In the spirit of hydroacetelyne's and Terracotta's conversation here, I've gotten trapped in car repair again. My daily driver's an older one -- and I strongly prefer decade-old used cars to start with -- so there's been some fun and !!fun!! involved.
This weekend had two targets.
I'd had a couple stalls on idle pop up recently, and that was the most critical; so far they'd only happened in parking lots while Christmas shopping and the car had restarted after seven or eight cranks, but 'tis the season for late construction and bad traffic and neither combine well with a dead car on a road. Nothing on an OBD-II reader, no check engine light, battery is older but still testing good, alternator power is good, wire harness is fine, spark plugs are less than six months old and were pregapped to this car (and had been fine for those six months), no obvious loose connections, issues do not seem to be tied to fuel level.
Crankshaft position sensor replacement. Bizarrely, it's actually easier to get to in this make and model than the oil filter, and while that's damning with faint praise (how do you get the oil filter wrench in there if when some hurried Jiffy Luber overtightens it?) , it's still a nice change from some other similar cars.
Still managed to cut my hands working on it. You'd think I'd listen to my own advice.
((To nerd out, both Grok and ChatGPT were able to diagnose this specific issue from ""I had a <Year> <Make> <Model> which had the engine stop suddenly while in a parking lot. Trying to restart the engine took several attempts. A tester gave good battery, alternator, and starter electrical values. What causes are likely?" On the other hand, neither of them gave complete and correct advise on replacement, both warning that I'd need to get under the car to fix it and not warning about how important proper fit and tightness were.))
Other one was a stereo head unit. I'd bought the kit (and wiring harness, and a 'one-size-fits-none' adapter from asian to DIN rails) a couple months ago with the dual goals of some alternative to country music, and a working backup camera, then left them on a shelf. Already too late to work on it before the weather gets chilly, but I was already cold, so here we go. The wiring harness was easy enough, though I'm glad I bought a new soldering kit -- absolutely love these things compared to the typical Weller. Asian-to-DIN adapter was absolute trash: not only did it require random parts of the plastic be snipped off in the instructions, the instructions weren't even close to right, so I ended up spending an hour in the cold with an X-Acto doing fit tests. Backup camera was found hard and left untried for now: finding a path to get from the head unit to the license plate on a hatchback that doesn't risk wires getting pinched every time I load up on groceries or Lowes detritus is gonna be some doing.
Separately, also helped a fellow robotics mentor/friend out with his recent purchase. He bought what he accurately calls a 'clapped out' sedan with the intent to get it into normal working order and flip it, and oh my god there was so much spray paint. On the upside, whoever started working it probably wouldn't know hot to prep cardboard for a can of rustoleum, but on the downside there's enough plastic on this thing that it didn't always care and what the actual fuck. Light blue spray paint. I hate those walmart-grade LED strips on cars, the previous owner used sheetmetal screws to hold them on, and they still seem sane compared to the spray paint.
gattsuru
20hr ago
DS9 has its stinkers, but at its heights it's easily the best Star Trek by a wide margin. I'd still put Babylon 5 a little higher, but it had steep competition from DS9 in a way that no other Trek of that generation, or the current gen, could or was even trying to supply.
gattsuru
1d ago
That's an impressive amount of effort for a Halloween costume, and some remarkably unobservant partygoers.
gattsuru
1d ago
... I'm almost afraid to ask, but did you have the pig's head facing in, or out?
gattsuru
1d ago
Yeah, that's probably a more honest engagement with the events. I've just seen a lot of people say it was a big important deal that tells us about falling modern standards, so it really bugs me that it's just such a mess.
gattsuru
1d ago
Is that…Trek?
Nextwave (cw: sound, mild profanity)! Only Warren Ellis comic I can unconditionally recommend. Very short series, sadly.
Not exactly unusual—didn’t Scott write about ADA enforcement in these terms? The main limiting factor is the difficulty of bringing a case. Technology has to have reduced that cost, so a given org can target smaller companies.
I think so, but I can't find it now. The ADA (and the Texas abortion pill thing) are laws or at least regulations on the books, even if they're probably getting stretched to their breaking points. A lot of this stuff isn't a strict rule, or even necessarily written down anywhere, so much as it's just Understood at ultimately one-on-one scales. Sometimes that's unavoidable: in aviation, I think you could get five opinions from three DERs if you bring up flutter analysis. But it ends up in a world where a lot of things are theoretically allowed, and you can even find people doing them quite publicly, but also prohibited.
