You Are Afraid of the Wrong AI

Over the past few weeks, we have been deluged with dire warnings that frontier models are on the verge of achieving artificial general intelligence (AGI), and that when they do, they could pose an extinction risk to humanity. In this paper, I argue that the claim that frontier models are on the verge of achieving AGI is based on conceptual confusions and misleading evidence. So, the rational credence in an imminent AI apocalypse should be extremely low. 

 

Still, I argue, AI poses significant risks. Frontier models are powerful, making excellent tools for criminals and cyber warriors. But the most serious risks are not posed by AI, but by the AI companies. The race to develop AGI using current methods could hasten the tipping point for catastrophic climate change—which itself poses an existential risk to humanity.