Mailbag XLI: Is AI Going to Wipe Us Out? | AI Extinction Risk, Doomscrolling, Climate Anxiety, Economic Fear, and the Future of Humanity

Is artificial intelligence going to wipe out humanity? Could AI cause human extinction within the next ten years? Are we living through the beginning of a global collapse, or are our minds being asked to process so many real and possible threats at once that the future itself has begun to feel like an emergency? In Mailbag XLI of The Observable Unknown, Dr. Juan Carlos Rey responds to listener Sarah F., who writes that she is “scared to death” by what she sees happening in the world. She’s worried about artificial intelligence. The economy. Natural disasters. Political and social frustration. The possibility that AI could eventually destroy humanity. And beneath all of it is a more personal question: How do you keep your common sense when everything arriving through the screen seems to say that the world is coming apart? The episode begins with the question Sarah most needs answered: No, there is currently no established scientific finding that AI will wipe out humanity within ten years. Researchers do take catastrophic AI risk seriously, but that is very different from having a reliable extinction timetable. The 2026 International AI Safety Report describes wide disagreement among researchers about future loss-of-control scenarios. Some experts believe sufficiently capable future AI systems could create catastrophic or even existential risks. Others judge those scenarios far less plausible. The same report says present-day systems show some early warning capabilities relevant to the debate but do not currently possess the combination of abilities required for a genuine loss-of-control scenario. That difference is central to this episode: “AI extinction is a risk researchers are studying” is not the same statement as “humanity has ten years left.” Will AI Cause Human Extinction? There is a serious scientific and policy discussion about AI existential risk. Researchers study whether future AI systems could become capable enough to evade human oversight, pursue long-term goals, conceal behavior, resist shutdown, gain access to consequential systems, or otherwise operate beyond meaningful human control. Those are legitimate research questions. They should not be trivialized. But neither should uncertainty be converted into prophecy. The 2026 International AI Safety Report says current AI systems lack the capabilities necessary for the kind of loss-of-control scenario associated with existential catastrophe, while also noting that relevant capabilities are advancing and deserve continued monitoring and safety work. So when someone says: “AI is going to kill everyone in ten years,” the appropriate question isn’t simply: “Could something terrible happen?” It is: What evidence supports that particular probability and that particular timetable? Possibility is not probability. Concern is not prediction. Prediction is not inevitability. And uncertainty is not a countdown clock. AI Risk Versus AI Doom The episode draws a sharp distinction between taking AI risk seriously and living psychologically as though the worst possible AI scenario has already been decided. Artificial intelligence can create genuine risks. Those include failures in high-stakes systems, cyber misuse, deception, misinformation, labor disruption, harmful autonomous behavior, concentration of power, and future systems that may become increasingly difficult to supervise. The 2026 International AI Safety Report specifically notes reliability failures, increasingly capable AI agents, uncertainty around loss of control, and the need for advance preparation even where timing and probability remain unclear. But serious risk analysis requires something doom culture often destroys: proportion. A catastrophic possibility deserves investigation. It does not automatically deserve certainty. Why Does the World Feel Like One Continuous Emergency? Sarah’s fear isn’t really about AI alone. Her letter describes a broader experience familiar to anyone caught between breaking news, social media, economic uncertainty, climate disasters, political conflict, war, technological disruption, and catastrophic predictions. Dr. Rey describes this experience as temporal collapse. A phone can present: A flood. A wildfire. A war. A market warning. A political crisis. An earthquake. A mass shooting. A frightening AI prediction. A climate projection. A layoff announcement. All in a few minutes. These events may occur in completely different places, involve different causes, and unfold across very different timelines. But the nervous system receives them through the same screen. One after another. Again. Again. Again. The world begins to feel less like a planet containing billions of lives and countless simultaneous conditions and more like one room in which everything is going wrong at once. Doomscrolling and the Nervous System This is one reason doomscrolling can become psychologically corrosive. The problem isn’t simply exposure to bad news. Some bad news is important. The deeper problem is the destruction of distance, scale, probability, and time. Something happening