Reflections on AI and Its Place in Society

It feels very challenging to articulate my perspective on AI, as I feel like there is so much I have yet to learn, and as it affects and has the potential to affect so many parts of society that to try and consider them all at once just feels overwhelming.

I do not view the technology as an unalloyed good, nor do I see it as an apocalyptic evil. While from a purely technical aspect, I understand framing it as a tool, when looking at the societal impacts I think that to refer to it simply as a tool is to reduce it to its physicality. I think that AI development is a race for global sovereignty; to ignore it, or to attempt to shove the genie back into the bottle, would be futile. In my opinion, the rational approach to take is to consider and prepare for future issues seriously, while also attempting to mitigate the very real issues we are currently facing.

The present feels to me undeniably messy, a mix of potential positive uses mixed with seemingly inevitable risks. On the one hand, I see immense utility in the use of AI in the NGO space, accelerating their reach and democratising access to vital information. Even those who most vehemently proclaim themselves to be anti-AI often seem to see the benefits of breakthroughs in fields such as healthcare. Yet, I also see generative models trained on stolen artwork, and reproducing the biases already present in anglophone western societies. A lesser-known example of this that particularly stuck with me recently was about biases related to animals. When being asked to generate an image of a salmon swimming in a river, a LLM dutifully generated an image of a neatly sliced, supermarket-ready salmon fillet swimming upstream in a river, reducing the animal to a product of human consumption.

Another point that I have mixed feelings about is on the environmental impact of LLMs, particularly since the energy consumption of text and image generation seems to be largely debated. I read one paper decrying the catastrophic environmental impact of Large Language Models, and another the next day arguing the carbon footprint is actually quite negligible and that we aren’t actually certain of how much energy is used per prompt. I’d like to look into this further.

The feeling of uncertainty is particularly acute when I look at the field of research I’m currently working in: education. I care a lot about AI literacy, and I feel a bit wary about how we could ensure AI education is made accessible to all. There’s a significant Global North/South divide, compounded by the fact that many teachers have not been trained in AI-literacy, and that many schools lack funding to purchase the necessary tools for students to proactively learn about AI.

Furthermore, I see the undeniable potential for AI to assist with classroom orchestration and personalised learning, but, at the same time, I don’t see it replacing the human element of schooling. Play and real, friction-filled human interaction are important for a child’s social development. Certain experiments, like the Alpha schools in the US, are working to try and deliver an AI-led education. While I find them to be very interesting case studies, I fear that their success relies heavily on having parents who possess the tech literacy, industry connections, and exorbitant wealth to guarantee some level of success to their children. I would be surprised if, even if its model were applied to all schools, such a system would actually work for the general population.

A final point on AI in the classroom that I feel is important to insist upon, whether for cases like the Alpha schools, or more general AI integration, is the fact that AI exposure isn’t the same as AI literacy. Thrusting an LLM into a student’s hands risks turning it into a crutch rather than a tool. China’s new AI curriculum, which focuses on teaching the actual mechanics and possibilities of the technology, seems like a robust way of building a genuinely informed generation, rather than a generation of dependent prompt-engineers.

Author: Natasha Litherland

The full article can be found here: https://uncertainty101.substack.com/p/current-perspectives-on-ai-in-general

Date: May 2026