AI Safety Expert Roman Yampolskiy WARNS About the TWO Likely Endings with AI

AI superintelligence poses an existential threat to humanity because it is inherently uncontrollable, making a negative outcome virtually guaranteed. Rather than pursuing general superintelligence, which functions as a replacement for human cognition, development should focus on narrow, task-specific tools that solve complex problems—such as protein folding or medical research—without enabling recursive self-improvement. Current efforts to build advanced AI are racing toward a system that lacks alignment with human values, creating a high probability of extinction. While some argue for design-based safety, the fundamental unpredictability of superintelligent systems renders traditional control mechanisms ineffective. Ultimately, the path forward requires international cooperation to restrict the development of general superintelligence, prioritizing human safety and the preservation of human agency over the pursuit of potentially catastrophic technological breakthroughs.