meduka
1mo ago
the base prior has to be that the 'woman wielding the unwieldy weapon was wrong'
I am, personally, never going to be upset about a woman (especially a girl!) carrying a knife around, no matter what the letter of the law says. It would be great if we lived in a world where the average adult man couldn't trivially overpower any woman they see due to inherent physical asymmetries, but that is not our reality. A knife makes it possible for a girl to defend against a predator.
Upthread there's a link about a gypsy grooming gang that was active in the city where this took place, the guy in this case may well be a gypsy, ergo I side with the 12 year old girl against the weird brown creep who was filming her.
I will concede that my support of the girl in this case is not unconditional. If there was evidence that she was trying to rob the "Bulgarian" man by brandishing her weapons, I would (begrudgingly) side with him. Merely scaring creepy men away from the place where you and your friends like to hang out is, actually, based.
meduka
4mo ago
The fallacy here is the assumption that in the counterfactual world where DOGE didn't cut these positions, the death toll would be (greatly?) reduced. The very blurb you quote suggests that in the best case a full time overnight forecaster provides a few minutes of heads-up via the emergency alert system. NOAA reports around 50-100 fatalities from tornados per year, with some outliers during extreme weather conditions. If we see an enduring spike in fatalities through 2025 and into 2026 and 2027, that would be evidence for your hypothesis. As of now, I'd say it's too early to tell.
How many deaths would there have been in Kentucky if there weren't Weather Service cuts? It seems impossible to know for sure. I couldn't find any information on whether an emergency alert was sent out in Kentucky (though I didn't look very hard) but if one wanted to make a case for these cut positions being important (rather than just accepting a statement from the Weather Service union) you'd want to dig up some data regarding how many tornados are "typically" caught -- and how quickly -- pre and post cuts to quantify the effectiveness of these local overnight forecaster positions.
I'm strongly anti-safetyist. The optimal number of yearly tornado deaths is not zero. The government could obviously reduce tornado deaths to zero if this outcome was prioritized at all costs. We acknowledge that there are diminishing returns and don't invest the resources to drive tornado deaths to zero. It seems extremely unlikely to me that the current resource distribution is optimal, though plausibly it's in a local minimum and moving out of it will cause some amount of pain.
meduka
5mo ago
I saw this unfold "live" and was mostly annoyed the meta-level "controversy" kept popping up in my For You. Some of the "best" Twitter threads are just a dozen tweets and screencaps of random books/papers tied together to make some coherent point that you can nonetheless swipe through in 5 minutes. Is such content plagiarism? I find the notion itself absurd; this isn't academic writing. If the content is primarily sourced from one particular author they should obviously be linked to (which he was), but even this is a selfish desire: if their writing is interesting, I'll want to follow them! I read Cremiuex's thread, and I skimmed the blog post he allegedly "plagiarized", and I prefer Cremieux's rendition, but it doesn't matter, because if Cremiuex hadn't tweeted it, I wouldn't have ever found the source blog, or author.
Trawling the web and packaging up good ideas other people have had into a format that is easily digestible (and visible!) is a public service in my book. Cremiuex's thread was around 1000 words, the first blog post linked was around 3000 (maybe 2700 if we exclude some of the tables/formatting?) Even if Cremiuex didn't verify any of the figures or include anything from the other sources he linked (I am not invested enough in this saga to check) this degree of editorializing is sufficient to evade the label of "low effort content theft" in my opinion.
"This guy on Twitter plagiarized 1/3 of my blogpost then linked to my site, woe is me." Come on.
meduka
5mo ago
I suppose I'm just not enough of a lawcuck to understand why this is being blown up into such an ordeal. The guy is an El Salvadorean citizen, was not in the US legally, and could have been deported to any country (besides El Salvadore) and then, from there, immediately deported again to El Salvadore and this would have somehow been fine. But because some braindead or politically captured bureaucrat rubber stamped his paperwork where he claimed he'd be in danger if he returned to his own country they granted a targeted stay of deportation, which precipitated this entire clusterfuck.
