You’re scared I’m going to share your name with companies? Why would you think I would do that? If that’s what you think of me, you shouldn’t be on my forum asking for lots of free help. It doesn’t make sense for me to put a lot energy into helping someone who doesn’t have a lot of respect for me, and it doesn’t make sense for you to put a lot of energy into your posts here if you don’t respect me.
Most people don’t try to withhold their real names or countries from me (especially not while/after asking for discounts).
Or did you somehow misunderstand and think I wanted you to share your identity publicly? I just want it DMed to me.
I used to talk to anonymous people often but I’ve been sockpuppeted repeatedly and people continue looking for ways to harass me today.
i wasn’t saying that. i didn’t mean that you will share my name with companies. since employers also search social media during hiring, i just don’t want to connect my real identity to career-related posts anywhere. it’s not about this forum specifically. and yes, i understood you meant DM.
i get the sockpuppet concern. can we do some other check that doesn’t link my full identity to this account? like a social media account?
If I don’t share your name with companies or anyone, then I don’t see the problem. Our DMs wouldn’t show up on hiring screener searches. And you could DM me with disappearing Signal messages if you wanted encryption.
I don’t see how an anonymous social media account would solve the problem. If I always ask for real name and country, people can only troll me once (unless they manage to fake a real name which is an extra barrier). But if I accept anonymous social media accounts as identity, a person with 5 of those could make 5 forum accounts and troll me 5 times.
agree. anon social media account won’t fully solve that. it is more than five years old tho.
the privacy issue is not about intention. much more private stuff gets leaked without anyone involved intending it. i’m extra cautious about this stuff from past experiences.
can you say more about how name and country will help?
I just explained that. This is a low effort question which doesn’t provide information about what you don’t understand or what else you want to know. You didn’t do your own analysis and you asked me to do more work. There were also listening problems and low effort questions in other topics.
You can DM me the social media account and I’ll take a peak but I don’t expect it to solve the problem.
And identity is just one issue. I remain concerned that you asked for a lot of help, praised our conversations, then called our conversations unproductive and said other negative things. So even if I didn’t ban you, I’d probably just stop reading your posts since I’m not convinced it’s a good use of my energy, and then you’d probably quit the forum anyway. When I tried to explain this and other problems to you, I got what looked like bad AI responses, with lots of common AI wording patterns, that didn’t really engage with my point. You didn’t appear to demonstrate conceptual understanding of what you’d written or what I said. So my concerns weren’t addressed. And I was already worried about using AI and lying about it from other posts.
you lied about using AI in multiple posts (more than half that I checked came out as 100% AI or similar; unprecedentedly high for not using AI)
you can’t think of ways to prove (in the normal soft sense) that you’re not using AI to write; even though it takes anyone else like a few seconds to think of something; the proof standard is lower than or similar to speedruns and creating proof for speedruns is self-evidently accessible.
you are evasive about your identity even though no one is asking you to publicly reveal it.
you seem unable to put yourself in Elliot’s shoes long enough to even think about this in a basic way.
The ID leaking excuse is insubstantial with the context you have now. You are not emailing your passport to some government bureaucrat or corporate wage slave or insecure kyc platform or porn website. Offering anything would be more than what you’re currently doing. yet you offer nothing and expect the privilege of discussing anonymously after lying.
if your natural writing style did trigger AI detection, it would be quite interesting and valuable. But I think you know why it’s not valuable (because it’s not your natural writing style). I just went and checked another one of your posts (chosen at random) and it also is detected as 100% AI. Yet your recent posts don’t show up as AI. Can you explain that difference? Why did your posts go from mostly seeming AI generated and now don’t?
Others have come up full-AI, mixed (part AI and part human), AI-polished, and shorter ones before or recent posts in this thread show as human. I’ve checked about 10 or so.
I’ve found it to be quite good and has improved over time. ~1 year ago I tested some generated text against it that was detected as human; but I went back recently and tested the same text and it’s now properly detected. They make improvements every few months which is nice. I’ve never seen purely human text get classified as AI.
there are many things i object to in this but i don’t wanna go into everything and look nit-picky.
on the privacy and safety point, you can see that i posted a digital hygiene link before this issue even came up. even my friends get annoyed with me sometimes for being too cautious about this stuff. i do think i’m more cautious than most people, but that’s because of actual past experiences.
for that reason i’m sympathetic to Elliot too when i say i understand the sock-puppet concern. i’m trying to address it in a way that could work for both of us. it’s not that i think Elliot would intentionally share anything. it’s just that the link would exist somewhere and i want to keep that as minimal as possible. i’ve come across multiple recent threads about doxxing and harassment on the latest page in the past few days itself. so given that there are people with hostile intent, and AI is making exploits cheaper, i hope you can see why i’m concerned just like i can see why Elliot is concerned.
i came here to discuss ideas and improve my understanding, not to troll. can you share the clearest example where you think even a charitable reading would say i was trolling with AI? i think one concrete case would be easier to discuss than detector scores or broad impressions.
In my experience, this kind of pattern works OK for coding but not for philosophy writing. I can steer an AI to get good code but not good philosophy text. I have to be the primary writer, and give AI smaller roles, to get good results. AI can be more helpful for writing technical documentation as a time saver. A difference with that kind of writing is there are no controversies, criticisms or arguments involved, and it tends towards simple factual sentences. I haven’t tried using AI for fiction writing but I’d have low expectations currently.
