Engagement-driven media will be the end of knowledge.
Sam Bankman-Fried (henceforth SBF) said something ignorant about Bayesian reasoning to Michael Lewis and many people called him on it, including some high-followership influencers. Then other influencers saw the engagement and jumped in with both feet and no understanding of the basics of reasoning with uncertainty.
To make sure that we all know what he said, here it is from Lewis's book:
The problem isn't the reasoning — not that most commenters have made any attempt at Bayesian reasoning, which is done by calculating.
No, the problem is more basic.
The problem is using probabilities in a situation where the information has been realized. Probabilities are for unobserved states of the world; when we observe the state of the world, there's no longer any use for probabilities.
Let’s say we shuffle a deck of cards, getting a specific outcome as follows
The prior probability of that particular outcome is approximately
0.0000000000000000000000000000000000000000000000000000000000000000000123979993085715 or 1.23979993085715E-68.
That is the inverse of the factorial of 52 (the “approximately” is because of the numerical limitations of the program used to compute it; it’s a rational number that continues forever).
After the outcome occurred, saying that it’s unlikely that it did occur because that prior probability is so low is nonsense.
The "actual probability" that the outcome occurred, given that we know whether it did has no meaning: it either did occur (no, that's not “probability 1”; it's an observation) or it didn't (again, not “probability 0”; again it's an observation).
The greatest known English writer (until now) may be a matter of taste, but it's not a matter of probability: writers are people who have written and published, so all the information needed for the assessment has been revealed (ignoring the matter of taste, which is not what SBF was commenting on).
By comparing the writing of Shakespeare with the other writers (in many cases a page or two of the latter will do, especially in the post-modern era...) we can order the writers from best to worst. That’s how Shakespeare gets the first slot (for the people who SBF was talking about). Even if there were ties, that wouldn't be a case for probabilities, but just for declaring a tie.
A pity that so many of the people commenting on SBF's point spent so many words trying to sound knowledgeable while proving that they didn't know that probabilities don't apply to realized events.
But they certainly got a lot of engagement.