The speaker discusses Andre Karpathy's recent podcast appearance, where Karpathy suggested AGI is ten years away. The speaker argues that this timeline is based on an outdated definition of AGI, which focuses on an AI doing "any economically valuable work as good or better than a human." Instead, the speaker proposes a definition centered on an AI system replacing an average knowledge worker, emphasizing the practical impact on human employment. The speaker contends that AI systems, through continuous iterative improvements and scaffolding around LLM limitations (like RAG and context management), are rapidly approaching the ability to replace a significant portion of the workforce, particularly the "bottom 50%" of knowledge workers whose performance has not improved. The speaker predicts AGI, by their definition, will arrive before 2028 with a 70% chance and before 2030 with a 95% chance, driven by massive investment and the low bar set by current human mediocrity in many roles.
Sign in to continue reading, translating and more.
Continue
