This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
A distraction from the war and ICE. I was thinking about posting in the fun thread, but it's not really a fun topic, though it may not be culture war either since I expect most people to be on the "this is bad" side. Maybe we should have a recurring "Butlerian Jihad Roundup" for posts like these?
Bots are taking over the internet. Corporate shills and (foreign) government propagandists have upgraded with virtual cybernetics. A related but lesser change is people using LLMs to reword their own posts (+ emails and other communications).
Some AI writing is obvious, but sometimes it's indistinguishable from (if not completely identical to) what a human would write. NYT has a quiz to distinguish human and AI writing. I did bad (3/5), but in my defense, I think most of the human examples are awful, making the quiz harder. See for yourself.
On Hacker News, it’s now so bad there's a new guideline, “don’t post generated/AI-edited comments”. Unfortunately, due to the extreme intellect of the average Hacker News commenter, it can be hard to distinguish their profound technological insights from even a markov chain trained on buzzwords. Indeed, looking at top threads I still notice lots of slop-like posts from brand new or previously inactive accounts, like this one. I've been sarcastic, but I really like Hacker News, and hope it finds a way to stop the slop.
Other networks are taking a different approach. For example, Meta has acquired MoltBook (the AI social network) in an effort to add even more bots to FaceBook. I’m joking — no wait, they may actually be doing that. Not content with the Metaverse, maybe Zuckerburg has become addicted to burning money on uncanny social experiments.
On the Motte, at least for now, I haven't seen any obvious bot posts. There were a couple AI-assisted posts (by "known" humans) over the past couple months that got called out.
How will social media evolve? Will people move to invite-only sites like https://lobste.rs and Discord? Will most people accept AI discourse as natural or even prefer it? Will AI discourse become so good that we prefer it? Right now, it seems even the best AI writing (prompted to be consice and human) is unnecessarily wordy and has certain tropes; but what if someone discovers how to train an AI on a specific human's writing, so that it's effectively indistinguishable?
There is no solution. There is no proof-of-work or proof-of-humanity that is not severely error prone, extremely laborious, or that avoids requiring some kind of totalitarian police state dedicated to monitoring every word written by a human, or every token outputted by every known LLM.
It can't be done, or at the very least it won't be done.
HN is the best parody of HN. There are plenty of (almost certainly human) users who could be trivially reconstructed by telling an LLM to write in the style of the biggest grognard pedant with arboreal-reinforcement of the anus it can envision.
Their attempt to ban "AI-edited" submissions is laughable, an attempt to close the barn-door after the horse was taken out back, shot, and then rendered into glue. There is no way to tell, distinguishing entirely AI written text is hard enough, let alone attempting to differentiate between an essay that was entirely human written, and one that took a human draft and then passed it through an LLM.
I intend to munch popcorn and observe the fallout. In all likelihood, a few egregious examples will be banned, alongside a witch-hunt that does more harm than good.
The majority of bot posts (that anyone can tell are bot posts) are spam that is caught by the moderators and never see the light of day. I can't recall a single example of us allowing someone in who we thought was human, and then finding a smoking gun that would make us conclude that it was a bot all-along.
I am on record stating that I do not see an issue with LLM usage, as long as a human is willing to vouch for the results and has done their due diligence in terms of checking for errors or hallucinations. I do not make an effort to hide the fact that I regularly make use of LLMs myself when writing, though I restrict myself to using them to polish initial drafts, help with ideation, or for research purposes. This stance is, unfortunately, quite controversial. Nonetheless, my conscience remains clean, and I would have no objections to anyone else who acted the same way.
None of the tools that purport to identify AI-written text are very good. Pangram is the best of the pack (not that that means very much). I've tested, and while the false positive rate on 100% human writing (my own samples) is minimal, the false negative rate is significant. It will take essays that have non-negligible AI content and declare them 100% human, or substantially underestimate the AI contribution.
And that is with no particular effort to disguise or launder AI output as my own. If I actually cared, it would be easy as pie to take a 100% AI written work, then make small changes that would swing it to 100% human by Pangram's estimation (or prompt an LLM to do even that for me). The tools help with maximally lazy bad actors, but that is their limit. Eventually, they won't even catch said lazy bad actors.
Asking the LLMs? No good. Even worse.
I took an essay I wrote myself (the only AI involvement was proof-reading and feedback, most of which I ignored). Then I asked Claude Sonnet to summarize the content in 100 words, then to itself write a prompt that would be used by another LLM to attempt to reconstruct the original.
I then asked fresh instances of Claude itself, as well as Gemini Pro, to write a new essay using the above as verbatim instruction.
I then took all 3 essays, put them in a single prompt, and then asked Claude, Gemini and ChatGPT Thinking to identify which ones were human, AI, or in-between.
You may see the results for yourself. Gemini's version of the essay was bad, and thus flagged by pretty much every model as either AI, or the "original" that was then expanded. The other two, including my own work, were usually deemed 100% human. Well, one is ~100% human, the other very much isn't.
Gemini in Fast mode:
https://g.co/gemini/share/0d4e6279bf8f
Gemini Pro:
https://g.co/gemini/share/119274d62e32
ChatGPT Thinking in Extended Reasoning mode:
https://chatgpt.com/s/t_69b3fad20c9c8191a27e3542685f20ba
Claude Sonnet with reasoning enabled:
I can't link directly, because the share option seems to dox me with no way of hiding my actual name.
Here's a dump instead-
https://rentry.co/oo4qkduk
Claude was the only one to correctly flag essay 3 as human, and that is likely only due to chance.
ChatGPT was the only model with memory enabled, and it failed miserably.
What else is there to say? Good luck and have fun while there's some hope of telling the bots apart from humans, if not humans using the bots.
With apologies to Descartes, "always has been". While cogito, ergo sum manages to demonstrate that I exist to myself (at least, I find the argument compelling), I've never been able to satisfactorily prove that the rest of the world and everyone else as I perceive it exists, and isn't some
big simulationdemonic manifestations or imagination.Yes, but consider a natural reply to a prompt. Previously, that was practically guaranteed to either be from a human, or bot that was (human) tailored for a limited generalization of that prompt. Now we have AIs that reply naturally to arbitrary prompts, it’s their purpose. Consequently, they produce natural posts, comments, etc.
More options
Context Copy link
Just a few days ago, I met a patient who was convinced that they did not, in fact, "exist". He believed himself to be a rotting corpse, and initially declined his antipsychotics on the grounds that a dead person had no need for medication (a valid argument, as opposed to a sound one).
After some debate, we decided to tell him that the drugs would prevent his "corpse" from decomposing and causing a stink that would inconvenience the rest of the ward. Pro-sociality intact, he found this a compelling argument, and swallowed them without any further fuss.
So no, not even "Cogito ergo sum" is foolproof. The universe, and the DSM, must account for even better fools.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link