The guy was married to a US citizen ("Jennifer Vasquez Sura", okay...) who had filed a restraining order against him. Not exactly Elite Human Capital. The wife had two children from a previous relationship who are "disabled". Garcia's own child is also "disabled". This context is supposed to engender some kind of sympathy, I suspect, but as someone who actually interacts with people of this socioeconomic strata I am more inclined to believe they were scamming government benefits, and the wife's current PR blitz is a consequence of her smelling blood in the water chasing a fat legal payout.
I will freely concede that it would be alarming if the Trump administration deployed this "strategy" to consign innocent American citizens to a third world gulag without legal recourse or due process, but I don't think Trump is "based" enough to do that. (No, the off-hand comment he made to Bukele about sending "homegrowns" does not count, as it was clearly about -- legally -- sending convicted criminals to serve out their sentences more cheaply than can be done domestically.) This attempt to force the executive to (presumably, temporarily) return one particular illegal comes across as political theater and legal chicanery. Frankly I'm hoping Trump makes a show of retrieving Garcia on Air Force One, landing in the US for a photo op, then clasping him in chains and loading him back on the plane, to dump him in Argentina or somewhere else -- from where he'll be repatriated straight into El Salvadore's "black site prison", hopefully for life.
There was some minor procedural error, therefore we must make an elaborate show of correcting it, at great expense, to achieve an outcome that will immediately collapse back to the current status quo. This is your brain on legalism.
meduka
6mo ago
Would you care to make a prediction as to when (or if!) LLMs will be able to reliably clear games on the level of Pokemon (i.e., 2D & turn-based)? If a new model could do it, would you believe that represents an improvement that isn't just "marginal"? Assume it has to generalize beyond specific games, so it should be able to clear early Final Fantasy titles (and similar) too, to cover the case where "Pokemon ability" becomes a benchmark and gets Goodharted.
Personally I don't think it's necessarily impossible for current models with the proper tooling and prompting but it might be prohibitively difficult; the problem is a little underspecified, for it to be "fair" I'd want to limit the LLMs to screenshot input and tools/extensions it writes itself (for navigation, parsing data, long-term planning, memory, etc). I don't think 2 years is too optimistic though.
(I asked 3.7 Sonnet, which suggested "3-5 years", o3-mini which said "a decade or more", DeepSeek said "5-7 years", and Grok "5-10 years".)
meduka
6mo ago
I played a few.
- Orbi Universo II: It looks like Democracy but mixed with Civ/GSG. The content in the demo was a little thin and the systems don't seem particularly balanced, tooltips could use some work, etc, but I'll probably pick up the full game when it comes out as it has some interesting ideas.
- Icaria: Factorio + programmable bots (with a visual scripting language). I understand this is a bit of a mini-genre of its own; I haven't played any of the other program-bots-to-build-a-base games. It seemed neat, but the demo ends just as it starts to get interesting. Could be a solid foundation, so long as there's enough content in the full version.
- Machine Mind: A cute little top-down tower defense-esque driving/shooter with customizable vehicles and resource gathering. My only concern is there doesn't seem to be much depth (no production chains) and it's rather casual. Full game will probably include a series of campaign missions that trickle out new mechanics which is not a design I'm fond of, unless it's handled well e.g., Thronefall.
- Private Military Manager: Tactical Auto Battler: I wanted to like this more than I did. I really want someone to make "Football Manager" but with PMCs. This is not that, it seems very linear and the decisions you make are not interesting.
- Outworld Station: I liked this a lot. Factory automation in space. Demo sadly ends just as it gets interesting, teasing you with ship-based logistics and then not letting you play with them. It's polished -- zipping around your starbase feels smooth, hauling asteroids back to your forge is satisfying, though combat is a little too trivial but could be fleshed out with e.g., more enemies, allied ships, co-op, etc. If they deliver on the mid & endgame then it could be really good.
