Questioning Linear Extrapolation of LLM Parameters

teortaxesTex · x · 2026-07-10

The author expresses skepticism towards linear extrapolations like "31Q total parameters and 243T active parameters," arguing that such scales might never materialize. They also suggest that parametric knowledge may not necessarily be the primary vessel for "Godmind" preferences.

This response challenges an AI 2040 scenario discussion, fundamentally questioning the reliability of using parameter scales to linearly predict the capabilities of future intelligent systems.

Related event: Superintelligence Predictions Spark Cognitive Divide(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →