A 16-Year-Old's Obsession of Poisoned Snails

Atishay JainANTH-2203Prof. BenessaiahMay 1, 2026

Final Learning Portfolio

A 16-Year-Old’s Obsession of Poisoned Snails

I still remember the very first day I learned about ChatGPT. It was a cold New England winter night and my buddies and I were cooped up around my monitor, fascinated. To say the least, the pairing of overly imaginative 16-year-old boarding school kids and a chatbot (which literally would obey nearly every command we gave it) somehow ended with my friend obsessing over ChatGPT’s fantasy stories about him eating poisoned snails and ruling the world.

After messing around on it for a few weeks and getting pissed off at how shitty its writing was, I began paying attention to how the chatbot would respond to me based on how I prompted it. If I was more assertive and direct, was it more assertive and direct? One weakness I remember documenting at that time in late 2022 was the chatbot’s inability to access present-day information. As someone who wanted more information to draw patterns from, my 16-year-old self had struggled to understand the technicality of if something so powerful could be built, why could it not simply access the internet?

Four years later, after several failed attempts of my father trying to convince me to sign up for AWS’s AI certifications, I find myself sitting in a cyborg anthropology class asking more unsettling questions; though this time, people’s lives are at stake. Over the course of the semester, I was grateful to have received the intellectual toolkit to ask these pressing questions, understand potential questions, and engage in meaningful debate, especially as we begin to reshape the entirety of the human future.

Meaning Lost in Translation Traversing far beyond language, translation is what allows members of humanity to create genuinely unique meanings for any circumstance they encounter. A stoic father may mean well for his child when he advises him to sleep early, but the child may assume otherwise. The father doesn’t know better. He grew up in severe poverty, where the only source of light he relied on was the sun. If he didn’t capitalize on all the daylight possible, he would automatically be unsuccessful in his education. Without his education, he would never get ahead. The father attempts to translate different parts of his life through the systems he inhabits: high-stress corporate America, nonsensical family politics, and deep spiritual grounding. The stoic father means well, but of course, the child may assume otherwise. In some way, the stoic father keeps adapting, and adapting, and adapting all while the complex system he lives in seemingly self-regulates.

I wrote in my last learning journal my inkling for analyzing power across several systems at once, and how this led me to discovering complex adaptive systems. Across the semester, I started learning more about algorithms, how they are shaped, and how they also seemingly self-regulate to an extent. Take content moderation for example. To users like myself, I am not actively thinking about all the “virtual checkpoints” or off-shore content moderation that takes place on the content I consume on social media. Just as the stoic father inhabits his own epistemic network, AI systems readily inhabit networks themselves as well. We have seen across the semester, several examples of tracking appearing, often where we least expect it, and what’s especially scary is how invisible these integrations are meant to be. I think Instagram may actively be somehow reading this now, so I’ll say something to make it jealous. TikTok’s algorithm is better, and it’s not close. I have no idea why, it just is.

Qualia

If all of our feelings, thoughts, and actions can be tracked and profited off (thanks Zuboff!), what is left? Let’s take the stoic father example again. In the epistemic network the stoic father inhabits, there are often unexplainable feelings he will encounter. The redness of the roses he gifts his wife. His inner child geeking after purchasing his dream car. That unexplainable feeling is known as qualia. Even if it were possible to map the stoic father’s life onto a chart with nodes and then translate that to an AI system, the qualia would simply not transfer. The stoic father would go back to his home village expecting to find the same chaatwala whose father’s, barber’s, daughter also happens to be his closest childhood friend, but instead he would find a QR-code replacement. The transaction takes a mere ten seconds. Designed to be completely optimized and frictionless, the AI system processed the payment correctly, and money went from the stoic father’s account to the vendor’s account (wherever the vendor was). Why should the stoic father blame the system? It did its job. But it also turns out the chaatwala’s, father’s, barber’s, daughter died of dysentery last year due to poor water conditions. At least they had QR codes.