- Terminal One: Airport simulator, like Airport CEO & SimAirport (both reasonably fun). Terminal One is 3D, which sets it apart. Except it's utterly unfinished, so much so I couldn't evaluate it; I placed my stands in the wrong direction and couldn't figure out how to delete them (maybe I was still in tutorial mode?); would not recommend the demo, but I like management games so I will probably give it another shot when it's out in EA.
Not part of Next Fest, but recent releases I've played:
- Microtopia: Factorio-like with robot ants. The neat thing is it inverts the logistics system you typically see in factory games: everything is delivered by your ants. Your ants also expire, so your base has to be designed around your queen, who poops out larva -- then you hatch them into workers and send them along a defined "trail" with branches based on logical conditions. They grow old and die, but you can merge and evolve your basic workers into more advanced types which resets their lifespan. So your entire factory ends up being a big loop, which is very satisfying to set up and watch in motion. It's cool, though the progression continually unlocks new features that have major implications for your base design, so if you need everything to be 'perfect' at every stage you're going to be in for a rough time.
- Heart of the Machine: I like Arcen Games, they're always interesting but tend to fall down on the execution. HotM is more polished than I was expecting (thanks Hooded Horse?) but it's probably best described as a sort of visual-novel with a strategy/management wrapper: the actual mechanics are not challenging and mostly serve as a framework to deliver the writing, but the writing itself and some of the story branches are decent enough. I'm also a fan of the nested game loop structure where "new game+" is actually a coherent part of the design and the content updates have been hitting nonstop since the EA launch. Hard to evaluate. I wish there were more systems and less narrative but I like the theme and the concept.
meduka
7mo ago
I mean one of his tweets being quoted in news stories was literally “I would not mind at all if Gaza and Israel were both wiped off the face of the Earth.”
Pessimistically I assumed this was the one that really got him fired (well, "asked to resign".)
Maybe if the left called him an antisemite rather than leading with the anti-Indian racism the charges would've stuck? Though maybe not, the Elon nazi salute thing passed from the zeitgeist very quickly, especially with the ADL coming out to defend him.
meduka
9mo ago
My biggest regret with my last build (besides choosing Intel and getting a sure-to-die 13900k which I'm loathe to RMA which would strand me without a desktop for 2+ weeks) was opting for air cooling. I've used Noctua fans for a decade now, they've always done well by me. The first (and last) time I tried an AiO water cooler was in 2013 and it was so terrible I wrote them all off entirely. Pump headers weren't standard on motherboards back then and I remember absolutely hating Corsair's software that was mandatory to keep running 24/7. (I'm not sure if I could've "just" pegged CPU_FAN1 to 100% in the BIOS and things would've worked; probably should've tested it.)
Anyway, despite having a case designed for maximum airflow and a huge Noctua fan, my 13900k will thermal throttle instantly if I run any benchmarks. Though it's fine in daily use and I haven't seen it exceed 80 while gaming, it is the principle of the matter... also the Noctua fan is so large I need to take the whole assembly off if I want to swap out RAM or m2s, which is rather annoying. Apparently AiOs have gotten better lately, so I'll give them another shot for my next build. Or just go whole hog with a custom loop, it doesn't look that hard.
There are definitely going to be massive blind spots with the current architecture. The strawberry thing always felt a little hollow to me though as it's clearly an artifact of the tokenizer (i.e., GPT doesn't see "strawberry", it sees "[302, 1618, 19772]", the tokenization of "st" + "raw" + "berry"). If you explicitly break the string down into individual tokens and ask it, it doesn't have any difficulty (unless it reassembles the string and parses it as three tokens again, which it will sometimes do unless you instruct otherwise.)
Likewise with ARC-AGI, comparing o3 performance to human evaluators is a little unkind to the robot, because while humans get these nice pictures, o3 is fed a JSON array of numbers, similar to this. While I agree the visually formatted problem is trivial for humans, if you gave humans the problems in the same format I think you'd see their success rate plummet (and if you enforced the same constraints e.g., no drawing it out, all your "thinking" has to be done in text form, etc, then I suspect even much weaker models like o1 would be competitive with humans.)
