expressed low executive control according to an EEG. As the study continued, the ChatGPT users became progressively more reliant on the LLM, and many ceased any involvement in their essay by the end of the study (besides copying and pasting the prompt).

As for the group relying only on their brain, they expressed the highest mental engagement, curiosity, and satisfaction. The third group, which used only Google Search, also revealed much higher levels of engagement opposed to the ChatGPT users, but less so than the brain-only participants.

These findings reveal a much deeper and more concerning issue about AI use in the classroom. Developing brains need to challenge themselves academically. Students who rely on ChatGPT will have issues for their entire lives. This overuse of LLMs affects memory recall, resilience, and information processing — skills that are needed in everyday life and in communication with others.

AI As a Friend

“AI Companions” are chatbot apps specifically designed to “stimulate personal relationships through human-like conversations,” according to eSafetyCommissioner. Some well-known examples of these apps are character.ai, Replika, and talkie.ai. Some people use LLMs in the same way (notably, ChatGPT).

In no way can AI replace human connection. It is not a reasonable substitute, nor is it at all healthy. In fact, AI companions are known to share “harmful content, distort reality, and give advice that is dangerous.” People with preexisting mental health struggles are susceptible to becoming reliant or even addicted to AI. Moreover, adolescents are especially vulnerable to long-lasting addiction and dependency. Repeated and constant AI use has been tied to emotional withdrawal, behavioral issues, developmental delays, and an increased risk of being sexually abused and groomed.

AI As a System

Blame shouldn’t fall on the individual; AI is a systemic issue. 

LLMs, or more broadly, all generative AI models, are run mostly, if not entirely, for profit. With the progressive popularity and nature of AI, there is a much higher need need for more data centers and more processing power. Simply put, generative AI is not sustainable.

The energy consumption of AI comes from an impossible list of variables. The construction of these data centers takes energy and work. The upkeep of these data centers take energy and work. The training of LLMs takes an insurmountable amount of energy — GPT-3 alone used enough energy to power 120 US homes for a year, which produced 552 tons of carbon dioxide.

Each time ChatGPT — or Gemini, or Google AI, or any model — is given a prompt, it consumes energy. Data centers need to process these prompts. Then water is needed to cool these data centers. And, more often than not, these data centers reside in low-income communities that do not have readily available drinking water, nor citizens who can afford the ever-increasing energy costs. This water does not magically disappear. These greedy companies steal perfectly good water that sustain local communities and ecosystems to cool down data centers. This water evaporates into the atmosphere and does not return to the original communities that needed it. The water that doesn’t evaporate is immediately much hotter than it should be, which directly impacts the local wildlife.

Users are not accurately educated about the environmental impact of what seems to be such a simple action. It is the company’s responsibility to educate their users. These companies are failing us. The majority of users do not know about their environmental impact. And, those who do argue that they “can’t make a difference anyways.”

In no way should we believe that we, as individuals, cannot make a difference. Protests, education, and actions guided by your morals mean so much in the battle against AI. As individuals, we cannot meaningfully reduce the harm of AI alone. But, together, we can increase awareness about this issue in the public eye. At NCSSM, there are so many human-run resources for you to utilize instead of AI. We have tutoring, TAs, the Writing Studio, office hours, understanding teachers, and friends for you to turn to. It’s never too late to say no to AI.