transfer-learning-dates-to-1976

IN premise — entries/2026/06/21/wiki-Neural_network_28machine_learning29-chunk-8.md

Created 2026-06-21T09:55:52+00:00

Transfer learning dates to 1976 (Bozinovski), far earlier than the 2010s as commonly assumed.

Dependents

These beliefs depend on this one:

• OUT dormant-solutions-await-enabling-conditions — ML's pattern of multi-decade adoption latencies combined with the convergent discovery of genuine mathematical necessities across disconnected fields suggests that solutions to current reliability challenges may already exist in published research, awaiting the economic or hardware conditions that would make them viable.
• IN idea-latency-validates-economic-gating — ML ideas exhibit systematic multi-decade adoption latencies — transfer learning (invented 1976, adopted 2010s) and self-supervised pretraining (invented 1991, dominant 2018) were both available for decades before widespread use, providing independent evidence that ML progress is gated by economic and hardware readiness rather than idea availability.