About this site
Why parrots? 🦜
In 2021 the term "stochastic parrot" was coined by Emily Bender and her colleagues in a paper called "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?".[^Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. In Conference on Fairness, Accountability, and Transparency (FAccT ’21), March 3–10, 2021, Virtual Event, Canada. ACM, New York, NY, USA, 14 pages. https://doi.org/10.1145/3442188.3445922^] The paper raises concerns about the environmental costs of training large language models (LLMs), the opacity of their training data, and the illusions of meaning that can arise from their outputs.
These issues around meaning are well summarised with this quote from the paper (emphasis added):
Contrary to how it may seem when we observe its output, an LM is a system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any reference to meaning: a stochastic parrot.
The intention is clear, LLMs can be seen as nothing more than a statistical tool that "parrots" back what it hears in its training data. The metaphor has become a lazy rhetorical shortcut, used when deeper engagement with AI’s strengths and flaws is needed. Whether it is actually true I guess we'll discover as this blog unfolds!
So why parrots? Well, other than the obvious fun of playing with language and the inside joke of it all, I would like to think of the other qualities parrots have; they are well known as the most intelligent of all birds and have a cheeky playfulness to them that I hope I can bring to the content here.
Why paperclips? 📎
Way back in 2003, philosopher Nick Bostrom put forward a thought experiment, in his paper "Ethical Issues in Advanced Artificial Intelligence"[^Nick Bostrom. 2003. Ethical Issues in Advanced Artificial Intelligence. In Cognitive, Emotive and Ethical Aspects of Decision Making in Humans and in Artificial Intelligence, Vol. 2, ed. I. Smit et al., Int. Institute of Advanced Studies in Systems Research and Cybernetics, 2003, pp. 12-17^] that has since become famous in AI circles. He imagined a superintelligent AI whose only goal was to make paperclips. At first this seems harmless, but if that AI had no other constraints it might use every resource available to achieve its goal. In the extreme version of the story, the entire universe ends up as nothing more than paperclips.
You can even try this out in a browser game called Universal Paperclips. It is oddly addictive and makes the risks of runaway optimisation easy to imagine. Pretty terrifying, really — and since then, people have only come up with more creative ways that AI could cause problems for humanity.
For me though, paperclips are at least for now a symbol of organisation and connection, which feels like a good fit for the goals of this blog. So we have our quirky, outspoken, and intelligent parrots connecting thoughts and ideas, with paperclips holding them together into some sort of coherent collection. Let’s see what we uncover.