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 -
Nate Silver just accidentally posted a link to an AI slop article. A quick delve into the article text makes it obvious that the contents were blatantly copypastaed directly from the output of ChatGPT. Various GPT detectors also agree, giving it a solid 100% confidence of being slop. Unfortunately, it seems that nobody in the replies even seems to have noticed or cared.
I'm of course already used to my google searches being clogged up by black hat SEO slop, but I expected it to just live in the shadows quietly collecting clicks and boosting pagerank. So it was sobering to see an aislop article just get posted like that by someone I regard as at least somewhat intelligent.
What does this say about the world? Are normies, even somewhat intelligent ones, incapable of distinguishing the most obvious stinky smelly chatgpt output? Or did hundreds of people read the headline and drop a snarky comment, and not a single one bothered to read the article? It's either a depressing anecdote about human nature and social media, or a depressing anecdote about the lack of intelligence of the average human.
Of course aislop grifters should be fedposted just like indian call center scammers, but sometimes I can't help but feel like the victims deserved it. But when they bother me waste 5 seconds of my time again, I am right back in fedposting mode.
Edit:
Since you idiots are out here defending the slop, these quotes are hallucinations:
Here's the full recording of his talk and you can check the Youtube transcription: https://youtube.com/watch?v=MPt8V3MW1c4 And before you ask, the fake article specifically claims these fake quotes were said at his Harvard talk, not at some other time.
So again the AI put totally false words into somebody's mouth and you apologists are defending it.
Alternative phrasing: are normie journalists, even somewhat intelligent ones, incapable of putting out articles better than even the most stinky smelly chatgpt output?
I mean, why is this a surprise development? Oh, the article has a bunch of made-up garbage, you mean unlike all the rest of journalism where everything is true and the media never truly lies?
Why should I give a shit if a journo "hand authors" his latest piece of crap or just pushes a prompt at an LLM? Text generation is their speciality, and at this point I'd be stunned if most of them weren't 90th percentile in general writing quality in comparison to humans. How many journalists do you even know of for their writing quality? Hunter S Thompson and ?? Hell, just prompt the LLM in the right way and I bet you could get to 99th percentile no problem.
Do you even have an argument here beyond just the words "aislop"? Can you articulate a point that taboos the phrase slop and similar terms?
Journalists, even partisan hacks, almost never lie. They produce propaganda through omission, and when they need to spread falsehoods, they will always hedge it though "anonymous sources." It's actually quite nice that they will explicitly tell you when you're about to read something made up.
As much as I hate journalists, they are quite good at writing. Some no-name journalist at AP has gone to school for writing and honed his craft for years. His writing is almost certainly in the top 99th percentile of writing skills, and certainly far better than yours or mine.
To repeat my previous comment:
Your opinion of journalists is too low. Journalists are rabid partisans, but they're generally not very stupid. But they pretend to be stupid quite often in order to serve their side.
Absolutely fucking impossible. Just try and give me a single example of LLM output that's remotely comparable to quality human work.
Hm... this was an attempt, and there's a few minor gaps that I spotted even with my weak knowledge of the field, and the output has demonstratably worked, but it would be interesting to see a skeptical eye analyzing it.
Firstly, this is not really a work of writing but just a bunch of bullet points thrown together. If we're talking about comparing the quality of prose, this isn't the ideal medium.
But let's delve into the slop:
I think you already fucked up. masks are more of a thing for electro etching. Not really the correct concept for ECM, but slopgpt doesn't point it out.
Nope. This is the oft derided slopgpt sycophanthty. What's unique about it? How is it different from "normal?"
What specialize electrolytes? Is it really widely available? Is chatgpt assuming you can just sprinkle some table salt in the water?
Does it reduce operational costs? Is electrolyte a major operational cost? How much does it cost?
False. ECM uses electrolysis as a mechanism of action.
Jewgle says that 4000 series aluminum is prone to cracking during welding, not during machining.
Voltage isn't the factor that matters when talking about excessive material removal.
Sounds wrong. The silicon should be alloyed into the metal at a molecular level, not embedded as silicon grit or whatever.
Very suspect logic.
Totally false and retarded.
Slopgpt just assumed this is a DIY process. What if we're talking about doing a commercial process with this setup?
pH just appeared out of nowhere with no mention of it before at all.
Sounds wrong
I don't believe acetone affects aluminum at all.
Beyond my ability to evaluate.
You already knew it would work before asking chatgpt, and knew what you wanted to do. How much value add did asking it actually provide? Did any of the problems that it predicted actually happen?
The specific series of events here was that I'd had some rough familiarity with ECM in general from the ctrlPew world, saw this video, tried to isolate the rough theory, and was curious whether it was possible to expand to cutting through thin metal with fine detail, asked the LLM those questions, sanity checked the numbers for any massive math or chemistry errors, then asked a chemist I knew regarding safety concerns. I had never tried it before, nor have I found any documentation of this specific approach beyond etching layouts that warned about leaving the process so long that inserts fell out (and shorted). After I ran it, I found a few other similar attempts, albeit still not entirely overlapping.
