Polly Wants a Better Argument - by SE Gyges

The “Stochastic Parrot” Argument is Both Wrong and Actively Harmful

Perhaps the most influential single paper on the public perception of LLMs is On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. It is, however, a bit of a mash-up, and credibly seems like it should have been at least two papers. One of those papers raises many valid concerns about the ethical implications and impacts of AI training and use. Another makes the claim in the title, that an LLM is a “stochastic parrot” operating “without any reference to meaning.” That core claim is either irrelevant or completely wrong in every detail, both in how it is commonly understood…

related reading

• Stochastic parrot - Wikipediaen.wikipedia.org
• The Persona Selection Model: Why AI Assistants might Behave like Humansalignment.anthropic.com
• Guardian Angels: LLM Personalization for Productivity and Security · Gwern.netgwern.net
• ChatGPT Is Nothing Like a Human, Says Linguist Emily Bendernymag.com
• LMCA_dataset.pdfandrew.cmu.edu
• What We’ve Learned From A Year of Building with LLMs – Applied LLMsapplied-llms.org
• Things we learned about LLMs in 2024simonwillison.net
• Towards a Typology of Strange LLM Chains-of-Thought1a3orn.com
• Discovering Language Model Behaviors with Model-Written Evaluations — LessWronglesswrong.com
• Taking LLMs Seriously (As Language Models) — LessWronglesswrong.com
• Against LLM Reductionismerichgrunewald.com
• The Future of Everything is Lies, I Guessaphyr.com