Over five percent of China's power comes from fission plants, and that's underrating it since they've got very high uptime compared to on-demand plants. As for why it hasn't scale up faster, China's political classes had very obvious mixed feelings about dependence on foreign-produced infrastructure for a long time, which only went harder as western regulatory overhead killed western nuclear power. While they've theoretically had 'domestic' production of nuclear plants since the mid-90s, they didn't actually manage serious production of the CNP-600s until 2010-2012... at which point the Fukushima disaster and its political fallout lead them to go back to the drawing board and start the production cycle again.
But they've put >3 GW of fission power online just in the last year. As bad as their political situation is for power construction, it's still beating the west's.
gattsuru
1d ago
Google's been trying something similar with Sima 1 and the more relevant Sima 2, though I'm not seeing anywhere near as much information about what the model parameters and configuration were for that one. Qwen-2VL-7B seems, intuitively, way too small to make this sort of deep analysis and decision making, and it's kinda weird that a lab environment didn't go to something like Qwen-2.5VL-32B. But 7B was also obscenely good at captioning videos and making problem solving analysis from it, and people had gotten some results, if not great ones before.
Unfortunately, a lot of the value in the study is going to depend on exactly what and how they tested the model, and there's really not enough detail here. An hour-long autonomous play session of 'finish this mission' is the big selling point, but I don't know Genshin well enough to say whether a) that mission was nontrivially different from training data or b) that it involved more than 'follow quest marker, spam A at enemies when lock-on-button does anything.
It'd be interesting to see more information about how well these models handle completely out-of-training problems, though. I've talked about using a MineCraft mod to see how well a model can create a 'new' solution, but these sort of games are trivially easy to present completely out-of-training problems, ranging from stuff as trivial as an enemy or attack that's changed color, all the way up to completely novel gameplay mechanics (eg, FFXIV threw in a "change color to reflect attacks" mechanic several years after initial release). I wouldn't expect an LLM to possibly one-shot every version of this, and some probably aren't possible for architectural reasons (eg, even if a model could go from vanilla minecraft to GTNH from, no plausible memory-constrained implementation would have the context window for even some mid-game recipes), but I think it'd say some interesting things regardless.
gattsuru
1d ago
To some extent, it does happen: one of the many swerves of the Peanut The Squirrel saga from last year was that the squirrel's owner was doing very gay4pay-looking onlyfans while having a wife. Which could be bisexual, but you can be bisexual and work blue collar, and a lot of the framing was more 'what straight guys think gay guys want' than what even the bottomiest gay guys actually want.
Which points to a part of the problem, if you're a straight guy trying to sell to gay ones. Look at fantasy (art or written) gay4pay or orientation play, and there's a lot of stuff that's not just going to be uncomfortable for a straight guy to try (though there's definitely a lot of that: forget taking a dildo, who wants to wax their chest), but will also just be very hard to credibly recognize or sell.
On the other hand, SquirrelDaddy was an OnlyFans making in the top 1% and maybe top 0.1% of male earners, so maybe my tastes are just weird.
That said, the available business strategies are less viable. The nice thing about gay courting's that there's somebody into everyone, and sometimes the breakdowns can come in surprising ways (eg, one of the top 0.001% male earners on OF is outright obese)... but there's not necessarily many people into anyone. The entire male side of onlyfans pulls in about a fourth of the revenue that the women's side does, and the top (hurr hurr) is saturated with tops that are, to be blunt, not possible for the average man even if he wanted to. Any of the highest-end creators in any space tend to have people who are doing that career as a full time job and a half, but to be blunt, almost all of the top male creators are genetic freaks. Ain't no amount of zinc and pineapple that's doing that.
I wouldn't recommend a gay (or 'I'm just posing naked for my fans, which I totally assume are women') actor go to Dubai for sex work, though. The UAE's more extreme punishments are theoretical even for their own inhabitants, but they have made life miserable for tourists in the past. While those punishments theoretically include M/F situations, they're much more likely to hit man-on-man ones.
gattsuru
2d ago
China actually started up a molten salt 'thorium' (eg, starting with uranium, then moving to thorium) reactor last year, with the first full thorium cycle this November. I'm not optimistic about its effectiveness, but that's more because it's a lot more complicated than it needs to be, rather than net energy or net cost problems.
