4 Comments

This explains the root of the problem:

https://www.lesswrong.com/posts/FaJaCgqBKphrDzDSj/37-ways-that-words-can-be-wrong

Young Wittgenstein's Theory continues to hold: almost all philosophical problems are language problems. But it's far too simple of an explanation for the hubristic Western culture of 2024, so blinded by The Science they can't see what's right in front of their eyes.

Expand full comment

Great work as always Tom. Have you looked at this paper: https://academic.oup.com/pnasnexus/article/3/2/pgae035/7591134?login=false

In an article in The Conversation they write: "A recent paper illustrated how large language models can be deployed to craft micro-targeted ads at scale, estimating that for every 100,000 individuals targeted, at least several thousand can be persuaded." Link: https://theconversation.com/disinformation-threatens-global-elections-heres-how-to-fight-back-223392

Seems like a gross overstatement to me - the article doesn't measure "persuasion" in an interesting sense - but I'd be interested what you think given you have much greater expertise.

Expand full comment
author

Thanks Dan. I don' t know that paper - I'll have a look and see what I think (no promises on speed!)

Expand full comment
author

i have a two part response this paper. Here's the first part https://tomstafford.substack.com/p/ai-juiced-political-microtargeting

Expand full comment