QR Codes

The stoic father’s story teaches us something urgent – ad hoc AI solutions are useless when there are inherent infrastructural flaws beneath them. A common theme we have seen is western tech imperialism dressed as altruism. In Kate Crawford’s Atlas of AI, she opens by saying that AI, in an effort to monetize power and control, is “ultimately designed to serve existing dominant interests” (Crawford, 8). Western biases have continued to appear and reappear, even in datasets that are being used to train models. I have reflected in the past on how often algorithms benefit from appearing automated and neutral, even though human decisions, most likely exploited labor, are shaping what users have access to. If platforms continue to appear as “neutral,” this propagates a misleading notion that AI systems are self-regulating once they have been trained sufficiently. AI system moderators readily make decisions which affect millions of users everyday, but do not design the rules of the system. Western ideologies cannot continue to be the de facto penetrating influence of the Global South (I hate this term and need to think of a new one); and so as long as the Silicon Valley elite continues to take lead, efficiency, optimization, and profit will breed. With efficiency, optimization, and profit comes bias. True diversity of perspective in AI systems is not something which can be coded or computed by an algorithmic model.

Some Nerdy Stuff to Support My Fears

In a specific case study I came across which dealt with AI healthcare implementation in developing countries, researchers have identified a pattern they call “pilotitis,” where AI pilots demonstrate initial technical success, but collapse once donor funding ends because the infrastructure to sustain them was never there to begin with (Joseph, 2026). Undoubtedly, longer-term institution adoption is difficult. However, some solutions the researchers proposed included addressing barriers such as “fragmented ecosystems, donor dependence, weak governance, limited trust, infrastructure gaps, and political complexity” (Ibid). They provided an example of an AI pilot in Brazil which was only truly successful when policymakers aligned the digital health goals originally lined out across the different levels of government. They also cited examples of cross-infrastructural success in India and east and sub-Saharan Africa. Though this case study, which was very interesting to read, focused on developing countries specifically, I started questioning the role that the powerhouses that are at the forefront of AI development have. Namely, the United States. xAI, Grok, Neuralink, Starlink, and whatever Elon Musk dreamt of and ‘link,’ have picked up unparalleled popularity in public awareness. This is not just a coincidence, but there is actually real empirical data which supports the hype of AI is unlike anything we have seen before, especially via anthropomorphism. According to an in-depth study conducted on the mechanisms of AI hype by Markelius et al. in 2024, ChatGPT reached 1 million users within a week of release and crossed 100 million active users within a month (Markelius et al., 2024). The paper then coins this as the fastest-growing app of all time. Among other arguments, they also caution that the current discourse surrounding AI – what most people call AI’s “Black Box” – obscures the underlying, and extremely complex, infrastructure of AI. The “mystification” of such technologies, they state, falsely attributes AI as an isolated entity from human, environmental, and social dependencies which are needed for its survival. I wholeheartedly agree with what these scholars concluded (and also empirical proof that society’s-almost-collective fear is genuinely real).

The Oh-So Ominous Future

Four years ago, I was a 16-year-old watching a chatbot write fantasy stories about poisoned snails. I thought the most pressing question was genuinely why it couldn't access the internet to tell me who won the 2022 World Cup. My reflections and the education I received in this class broadened my perspectives entirely. The right question is what happens to the stoic father when the systems built around him are optimized for everything except him. This class and course material gave me the intellectual pursuit that makes me excited to learn, grow, and question; that feeling is extremely rewarding for me as a college student.

As innovation speeds, regulation falters. I will change that.

Citations:

Crawford, K. (2021). The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial

Intelligence. Yale University Press. https://doi.org/10.2307/j.ctv1ghv45t

Joseph J. (2026). From pilot to policy: why AI health interventions fail to scale in developing countries.

Frontiers in digital health, 8, 1699005. https://doi.org/10.3389/fdgth.2026.1699005

Markelius, A., Wright, C., Kuiper, J., Delille, N., & Kuo, Y. T. (2024). The mechanisms of AI hype and

its planetary and social costs. AI and Ethics, 4(3), 727-742.