3 Comments

So true, and the writing rings clear as a bell. I am reminded of Malcolm Gladwell's ideas on how statistical randomness, or basically, luck, will sometimes skew the odds in our favor (or out of it).

Expand full comment

Interesting thoughts. I find the science and mathematics heavy-tailed distributions to be illuminating across many domains of human experience. Perhaps, most obviously, wealth distributions, which exhibit similar effects to what you discussed.

Expand full comment

> “In other words, we can easily deceive ourselves into thinking we understand why something rose to the top, even though in reality chaotic happenstance played a large role.”

Ah yes, the oneiric effect. Certainly does help with crafting an analogous understanding of how some LLMs work.

I’ll be interested to see how you address the primacy effect and how it connects to recall testing for long-context LLMs in your next post.

Expand full comment