A lot of the various cost problems with nuclear plants reflect political willpower, rather than actual material costs. That's most serious in the United States where we've intentionally made them several times harder to produce at the same time that the control and construction technology has gotten much much better, but most western governments have done something similar. (or just had politicians launch rockets directly at the construction sites.)
There's a revealed preferences sense where, if you can't solve those political problems, you can't produce power at price, and it's not entirely wrong. But it's misleading to treat it as a physics problem.
gattsuru
2d ago
Coyotes are pretty well-hated and oft-culled because of their attacks on pets and livestock, though, and there's also the 'dingos ate ma baby' option of simple incompetence. That said, if you really want to go nuts on coyote conspiracies, the degree that coyote populations have exploded and the individual coyotes themselves have gotten much smarter in <10 generations is a real fun question.
For fun conspiracies I actually believe:
- Piggate wasn't real. For all it Took Down A National Government, Cameron was already a political dead man walking before the drop, and it was just a really convenient way to force him out without actually engaging with the political controversies that had undermined his party. The same behavior is totally consistent with an already-unpopular prime minister getting smeared by a schmuck he'd pissed off badly enough, and then found that none of his 'friends' were willing to pay the political capital to back him up.
- There was a coverup one direction or the other for the Bloomberg Supermicro thing: either a lot of people who could prove it were told not to do so at the risk of destabilizing international relationships, or a lot of people who could disprove it were told not to do so lest they destabilize US financial markets (and get blackballed). I'm not very confident on this one, but it's just such a weird goddamn story.
- A number of serious industrial or transportation sector accidents were really Reinvented Suicide As A Group Activity, but various incident analysis groups have instead used them for purposes ranging from getting unrelated political goals to deflecting from local political or social problems to just shaking down foreign businesses for cash. There's been a handful of these situations where jurisdiction friction has lead to them getting 'caught' -- aviation is particularly prone to it, with SilkAir 185, EgyptAir 990, and the recent Air India 171 -- but I think they're far more common than anyone wants to admit or even mention publicly, especially since there's a risk that publicizing them could incentivize further or larger attacks of the same kind. Basically, most large countries have a bunch of CEAF 5735 in a thousand different fields. The SL-1 incident is the safest one to mention, but there's some electrical and chemical processing examples from the tens to hundreds of deaths.
- A lot of 'advocacy organizations' related to industry regulations are wholly-owned government groups, and are explicitly-but-nonpublicly threatening to bring the weight of those government orgs to bear if targeted companies don't agree. Yeah, boring, almost too obvious to be worth mentioning for the obvious cases, between Ofcom and NCOSE existing, and
XTwitter's recent fine in the UK. But there's a lot of these orgs running at <100 person levels regulating through smoke-filled backroom deals; a lot of what's 'weird' about the modern era is just the ability of those orgs to impact companies with large impacts but not the large scales of pre-internet companies.
gattsuru
2d ago
I'd be surprised if we don't still have at least discount theatres in twenty years playing movies that have been out on 'conventional' digital services for months or years -- without the weird cost behaviors downstream of the studio system, movie theaters have sizable fixed asset costs and trivial operating costs -- but they're definitely going to be labors of love.
Weird that it's become one the more implausible Mystery Science Theatre 3000 assumptions. Or, hell, Snarkout Boys and the Avocado of Death bits.
gattsuru
3d ago
It's unfortunate that the most serious predictive engagement with the question of an AGI breakout remains a novella with the framing of a My Little Pony fanfiction. That's not as damning a praise as it sounds at first glance, but it's a strange issue given how much ink has been spilled and how omnipresent the technologies have become since 2012.
gattsuru
4d ago
Gorilla Tape's one of those things that seems like a marketing scheme -- I'm not very impressed by Gorilla Glue -- but it's genuinely worth keeping around. Some other fun options:
- VHB tape. Doesn't have any of the nice benefits like easy reapplication or simple tear. It just sticks things together, and never lets go, and it's just freakishly strong and vibration-resistant. Temperature cycles, weather conditions, high winds, literal fire, doesn't care. I have bent quarter-inch aluminum plate trying to remove things that had been stuck on with (too much) VHB: outside of some careful solvent use, the only real safe way to remove the stuff is to saw it off, and it will leave residue.
