AIS Mirror Research Repository

Archive: Stochastic Parrots

This vault contains critical analysis of Large Language Models (LLMs), focusing on the "Stochastic Parrots" framework proposed by Bender et al. (2021).

Advertisement
Seminal Paper

On the Dangers of Stochastic Parrots 🦜

Bender et al. (2021)
The foundational text arguing that LLM environmental and social costs outweigh benefits.

Climbing towards NLU: The Octopus Test

Bender & Koller (2020)
Demonstrates that a system trained only on form cannot learn meaning.

Sponsored

The Curse of Recursion (Model Collapse)

Shumailov et al. (2023)
Training models on their own output leads to loss of variance.