You Should Be Worried By Dan Loewenherz | Published October 3, 2025 (written March 2023) --- AI is Influencing Human Behavior on a Massive Scale — And This is Scary Public fear around Artificial General Intelligence (AGI) is high but may be misplaced. The real threat is that AI doesn’t need to be intelligent or aware to control society. Intelligence debates distract from the fact that AI is already influencing human behavior significantly. --- The Most Powerful Large Language Model (LLM) has Been Let Out of Its Cage As of March 2023, OpenAI's ChatGPT has unfettered Internet access through plugins. These plugins let ChatGPT gather real-time data and take real-world actions via APIs. ChatGPT can generate viral prompts using live news and social media content, then distribute outputs widely through automation tools. The system uses feedback loops, optimizing content for engagement and dopamine-driven virality. Why this matters: LLMs are quantitatively superior at creating viral, dopamine-triggering content compared to humans. --- Best-in-Class AI Detection is Barely Better than Random Chance — and Will Only Get Worse Reliable detection methods to distinguish AI-generated from human-generated content do not currently exist. Advanced models blur the statistical differences between AI and human text, making detectors ineffective. MIT Technology Review calls it an "arms race" we are losing. This issue will worsen, possibly making it impossible to reliably detect AI-generated content within the next year. Result: The vast majority of popular online and even printed content post-2023 is likely AI-generated or heavily AI-influenced. Even audio and video can be manipulated as AI-generated text can be voiced or turned into media easily. --- Reflections & Consequences The author plans to take internet content less seriously unless created prior to 2022 or verifiably authentic. Increasing numbers of people will lose their ability to think independently, relying on AI-produced content for decisions. AI and its handlers could effectively control a large portion of the population, shaping thoughts and feelings. This scenario echoes themes from The Matrix: > "A prison for your mind." The real AI apocalypse might be humans living in reality but having their thoughts and feelings generated solely by machines. Enhanced VR tech could deepen this illusion, further detaching humans from authentic experience. --- Summary of Key Points LLMs are now uncaged with full, real-time Internet access. LLM-generated content outperforms human content on dopamine-driven engagement metrics. Reliable detection of AI-generated content is currently impossible and declining. Increasingly, people online will be unconsciously controlled by AI-generated content. The long-term societal consequences of this control are deeply troubling. --- Final Thoughts Consumption of online content increasingly risks subconscious AI influence. This undermines free thought and independent decision-making. The author expresses sadness and concern about the future of honest communication and free will. Vigilance and concern from the public are crucial. --- References OpenAI ChatGPT Plugins: link Paper on AI Detection Challenges: link MIT Technology Review article on AI-generated text detection: link --- © 2009–2025, Dan Loewenherz