somewhere becomes emotionally immediate. Something that might happen later begins feeling present. A projection begins feeling like an event. A possibility begins feeling probable. A probability begins feeling inevitable. And eventually the body is responding not only to what has happened but to thousands of things that could happen. That is an enormous burden to place on a nervous system built to respond to much smaller fields of immediate danger. Present, Plausible, Predicted The episode offers a simple framework for recovering proportion: Present. Plausible. Predicted. A present threat is occurring now. A plausible threat could occur and has enough supporting evidence to deserve preparation. A predicted threat is a forecast, scenario, projection, or argument about something that may happen in the future. These categories aren’t interchangeable. If your house is burning, act. If wildfire danger is approaching your community, prepare. If someone online announces that civilization will inevitably disappear in ten years, investigate before allowing the claim to occupy your nervous system as though the event has already begun. The governing principle is: A possible future is not automatically a present emergency. Catastrophic Thinking and Future Anxiety Catastrophic thinking performs a peculiar psychological trick. It makes the nervous system experience a possible event before the event exists. A person can spend years emotionally inhabiting: The recession that might come. The war that might expand. The disease they might develop. The climate catastrophe that might occur. The AI takeover that might happen. The relationship that might end. The job they might lose. The mind begins treating imagined futures as current environments. And even when the feared event never arrives, the years spent living inside it are still gone. That doesn’t mean people should ignore future risk. It means preparation and pre-suffering are not the same thing. Is the Economy Really Collapsing? Sarah also describes the economy as “falling apart.” Economic hardship can be completely real at the household level even when national statistics don’t describe economic collapse. Housing costs, debt, food prices, job insecurity, medical expenses, stagnant opportunities, or personal unemployment can create genuine financial distress. But national collapse is a larger claim. As of the latest available U.S. figures, real GDP increased at a 1.5% annual rate during the second quarter of 2026. In August 2026, unemployment was 4.1%, with nonfarm payroll employment increasing by 162,000. Those statistics don’t mean everyone is prospering. They do mean that “the economy is collapsing” is not a precise description of the current national data. This is another example of the distinction between: “I am under economic pressure.” and “The entire economic system is collapsing.” The first may be painfully true. The second requires separate evidence. Climate Anxiety and Natural Disasters Sarah’s concern about natural disasters also deserves care rather than dismissal. Climate risk is real. The World Meteorological Organization’s 2026 update projects that global temperatures are likely to remain at or near record levels through the coming years. It estimates an 86% chance that at least one year between 2026 and 2030 will surpass 2024 as the warmest year on record, with continuing consequences for climate-related hazards. That is serious. But serious information still needs context. Climate risk does not require spending every hour psychologically positioned inside every flood, fire, hurricane, drought, and heat wave occurring anywhere on Earth. Awareness should produce informed behavior. It should not require permanent neurological occupation. The Difference Between Information and Occupation One of the deeper themes of this Mailbag is that information can cross a threshold and become occupation. You begin trying to remain informed. Then every alert feels urgent. Every projection feels personal. Every disaster feels nearby. Every expert disagreement feels like evidence that nobody knows anything. Every uncertainty becomes another reason to keep searching. The search creates more fear. The fear creates more searching. And before long, information has stopped serving judgment. Judgment is serving information. That is the moment to change the dose. How to Stop Doomscrolling Without Becoming Uninformed The episode proposes a practical experiment. For one week: Choose one or two serious news sources. Check them deliberately rather than continuously. Avoid catastrophic prediction content late at night. Don’t let recommendation algorithms decide which danger gets another hour of your attention. When a frightening claim appears, ask: Who is making this claim? What evidence supports it? Is it describing something happening now? Is it a scenario? What assumptions have to be true for it to occur? Do serious experts agree? Where do they disagree? What evidence would change the claimant’s mind? Can the claim be falsified? This isn’t denial. It is epistemic discipline. AI Anxiety and Uncertainty Artificial intelligence is particularly good at provoking future anxiety because it combines three