Email me at assadiguive@gmail.com, if you want to discuss anything I posted here or just chat.
Guive
Karma: 313
What are your thoughts on Salib and Goldstein’s “AI Rights for Human Safety” proposal?
I don’t know why Voss or Sarah Chen, or any of these other names are so popular with LLMs, but I can attest that I have seen a lot of “Voss” as well.
“I don’t want to see this guy’s garbage content on the frontpage” seems a lot more defensible than “I will prohibit him from responding to me.”
Sorry, I should have been clearer. I didn’t really mean in comments on your own posts (where I agree it creates a messed up dynamic), I mean on the frontpage.
Guive’s Shortform
LessWrong has a block function. Like with Twitter, I think this shouldn’t be used outside of the most extreme circumstances, but Twitter also has a mute function which prevents you from seeing someone’s content but still let’s them respond to you if they want to. Does LessWrong have anything like that?
I think the extent to which it’s possible to publish without giving away commercially sensitive information depends a lot on exactly what kind of “safety work” it is. For example, if you figured out a way to stop models from reward hacking on unit tests, it’s probably to your advantage to not share that with competitors.
I’m not sure that’s even true of leading questions. You can ask a leading question for the benefit of other readers who will see the question, understand the objection the question is implicitly raising, and then reflect on whether it’s reasonable.
Vietnam was different because it was an intervention on behalf of South Vietnam which was an American client state, even if the Gulf of Tonkin thing was totally fake. There was no “South Iraq” that wanted American soldiers.
Also, I bet most people who temporarily lose their grip on reality from contact with LLMs return to a completely normal state pretty quickly. I think most such cases are LLM helping to induce temporary hypomania rather than a permanent psychotic condition.
This feels a bit like two completely different posts stitched together: one about how LLMs can trigger or exacerbate certain types of mental illness and another about why you shouldn’t use LLMs for editing, or maybe should only use them sparingly. The primary sources about LLM related mental illness are interesting, but I don’t think they provide much support at all for the second claim.
It took me a minute to read this as an exclamatory O, rather than as “[There are] zero things I would write, were I better at writing.”
Can you be more concrete about what “catching the ears of senators” means? That phrase seems like it could refer to a lot of very different things of highly disparate levels of impressiveness.
It is not a paraphrase; the denotation of these sentences is not precisely the same. However, it is also not entirely surprising that these two phrases would evoke similar behavior from the model.
Interesting post. Just so you know, there are a few stray XML tags that aren’t rendering properly.
Personally, I don’t have mixed feelings, I just dislike it.
Agreed. Some of the content I post on Twitter could eventually make its way into a LW post or comment, but it would have to be completely rewritten. I think I have seen some people crosspost Twitter threads on LW, and found it annoying.
I’m most intrigued by the argument that as state-of-the-art models get larger, this work will be comparatively more accessible to smaller groups.
Do you think this prediction has been borne out? It seems like it has.
What kind of “research” would demonstrate that ML models are not the same as manually coded programs? Why not just link to the Wikipedia article for “machine learning”?