Also if I wanted to talk to AI I’d generally just talk to Claude myself. No one should think they’re talking to a human but actually they aren’t. I don’t want people to be tricked and have their time wasted.
Also people shouldn’t post AI arguments, even in their own words, without disclosure. Because then person they’re talking with is arguing with AI, not with a human who knows what they’re saying. There are more borderline cases like when an AI reminds you of arguments you already knew or it tells you some that you look up and actually learn about yourself.
Are you using the Discourse app or a browser? I have found the app buggy on iOS for months, maybe over a year – it works ok to read the forum but the UI often breaks when I try to reply to anything. It’s not consistent and sometimes has worked then broken again. There’s also at least one third party ios app compatible with discourse forums but I forgot the name.
I wonder if it’s just not good compared to you. Maybe it’s better than a good amount of current philosophy writers?
Aren’t most philosophers bad? Also, maybe it depends on how much the AI was trained on a given philosopher or philosophy topic? Maybe it’s much better at writing text based on Socrates’ ideas? Perhaps I’ll do some testing and report back.
Can you give an example of smaller roles you assign it?
Yeah, but because any AI writing has to be disclosed, if it’s 90% human, 10% AI, or 50% human, 50% AI (or even 20% human, 80% AI writing) you’re probably not going to get the same experience as talking to Claude yourself.
For example, the AGI thread with you and Oracle. I don’t know what percentage of his writing was AI, but it was still a super informative thread in my opinion and I doubt you would have had that experience talking to Claude. If you just talked about it with Claude, maybe Claude misses things. Maybe because you’re directing it, it takes some info for granted and doesn’t ask the right questions (because it’s not curious about anything and isn’t able to understand things conceptually, whereas Oracle was obviously very curious). I learned a lot from that thread. That’s part of why it was so surprising that Oracle claimed your back-and-forths were unproductive or something. You gave a ton of value, even if you were concerned his writing was AI. It reminds me of scenes in One Punch Man where citizens deny Saitama’s greatness and attack him, despite all he’s done lol.
Yeah, it seems pretty obvious you shouln’t post AI arguments you don’t learn yourself. Like how do you know the AI is making a good point if you don’t understand it? But, I’m sure people do it.
I’m using the browser. I haven’t noticed any issues until recently.
Suggest objections that other people would make that aren’t already covered.
Suggest things missing from my essay that I should have included.
Do research and find evidence that contradicts or supports my essay.
Find academic papers that make relevant arguments (I’d usually be more specific like papers that make similar arguments to one of my arguments or papers that argue against one of my claims. Or I’ll know people argue for P using reasons or approaches W, X and Y, all of which I already reject, and I’ll think someone should have tried Z approach and ask if anyone has.
Ask AI to find cites for things you already wrote that could be added (don’t just add them – treat them as leads).
You can ask AI to search for a quote of someone making a specific point and give you a cite.
If you talk with the AI a bit about your subject, this can often trigger your own memories. The AI can say something wrong that you already know a refutation of or bring up a sub-topic you didn’t mention.
I hear people get AIs to interview them. I haven’t experimented much with having the AI ask me philosophy questions but that seems reasonable.
I think its pretty bad compared to a lot of random philosophy professors or grad students. Maybe I’m wrong and the stuff I see (e.g. reddit, papers, blogs, books, podcasts) is unrepresentative and a lot of philosophers are worse and just don’t say much in public. But a lot of AI philosophy writing is worse than a lot of non-philosophers who post on LessWrong, Reddit or Hacker News.
Yeah I think this can be useful for at least 2 broad and different categories.
One is if you just have very little interview experience, or feel you suck at them, or struggle to think of what to say. It can just be general practice.
But even if you are good at them, it could be good if your interview is going to be from a well-known source. For example: if you are interviewing for a major company with documented interview practices, or if you are going to be interviewed by a public figure.
Related but slightly different: I know someone who was preparing to defend a thesis and one of the people who they would be defending against has had a lot of work published. So they asked an LLM to ask them the kinds of questions that specific public academic might ask. My understanding is that this was helpful for them to get ready for the defense quickly.
As someone who is probably irrationally wary of AI, asking it to help you find stuff like quotes or arguments or facts (whether in philosophy or anywhere else) does seem useful. Of course you should check after to make sure it was correct. But it can be very time consuming to find that sort of thing on your own so I can definitely imagine how it would be helpful.
Oh I didn’t mean it for practicing for an actual interview (which would be fine too). You can have the AI interview you to help you get your thoughts out as preparation for writing an essay.
One common theme across lots of these use cases is that they still require you to write down what you thought in your own words, rather than having the AI write down “your” words for you.
these are private forums. getting you in would require giving up a lot more privacy than just name and country. it would also link those spaces and people to this dispute which i definitely don’t want.
I can try. I already tried addressing and clarifying some of this above and asked in #80 and #91 for the issues to be made more concrete because I think that’s where I can answer best.
Can you give me the clearest example for each?
Edit: I understand that AI concern is broader so I can try to address that separately instead of only through specific examples.