- Gaffers. Yes, it's hilariously expensive for glorified duct tape. If you absolutely, positively, need it to last less than six months on any smooth surface, and come off without leaving residue no matter how absorbent the surface material, there's no competitor.
- Camper Mounting Tape. This is a fun one because it's useless for adhering two things together. What it does do is provide an easy, fast, seal. Waterproofing? Slap. Vibration dampening? Slap. Need a weak vacuum to hold overnight? Slap.
- Self-fusing silicone tape. It can't replace every possible use for vinyl or pvc electrical tape -- it's very bad for sealing off actual connectors, or for small areas, or anywhere you can't wrap fully around the surface -- but >90% of situations, it's just so much more reliable for long-term application and several times more pleasant to actually work with. Zero residue, great UV and chemical resistance.
gattsuru
4d ago
I can't imagine what a similarly-constructed paper from a progressive view would even look like. The only half-decent analogue I can think of is if the progressive response contained poorly cited infographic statistics, in which case it would at least gesture toward empiricism and the ways of knowing endorsed by the psychological sciences.
We did have recent Darwin discourse, if you want in-the-wild examples.
I'm not sure that 'the ways of knowing endorsed by the pyschological sciences' are anywhere near what you want to motion toward as a different class of thing, or that it's clear from either the rubric or the typical essay in this category of course that such empiricism is actually supported or required, given the quality of academic psychological research. Maybe if schools weren't treated the Stanford Prison Experiment like a real experiment rather than a play it would have bite.
But ignoring that for now, there's a lot of pretty well-regarded sources that are respected in modern psychology and have little more than ipse dixit behind them. I'm extremely skeptical that a writer pulling from Julia Serano to talk about trans rights would have gotten this style of response, but they've got about the same experimental foundation.
That's... kinda the problem. If the quality of thought and writing from the graduates of these programs were better, you could just motion about this slop being slop. But then you look at the professor's response, and it's not like it's doing any better, either! Look at the middle lead from the professor:
You argue that abiding by normative gender roles is beneficial (it is perfectly fine to believe this), but to then say that everyone should act the same, while also saying that people aren't pressured into gendered expectations is contradictory, especially since your arguments reflect a religious pressure to act in gender-stereotypical ways. You can say that strict gender norms don't create gender stereotypes, but that isn't true by definition of what a stereotype is. Please note that acknowledging gender stereotypes does not immediately denote a negative connotation, a nuance this article discusses.
There's a lot of this whole disagreement that makes me want to slap everyone involved -- including the student -- in the face with an embossed copy of "https://www.lesswrong.com/posts/cFzC996D7Jjds3vS9/arguing-by-definition", and I get that a) they probably haven't read it, and b) the professor has to write comments on a lot of bad essays and is only getting in national news for the worst. But look at that claim, and compare:
Gender roles and tendencies should not be considered “stereotypes”. Women naturally want to do womanly things because God created us with those womanly desires in our hearts. The same goes for men. God created men in the image of His courage and strength, and He created women in the image of His beauty. He intentionally created women differently than men and we should live our lives with that in mind....
Gender roles and tendencies should not be considered “stereotypes”. Women naturally want to do womanly things because God created us with those womanly desires in our hearts. The same goes for men. God created men in the image of His courage and strength, and He created women in the image of His beauty. He intentionally created women differently than men and we should live our lives with that in mind.
Yes, this is just ipse dixit and incompetently written, but it's not making the argument that the professor is criticizing; to rephrase it in left-friendly terms the student's argument is that a lot of what people present as the result of stereotypes are really underlying interests (aka the Damore), and people would enjoy their lives better if they were allowed to act in alignment with those goals. This might be (almost certainly is) wrong! But it's not the same as "to say everyone should act the same". Worse, the professor's contradiction between "everyone should act the same" and "while also saying that people aren't pressured into gendered expectations" is a textbook philosophy error.
