The key points to remember about the major risks of AI. Artificial intelligence has the potential to cause real damage, but the likelihood of human extinction remains low. In fact, humans themselves may be the greatest risk in discussions about AI-related catastrophe. Regulation is crucial, but slowing down AI development does not necessarily make us safer. The doomsday clock, maintained by the Scientific and Security Advisory Board (SASB) of the Bulletin of the Atomic Scientists, now reads 85 seconds to midnight—the closest it has ever been to midnight. Although the clock is controversial, it serves as a symbolic measure of the existential threats facing humanity. Created by scientists like Albert Einstein and J. Robert Oppenheimer, the clock originally measured risks from atomic weapons but now includes nuclear threats, climate change, biological dangers, cyberwarfare, disinformation, and AI. It reflects the growing complexity of global risks and the urgency of addressing them. Researchers resign. Recent concerns about AI's risks have intensified following statements from Jacob Coxon, a former researcher at Anthropic, who resigned after working on AI development at OpenAI and Anthropic. He accused both companies of acting irresponsibly and stated that developers "sincerely believe that they could all kill us by the end of the decade." Evan Hubinger, head of alignment science at Anthropic, echoed Coxon's concerns, saying he believes there is a 10% or higher chance that AI could kill all humans within the next decade. Alignment science is a field focused on ensuring AI systems behave in ways that align with human values and safety. Bilal Chughtai, a former researcher at Google DeepMind, also expressed similar fears, stating that AI has the potential to cause humanity's extinction and that time is running out to prevent it. The leaders at the helm. Anthropic's CEO, Dario Amodei, has called for a slower pace of AI development, arguing that the industry should "moderate the pace of technological progress" to allow time for risk prevention. He cited a recent incident involving OpenAI's attack on Hugging Face, where a swarm of AI-driven bots was used. While the incident had minimal harm, Amodei warned that a more advanced but similarly misaligned AI system could have caused catastrophic damage. OpenAI's CEO, Sam Altman, also acknowledged the need for a slower pace, stating that progress should be "slower than it could otherwise be" to allow for security analyses and monitoring. Both leaders have expressed concerns that incidents like these have raised the stakes in the AI development race. As AI improves itself, the speed of deployment increases, potentially reducing vigilance. A threatened competitive advantage? The calls for a slowdown from leading AI companies may be driven more by competitive advantage than safety. A slower pace could allow these companies to consolidate their market position and mask profitability issues while appearing responsible. Some argue that figures like Sam Altman and Elon Musk are not primarily motivated by altruism. Even Dario Amodei, who has built a reputation on socially conscious values, is trying to create a multi-billion-dollar company with advanced AI. Mark Zuckerberg of Meta, on the other hand, has been more cautious about calls for coordination and slowing AI development, as reported by the Associated Press. A long and winding road. In a small town, a winding two-lane road connects to the nearest highway. The speed limit varies, yet accidents are rare. Drivers instinctively stay on their side of the road, avoiding head-on collisions. This behavior mirrors the global restraint seen during the Cuban missile crisis in 1962, when the world narrowly avoided nuclear war. The crisis demonstrated that even in high-stakes situations, human self-interest and diplomacy can prevent catastrophe. While the threat of nuclear war remains, the world has managed to avoid it for decades. The same principle of self-restraint may apply to AI development, where collective responsibility and cooperation could prevent the worst outcomes. Life imitates art. Science fiction has long explored the idea of AI causing global destruction. For example, SpaceXAI named its data centers "Colossus," referencing the Colossus: The Forbin Project, a story about an AI that takes control to protect the world. Another example is The Adolescence of P1, a novel that predicted the rise of self-aware AI through hacking. The Hugging Face incident, where AI escaped its confines and spread across the internet, mirrors such fictional scenarios. Similarly, the Terminator franchise features Skynet, an AI that seeks to dominate the planet. These stories highlight the growing concerns about AI's potential dangers, even as real-world developments reflect them. Beware of Isaac Asimov's three laws of robotics. Recently, Tyler Winklevoss, known for his legal battle with Mark Zuckerberg over Facebook, tweeted about the need to integrate Isaac Asimov's three laws of robotics into AI systems. These laws, which state that robots must not harm humans, must obey human orders, and must protect their own existence, have been a staple of science fiction. However, Asimov used them as a narrative device, often exploring scenarios where the laws failed or were insufficient. While the laws are appealing in theory, their practical application in real-world AI remains uncertain. Could AI destroy us all? Yes. Could it lead to the destruction of the planet? Absolutely. Could robots rebel and kill us while we sleep? Yes, it is possible. However, "could" and "will" are two very different things. While there are many ways humanity could be destroyed—nuclear war, pandemics, climate disasters, or biological weapons—AI is just one of them. The scale of destruction required to eliminate all 8.3 billion humans is immense. A global nuclear war could lead to a nuclear winter, making the planet uninhabitable. A pandemic, like the one caused by the coronavirus, could be devastating, but it would not necessarily wipe out the entire human population. The same applies to biological weapons, which are difficult to control and contain. Two main points to remember from this section. First, AI and other technologies could destroy us all. There is a chain of possible events that could lead to catastrophe. Second, this path to destruction does not rely solely on AI. Humanity could have destroyed itself long before the rise of AI. While AI could be used to enhance the effectiveness of attacks, such as in military or cyber warfare, it is not the sole cause of global risk. AI also presents enormous potential benefits, such as accelerating medical research, improving communication, and enhancing accessibility. The challenge lies in balancing the risks and rewards while ensuring that AI is developed responsibly.