While artificial intelligence continues to generate headlines about existential threats, experts argue that the real dangers are far more mundane and immediate. Current AI systems excel at writing and coding but still lack critical capabilities necessary for true artificial general intelligence, including consistent reasoning, long-term planning, and reliable performance beyond their training data. Claims that we’ve already achieved AGI remain highly contentious among researchers.
The catastrophic scenarios depicted in popular culture—rogue AI systems gaining independence and attacking humanity—overlook a fundamental reality: advanced AI systems depend entirely on fragile infrastructure including electricity, data centers, specialized computer chips, and human technicians. A truly intelligent system would recognize this vulnerability and understand that destroying its support network would be self-defeating. More pressing concerns include AI-enabled cyberattacks disrupting critical infrastructure and the potential for premature workforce displacement as companies pursue cost-cutting before AI systems genuinely prove capable of replacing workers.
Perhaps more likely than either apocalyptic or utopian scenarios is what some call “the great disappointment.” The enormous expense of training and operating frontier AI systems—consuming vast amounts of electricity, water, and capital—may prove unsustainable unless costs drop significantly. Meanwhile, genuine progress in specialized applications like drug discovery, weather forecasting, and climate modeling remains overshadowed by the race to build incrementally improved general-purpose chatbots.
The tragedy would be abandoning AI research entirely due to panic, potentially halting beneficial advances in medicine and environmental science. Appropriate caution about AI risks should sharpen engineering and governance rather than distort priorities and understanding.
Originally reported by Livescience →
About the Author