Or, later, compare the professor's:
Additionally, to call an entire group of people "demonic" is highly offensive, especially a minoritized population." You are entitled to your own belief, but this isn't a vague narrative of "society pushes lies," but instead the result of countless years developing psychological and scientific evidence for these claims and directly interacting with the communities involved.
to
Society pushing the lie that there are multiple genders and everyone should be whatever they want to be is demonic and severely harms American youth. I do not want kids to be teased or bullied in school. However, pushing the lie that everyone has their own truth and everyone can do whatever they want and be whoever they want is not biblical whatsoever.
The latter is written very poorly, so for a casual reader, the confusion is understandable. But diagram the sentence out. "Society pushing the lie" "is" "[D]emonic", not an entire group of people. There's a fair critique that the student isn't engaging with the argument being presented, but in turn, it's undermined if the academic measuring this stuff can't do much better.
gattsuru
6d ago
This rather famously resulted in some awkward loopholes around the capybara. Thankfully, 1800s Catholics had not yet discovered the swamp rabbit.
gattsuru
6d ago
gattsuru
6d ago
There was plenty of consensus that he was "bad" in some nebulous way, but when I asked repeatedly what was wrong I was only ever given vague runarounds and examples of posts that proved my point like this one, where I disagree with Darwin's political point, but in terms of debate etiquette and rule-following his detractors are massively worse than he ever was.
No.
There's a difficult question in the difference between "your product" and "my thinking or perspective".
For a toy problem, I threw together a short story today, about 1200 words in the original draft, zero AI. Wrote it in FeatherPad (a notepad-like), so not even spellcheck.
Except no, that's a bit of a lie. I Googled a well-known phrase about birds in gilded cages, because I wanted to play on the original text, and Google's awful built-in LLM did give me some interpretations, even if not the one I actually went with. Still, I can't say I completely avoided influence from it. In this case, I am doing the thing and the thinking, the AI's just helping save time doing it. Am I outsourcing my product to Google? To a song from the 1900s?
After finishing the draft, I shoved the full text as a file upload into Grok with a request to check for spelling errors, overall coherence, typographical errors, and clarity. It caught a dropped fragment I didn't, and had a few suggestions. I took some, and didn't take others, but every keystroke going into the final product I wrote by keyboard, no copy-paste. In this case, these fixes are something I could do, but some parts probably not as reliably as the AI, while other parts, the AI is doing stuff I don't want.
Okay, what if I wrote the original really badly: a rough outline that hit all the same story beats and general concepts, but without any of the stylistic techniques, writing quality, or many specific segments I wanted in the story. That'd clearly be against the rules for where I submitted the story, and probably not result in anything nearly as good. And since I'm not a good writer, that's really damning it with faint praise.
Actually, let's try it: Grok (story) and ChatGPT (story).
For a tl;dr, not great, not terrible. Definitely didn't get the themes down, and the humor's not great, but ChatGPT's tone is mostly on the right tack, and I could probably inject a lot more dry comedy if I pushed it harder or feed it back in on itself with instructions to crank those aspects up. I'm not exactly a prompt expert. They did a good enough job on the 'enrichment' lists that I'm pretty glad I didn't try them before writing my own version, because I'd have probably gotten stuck on them and not moved to the ones I did use. And even if those were 'worse' from a realism perspective they were imo 'better' from a thematic writing one.
These LLMs clearly produce a better product than the outlines I put into the prompts. Whether it's better as a product than the short story I actually wrote is pretty dependent on what you're measuring; I'd argue I went quality over quantity, but writers get paid on the latter. The only real ideas the LLMs shove out are the enrichment concepts, and I could demonstrably come up with different ones. Indeed, a lot of the token cost for the LLMs is poking it into even moderately-good ideas. Does that outsource my ability to do things that way?
... except ironically, a guest-mode story prompt came up with an additional good theme to introduce: spelling out the main character's lack of trust in himself. It's a theme that works great for most horror fans that this story archetype is built around, but it's not really one I could see without being prompted. I don't think it's necessary for the product, but it's definitely a different idea.
I dunno where on that line Ademonera's post falls. I'd be less happy with it if they were using ChatGPT to source names and events, and then just using the names and events without reflecting on what they actually meant (or if they even meant anything). I wouldn't really care if it were glorified spellcheck, since at least it's not another Grammerly user. And if it's that messy spot where they had an outline of all the material they wanted to hit, but let the LLM reshuffle it into a narrative... well, I guess I wouldn't really see the point.
But the problems are separate ones.
More options
Context Copy link