Mitchell narrows 'stochastic parrot' to LLMs, not all AI
TL;DR
- Mitchell's essay argues the 'stochastic parrot' label applies to large language models, not to AI systems in general.
- The 2021 paper, co-authored with Bender, Gebru, and McMillan-Major, was always framed around language models specifically.
- She retains the phrase as a tool against what the piece calls 'anthropomorphic traps' where people reflexively humanize AI.
Margaret Mitchell, co-author of the 2021 paper 'On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?', has published a Medium essay titled 'No, AI is not a Stochastic Parrot.' Per summaries of the piece, the argument is narrower than the title suggests: large language models are still stochastic parrots; broader AI systems are not.
The original paper was written with Emily M. Bender, Timnit Gebru, and Angelina McMillan-Major, and the 'stochastic parrot' metaphor was meant specifically for large language models, not for AI systems in general. Mitchell is reportedly pushing back on how the phrase has been flattened into a one-liner dismissal of LLMs, and reclaiming it for a different job: fighting the tendency towards 'anthropomorphic traps' where people humanize AI. Mitchell was fired by Google in February 2021 in connection with the paper and later joined Hugging Face.
Five of the researchers we track in Who's Who had the link up within days.
Shared on Bluesky by 5 AI experts
-
I reposted this, but tacking itonto this thread as well bc I think it's useful context. Margaret was also a co-author of the Stochastic Parrots paper and recently offered some thoughts on which parts of the analogy hold …
View on Bluesky →
Originally reported by medium.com
Read the original article →Original headline: No, AI Is Not a Stochastic Parrot