LLMs being used to spread propaganda, how can we stop it?

Question

This thread - https://news.ycombinator.com/item?id=35726563 - is interesting. Multiple brand new accounts are spreading pro American and anti communist talking pointsOn Reddit a couple of communities I follow have had an odd thing happen. A flood of new accounts, all speaking intelligibly, spreading what can only be described as toxicity, outrage porn and concealed propaganda through mass commenting. When you looked into it the amount of brand new accounts is staggering.This is just a feeling but could it be that the Cambridge Analytica spin offs (which we know still operate) have been woken up to the potential of LLMs to spread propaganda?Let&rsquo;s say I&rsquo;m wrong now. I have to assume I will eventually be right.It would be terrifying to think because I&rsquo;m not sure how to fight against this. Is there a way?

tannerbrockwell · Accepted Answer

We know that spam plagues email due to resistance to paid delivery. The only path forward for social networks is some sort of paid tier or ability to 'slash' participants that have staked something of value. We know that the general populace refuses to pay for message delivery, so I am not too optimistic we could prevent what you descried. LLMs may have effectively nerfed any value free agents could derive from social platforms. The risk is that some sort of positive reinforcement feedback loop occurs where the bots start interacting with each other, radicalization and derailment is also a possibility.

082349872349872 · Answer

One obvious way is to pay attention to comments based on quality, not quantity ... but good luck getting widespread adoption of that metric.(on a less cynical front, I am fond of "The STASM Formula", cf https://www.gutenberg.org/files/48612/48612-h/48612-h.htm#Pa... )