psychologically powerful ingredients: Rapid change. Deep uncertainty. Potentially enormous consequences. When consequences are enormous, people naturally focus on worst cases. When change is rapid, older reference points feel unstable. When uncertainty is deep, confident predictions become psychologically attractive. The result can be a strange marketplace in which the loudest person is often the person least burdened by uncertainty. But “nobody knows exactly what happens next” doesn’t mean: “The worst prediction is probably true.” It means: Nobody knows exactly what happens next. Can Humanity Control Artificial Intelligence? There is no guarantee. There is also no established finding that control will inevitably fail. AI safety research exists because technological outcomes are partly affected by what researchers, developers, governments, institutions, users, and societies do before dangerous conditions emerge. Technical safeguards matter. Evaluation matters. Governance matters. Deployment decisions matter. Security matters. Human judgment matters. The fact that researchers are studying AI loss of control is not evidence that loss of control has already become unavoidable. It is evidence that people are trying to understand the problem before the most dangerous versions of it exist. What Does AI Mean for the Future of Humanity? The honest answer is that nobody knows exactly what AI will mean for humanity fifty years from now. AI may alter employment. Education. Medicine. Scientific research. Defense. Art. Media. Politics. Relationships. Economic organization. Human attention. The distribution of knowledge and power. Some consequences will be beneficial. Some will be harmful. Some will be strange enough that our current vocabulary will fail us. Human beings will make mistakes with AI. They will also build safeguards, discover failures, argue over rules, revise systems, create new norms, and adapt. The future is open in both directions. That matters. Uncertainty doesn’t belong only to catastrophe. It belongs to survival too. You Don’t Have to Carry Humanity Through Breakfast Perhaps the most important line in the episode is addressed directly to Sarah: You don’t have to carry the survival of the species through breakfast. There are global problems that deserve global attention. But an individual human being still has to inhabit a life. Eat. Sleep. Work. Love people. Answer messages. Pay bills. Walk outside. Take care of a body. Help where help is possible. Vote. Learn. Prepare. Build something. Remain reachable to joy. The existence of serious global risk does not cancel the legitimacy of ordinary life. What Can You Actually Do About AI Risk? You may not control the trajectory of artificial intelligence by yourself. You can still decide how you relate to it. Learn from serious sources rather than prediction markets disguised as entertainment. Distinguish present systems from hypothetical future ones. Support thoughtful AI safety research and governance. Become more literate about how AI works. Pay attention to how AI affects your profession and community. Question both utopian and apocalyptic certainty. And protect your own cognitive environment from people whose business model depends upon keeping you frightened enough to click again. Agency begins where the actual radius of action begins. The Future Doesn’t Need You Terrified Fear can be useful. It draws attention toward danger. It can motivate preparation. But sustained fear can eventually reduce judgment rather than improve it. The goal isn’t to become unconcerned about the future. It’s to remain capable of meeting it. Humanity has never possessed a guaranteed future. Neither technological progress nor catastrophe comes with a warranty. What we have is uncertainty, knowledge, argument, preparation, adaptation, error, correction, and choice. That is less emotionally satisfying than prophecy. It is also much closer to reality. MAILBAG XLI: Is AI Going to Wipe Us Out? is for listeners interested in AI extinction risk, artificial intelligence safety, AI existential risk, AI doom, AI anxiety, AGI risk, artificial general intelligence, loss of control, AI alignment, future of humanity, catastrophic risk, doomscrolling, climate anxiety, economic anxiety, future anxiety, catastrophic thinking, existential anxiety, nervous system regulation, information overload, mental health and news, uncertainty, technological change, AI predictions, AI safety research, human extinction, AI and humanity, climate change anxiety, economic collapse fears, media literacy, epistemic discipline, and how to stay sane in an age of constant crisis. The future isn’t guaranteed. Neither is catastrophe. And uncertainty belongs to both sentences. A possible future isn’t a present emergency. The future doesn’t need you terrified of it. It needs you capable of arriving there. The Observable Unknown is a podcast exploring consciousness at the intersection of neuroscience, culture, and lived experience. It is written and hosted by Dr. Juan Carlos Rey of drjuancarlosrey.com and crowscupboard.com, an interdisciplinary scholar whose work bridges neuroscience, philosophy, and the interior dimensions of human experience.