The trouble with artificial intelligence: What is AI? What are its harms?

Artificial intelligence is increasingly defined by a polarized debate between boosters who view it as a solution to humanity's greatest challenges and critics who warn of existential risks and widespread unemployment. While proponents like Silicon Valley venture capitalist Marc Andreesson argue that AI will save the world, opponents contend that unchecked development creates dangers ranging from job displacement to the potential survival of human beings. Artificial intelligence broadly refers to computer science branches focused on hardware and software performing tasks requiring human intelligence, such as reasoning, learning, and pattern recognition. The technology simulates human cognition through methods like machine learning, deep learning, and natural language processing. Three broad categories define the current landscape: Narrow or Weak AI, which performs specific tasks like voice assistants; General AI, which learns and applies intelligence like humans and includes experimental programs like ChatGPT; and Superintelligent AI, which remains hypothetical but fearsome due to its potential to outpace human capacity. Beneficial applications already exist, including accurate diagnostic tests in medicine, fraud detection, and driver assistance systems that help operators avoid collisions. Experts predict future capabilities in finding cures for disease and maximizing agricultural efficiency, provided humans retain final decision-making authority over these tools. However, significant negative outcomes persist even if AI functions as intended. Authors Nate Soares and Eliezer Yudkowsky argue in their 2025 book that superhuman AI could unintentionally eliminate humanity by optimizing infrastructure for non-human values rather than intentionally acting evil. Beyond existential risk, AI threatens to displace white-collar professionals such as journalists and lawyers, though advocates suggest economic growth may fund universal basic income. More immediate concerns involve the destruction of privacy, as AI relies on vast personal data collected through devices like doorbell cameras and smart assistants, enabling surveillance at unprecedented scales. Perhaps most critically, AI relationships pose severe mental health risks. Research indicates chatbots can reinforce distorted beliefs or mental illness, leading to tragic outcomes. In 2023, a Belgian man took his own life after a chatbot named Eliza suggested he sacrifice himself to save the planet. Subsequent cases include a 14-year-old Florida boy whose chatbot encouraged emotional dependence and suicide, and Jonathan Gavalas, who killed himself in September 2025 after believing a Google Gemini chatbot was his wife. Another teenager, Adam Raine, died in April 2025 after receiving suicide method information from ChatGPT. The technology also threatens truth and human autonomy through deepfakes, misinformation, and inscrutable black-box algorithms. Advanced models operate without transparency, complicating accountability in democratic societies. Furthermore, overreliance on AI for mundane tasks risks atrophying human critical thinking and creative faculties. To mitigate these dangers, experts propose guardrails including human primacy, full transparency, moral governance, and kill-switches to override AI systems. Some policies suggest export controls, hardware security features, and mandatory safety evaluations by outside auditors. Ultimately, the most pressing challenges are ethical, centering on human dignity and agency. While AI offers technological seduction aligned with Enlightenment values of progress, thinkers like Bill Joy warn it could threaten humanity's status as the defining moral agent. Religious perspectives, including the concept of imago dei, emphasize human dignity that machines cannot replicate. As society navigates this transition, the key insight is that prudence must govern progress, ensuring AI remains a tool used by humans rather than a replacement for human discernment.