AI Risks Are Already Here, Not Just in the Future

Recent warnings from former researchers suggest artificial intelligence could pose an existential threat to humanity by the end of the decade. However, a University of Washington professor argues that this catastrophic scenario is unlikely. Instead, he believes the more pressing danger is the subtle and immediate impact AI is already having on human lives, particularly in areas where oversight is lacking.

Chirag Shah, an adjunct professor in the UW Information School, notes that AI systems are increasingly used to screen job applicants, process claims, and make other high-stakes decisions. He describes this as a significant risk because these tools often operate without sufficient testing or human supervision. The concern is not that AI will suddenly turn hostile, but that it is already influencing critical aspects of society in ways that may be flawed or biased.

The danger of autonomous action

As AI moves beyond generating text to performing actions, such as opening accounts or filling out forms, the risks become harder to predict. Shah explains that while these systems may perform correctly most of the time, the rare failures can have serious consequences. Because these errors are unpredictable, it is difficult to know exactly when or how a system might make a mistake that harms a person.

This unpredictability creates a unique challenge for developers and users alike. Unlike traditional software, which behaves in fixed ways, agentic AI can take unexpected steps to achieve its goals. This means that even well-intentioned systems can produce negative outcomes simply because their decision-making paths are complex and not fully transparent to humans.

Erosion of critical thinking skills

Shah argues that the most profound risk is the gradual loss of fundamental human skills. As people rely more on AI for writing, analysis, and problem-solving, they may stop practicing the critical thinking and judgment required for these tasks. He compares this to learning multiplication tables, suggesting that even if calculators are available, the underlying skill remains vital for human development and professional competence.

This issue is particularly acute in education, where students and teachers are adapting to tools that can complete assignments. Shah warns that if we outsource our thinking to machines, we weaken the foundations of our education and professional training. The goal should be to maintain human agency and critical reasoning, even as technology becomes more capable.

Geopolitical pressures complicate safety

The debate over AI safety is further complicated by the global race between the United States and China. Both nations view advanced AI as a matter of national security, which creates little incentive for international cooperation on safety standards. This competition drives rapid development and deployment, often prioritizing speed and capability over careful testing and ethical safeguards.

According to GN technics/ai (en-US), this geopolitical context makes it difficult to implement the kind of global oversight that might be needed to manage AI risks. As countries compete to lead in AI, the focus remains on innovation and dominance, potentially leaving the public to deal with the unintended consequences of these powerful tools.