In machine learning, a stochastic [stuh-kas-tik] parrot is a large language model (LLM) that is good at generating convincing language, but does not actually understand the meaning of the language it is processing. The term was coined by computational linguistics professor Emily M. Bender in the 2021 artificial intelligence research paper ‘On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?’
Stochastic means ‘random and involving chance or probability.’ A stochastic parrot,’ according to Bender, is an entity ‘for haphazardly stitching together sequences of linguistic forms … according to probabilistic information about how they combine, but without any reference to meaning.’ Less formally, the term refers to ‘large language models that are impressive in their ability to generate realistic-sounding language but ultimately do not truly understand the meaning of the language they are processing.’
The analogy highlights two vital limitations. First, the predictions made by a learning machine are essentially repeating back the contents of the data, with some added noise (or stochasticity) caused by the limitations of the model. And second, the machine learning algorithm does not understand the problem it has learnt. It can’t know when it is repeating something incorrect, out of context, or socially inappropriate. Because of these limitations, a learning machine might produce results which are ‘dangerously wrong.’
Bender’s paper covered the risks of very large language models, regarding their environmental and financial costs, inscrutability leading to unknown dangerous biases, the inability of the models to understand the concepts underlying what they learn, and the potential for using them to deceive people. The paper’s coauthors, computer scientists Timnit Gebru and Margaret Mitchell (using the pseudonym ‘Shmargaret Shmitchell’), were Google employees when it was published. Google asked Gebru to withdraw the paper or remove her name from it, and when she refused, they terminated her employment. Gebru and her supporters argued that she was fired for raising concerns about the ethical implications of Google’s AI research. In January 2021, Mitchell was also fired, allegedly for attempting to exfiltrate files pertaining to Gebru. Google’s response to the paper and its authors led to a protest by Google employees.
The Daily Omnivore 