There were a few things it provided that I didn't think of or read from what I could find. Some of those would have been obvious to a human with machining background -- flow rate of fluid at higher cutting speeds are a common thing with air-blast or coolant flooding in traditional CNC, and the theory behind the pump assist here is related to that -- but others were not. I did not know a Faraday's Law could be applied to this case, and specifically having 'material removal rate' rather than any synonym helped me track it down to confirm rather than get a half-dozen links about inductance.
Conversely, you're right that there are things it definitely neglected to say. Getting any paint or adhesive to stick to nearly any common type of aluminum without a lot of prep and elbow grease is a massive issue that both is a common fault in other environments like aviation or building work, and also ignored by a lot of novices; the transfer paper for getting fine detail vinyl from backing to workpiece was a pain in the dick.
ECM isn't really the right tool at all for this job; it's just a tool I had relatively available, and mask vs paint vs tools specialized for the work piece is part of that. Yes, a process like [PCM](https://youtube.com/watch?v=bR9EN3kUlfg] would near-certainly have better and faster results, especially at scale, but I don't exactly want to be messing with high-pressure ferric chloride, either. The question is whether minimizing surface area and using readily-produced masks could work at all in environments where creating a form or to-purpose tool was not practical (in this case, very bad detail work), for any function at all, rather than whether it was the right way to go about it. That's why it was a process with very little documentation available, and worth asking either an theory-focused expert or doing this sort of search.
There's a fair criticism along it failing to XY problem me, but in this case I was genuinely interested in the method as much as the end result.
Commercial ECM dielectric tends to be pretty pricey anywhere I could find willing to ship to a residential area, and worse, only available in fairly large quantities that vastly exceed my use case. It makes sense for commercial ECM machinery, since it can be filtered and reused and you need volume but don't replace it much, but a lot of what the literature points at it performing better at wasn't practical for small-scale operations without specialized tools. I haven't been able to find what typical costs for conventional ECM were, but for my use case it would have vastly exceeding the rest of the costs by an order of magnitude (which was a Harbor Freight pump, paint and vinyl, a few drop shipment blanks, and an old power supply that had been on the bench for a while), and probably cost more than getting it made by a local CNC mill company. Mixing my own to replicate commercial electrolytic might have been possible at a reasonable cost (literature suggests sodium hydroxide and sodium nitrate or potassium nitrate with a bunch of additives?), but in turn would probably have gotten me put on a list.
Table salt worked, though I did measure by mass, mixed for a long period under mild heat, and passed through a coffee paper filter to exclude large granules, and the LLM did not mention that. Dunno if it was necessary. I've also seen ethanol or polyethylene glycol recommended as additives (presumably by humans).
It probably depends on the specific material, but 4032 is pretty famously hard to machine, and it was the material I used here (no, I have no idea what thin plate 4032 was doing in a drop shipment). I'd tried high speed milling this stuff previously, but even with an 0-flute and tiny depth of cut it was extremely prone to gumming up the cutter in mere inches, and even when dialed in would split and crack at corners, especially sharp inside angles.
I tried 12, 24, and 72 volt, and localized buildup was much worse at higher voltage. Voltage may not control removal, but it did seem to control how quickly material to remove popped up.
I don't have the tools to analyze this one, but finish was Not Great Bob in 4032 or 4045, especially on large flat plates, and whatever it was that did precipitate was much harder and less prone to clearing without blasting the area with fluid than with 6061.
Yeah, that's one of the more serious errors; 4000-series might not be the most corrosion-resistant aluminum alloy, but the 5% or higher-silicon ones are pretty high up there.
I think that the correct answer would to question wtf we were doing. There are few, if any, arguments for running this approach at scale or commercially.
At least some of the literature (eg here) suggests controlling pH and especially localized pH weirdness is a valuable thing for the normal ECM processes. Didn't seem to matter here, and it definitely isn't clear about why, so point.
The acrylic paint and acetone, or the vinyl pull? Unless it's had a couple weeks to cure acetone sloughed the paint right off. Vinyl was messier, with large surfaces of the cricut material falling off and tiny ones turning into a million flakes.
There are some alloys that you don't want to mix, with 2000-series being the worst, but yeah, 4000-series usually won't care. I did actually have problems with nail polish remover causing a gray mottled texture, but I think that was just contamination and it buffed out.
What I mean to say is that you aren't even doing "ECM" at all. You are doing electro-etching. Kind of nitpicking on the terms here but disappointing that it wasn't pointed out.
I think that would usually be an issue with the speeds and feeds. One thing with aluminum is that you are actually punished if your feed rate is too low. But I'm not a machinist so maybe that's not it.
Anyways chatgpt is a useful tool for vibing with, but the output result definitely isn't publication quality.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link