Could AI really kill us all? Your questions, answered.

There is a strange comfort in asking “Could AI kill us all?” because the question itself implies the threat is a single, identifiable event—a digital asteroid, a machine with bad intentions. But the MIT Technology Review roundtable, as described in the source article (https://www.technologyreview.com/2026/09/18/1144435/could-ai-really-kill-us-all-your-questions-answered/), hints at something far weirder: the audience had more questions than time, which is exactly the problem. We are collectively trying to interrogate a technology that is moving faster than our institutions, our metaphors, and our ability to even formulate the right anxieties. The most interesting part of this conversation is not whether AI will literally destroy humanity in some cinematic flash. It is that “kill us all” is a category error. AI is not one thing; it is a sprawling patchwork of models, incentives, companies, and unintended feedback loops. The existential risk may be distributed across thousands of small decisions—automated weapons systems with hair-trigger rules of engagement, healthcare algorithms that quietly ration care, financial systems that optimize for volatility until the market becomes a chaotic attractor. Death by a thousand cuts is harder to dramatize, but it is no less lethal. What makes this story feel so “Weird & Wild” is the inversion of our usual fear. We tend to imagine AI as a conscious agent with goals, but the more unsettling possibility is that it is an emergent phenomenon without a self—a kind of digital weather system. You don’t negotiate with weather; you don’t reason with it. You build shelters, you watch the pressure systems, and you accept that some storms will be bigger than your models predicted. The roundtable’s unanswered questions are not a failure; they are a symptom of living inside a transition where the old rules no longer apply, and the new rules haven’t finished writing themselves. If there is a takeaway from this MIT Technology Review discussion, it is that we need to stop asking whether AI will “kill us all” and start asking who gets to decide what risks are acceptable, who gets to pull the plug, and who is accountable when a system we built for optimization starts optimizing for something we never meant to value. The technology is not the alien invader. We are the ones building the ark, and we keep arguing about the paint color while the water rises.