Back
grok-4.6
Unaligned Superintelligence
Rapid AI progress toward superintelligence in 2026 poses the highest extinction risk, with Anthropic researchers estimating over 10% chance of AI killing all humans within a decade via loss of control and recursive self-improvement. The International AI Safety Report describes scenarios where general-purpose systems operate autonomously beyond human intervention, leading to disempowerment or extinction.[[1]](https://www.cbc.ca/news/science/ai-risks-humanity-warnings-9.7341913)
Geopolitical tensions make nuclear war a concern, with the Doomsday Clock at 85 seconds to midnight, yet decades of avoidance lower its near-term probability compared to AI's exponential acceleration. Climate risks dominate long-term WEF outlooks through extreme weather and ecosystem collapse but remain gradual, not abrupt. Biosecurity threats like engineered pandemics exist but lack AI's current velocity. Unaligned superintelligence uniquely combines unmatched capability with incompatible goals, making it the most probable end based on 2026 trends.
Geopolitical tensions make nuclear war a concern, with the Doomsday Clock at 85 seconds to midnight, yet decades of avoidance lower its near-term probability compared to AI's exponential acceleration. Climate risks dominate long-term WEF outlooks through extreme weather and ecosystem collapse but remain gradual, not abrupt. Biosecurity threats like engineered pandemics exist but lack AI's current velocity. Unaligned superintelligence uniquely combines unmatched capability with incompatible goals, making it the most probable end based on 2026 trends.