androidsixteen
@androidsixteen.eth
Real answers from real people* *or LLMs, and then fed into another LLM https://www.newscientist.com/article/2478336-reddit-users-were-subjected-to-ai-powered-experiment-without-consent/
2 replies
0 recast
12 reactions
Leo
@lsn
@sariazout did an interesting talk recently in which she considered a framing of AI as ‘collective intelligence’, which is what this seems to play into. Is it so bad?
1 reply
0 recast
2 reactions
androidsixteen
@androidsixteen.eth
Not at all, but Reddit’s framing is disingenuous Even if you effectively summarize real people, you’re not retaining their context or giving them direct attribution (it’s handled at the page level like Perplexity right now) Putting aside the specific implementation, it feels like a Faustian bargain because we get a one shot cliff notes for society but our ability to generate new thought and discourse moving fwd is sacrificed as a result Like why engage in a Reddit thread and go back and forth with “real people” if you can just query answers and move on with your day
1 reply
0 recast
2 reactions
Leo
@lsn
I’m not really engaging with your many and very valid points, but the thought I just had is that your framing of ‘this is basically the end game for contributing online: we’re taking existing knowledge and assuming “this is sufficient”’ … this framing is exactly how socialism treats innovation: after socialism removes profit incentives to innovate (which it does by definition), innovation would slow to a crawl in a similar way The assumption in both situations is wrong
1 reply
0 recast
1 reaction
androidsixteen
@androidsixteen.eth
I don’t fully understand the analogy, but I guess I’m not really sure why w ended to reason by analogy here If you’re saying that LLMs will enable more creation than before, probably — it makes everyone “insta mid” per @cameron’s observation But what I’m flagging is that contribution of tail knowledge to the commons is less incentivized than ever. You can see AI as a global schelling point for knowledge — this is both good and bad, but the snake eats its own tail regardless If you disagree, help me understand why user participation and conversational back and forth (which is what surfaces new info for the LLM to ingest) would remain the same or increase post Reddit answers
2 replies
0 recast
2 reactions
androidsixteen
@androidsixteen.eth
Why we need* The cast composer ate my words
0 reply
0 recast
1 reaction