I agree that any AI that can't complete these tasks is obviously not "true" AGI. (And it goes without saying that even if an AI could score 100% on ARC it wouldn't prove that it is AGI, either.) The only metric that really matters in the end is whether a model is capable of recursive self-improvement and expanding its own capabilities autonomously. If you crack that nut then everything else is within reach. Is it plausible that an AI could score 0% on ARC and yet be capable of designing, architecting, training, and running a model that achieves 100%? I think it's definitely a possibility, and that's where the fun(?) really begins. All I want to know is how far we are from that.
Edit: Looks like o3 wasn't ingesting raw JSON. I was under the impression that it was because of this tweet from roon (OpenAI employee), but scrolling through my "For You" page randomly surfaced the actual prompt used. Which, to be fair, is still quite far from how a human perceives it, especially once tokenized. But not quite as bad as I made it look originally!
meduka
9mo ago
the position that this proves that all Arab immigration to Europe should be cut off, because even the apparently liberal/assimilated ones are still ticking time bombs of potential violence
Well, wignats will have a field day. (Maybe that was his objective?)
Certainly the more moderate right will find it uncomfortable. An educated, apostate Arab who's vehemently against the Islamification of Europe -- well, if Germany is anything like the USA, he'd be held up as one of the "good ones" and a solid ally. (The moderate right is of course desperate to latch onto any token PoC so they can assert that they're not racist!)
I don't think you can quite square this circle without accepting an ethnonationalist framing, so I expect this to be swept under the rug. It looks bad for Arabs, obviously bad for the pro-immigration left, bad for the moderate right; the only people who can point to this incident as confirming their priors are the ones saying these immigrants are fundamentally incompatible with Western civilization by virtue of their ethnicity, regardless of their professed views.
Assuming this wasn't some 4D double layered false flag: https://x.com/stillgray/status/1870306075695546383 (this reads like premium copium to me, but, I guess it's not impossible.)
meduka
10mo ago
I ended up just brute forcing via log bots, but I did read an interesting post on the subreddit that suggested a "main river" architecture (compared to the typical "main bus"): all spoilables go on a giant belt that ends with a bunch of heat towers where they're promptly incinerated for power. You pull from the river, process the material, feed the results back onto the river. The result is that all your spoilables are always fresh, the "river" never stops flowing, and you avoid any awkward clogs. Viewing Gleba as, basically, a flow system vs. the stock system you see on the other planets seems like it'd greatly simplify logistics. Personally I didn't build a single heat tower until Aquilo which is an obvious missed opportunity in retrospect.
meduka
10mo ago
With Harris shaping up to receive 10-15 million fewer votes than Biden in 2020, has anyone here updated with regards to the chance of fraud in 2020? This graph is floating around Twitter: https://x.com/zerohedge/status/1854144250562429081
Is Harris truly so unpopular that 15 million Democrats just stayed home? Or was the overwhelming shift to mail-in ballots in 2020 the key to Biden's victory, giving barely motivated Dems an opportunity to fill in a ballot and mail it out rather than drag themselves to the polls? Republicans in several swing states managed to pass legislation tightening voting laws -- maybe that helped? Or is America simply not ready for a woman president (much less a "black" woman president)?
meduka
11mo ago
So, I was in a pretty similar position to you, though my procrastination and absenteeism started in middle school and never really cleared up. My last year of high school I was going to class less than once a week to hand in assignments or write tests and spending the rest of my time at home reading, programming, or playing video games. The school tolerated this because I got an exemption from a psychiatrist (who I was forced to start seeing after I said, basically, "If I have to waste another fucking year of my life in that place I'm going to end up killing myself," to my mother when explaining why I kept skipping class.)
The psychiatrist diagnosed me with "social anxiety" which I didn't agree with at the time, and still don't, but I played along because it at least meant being able to graduate on time.
I moved out at 18 to go to college. Predictably, it did not go very well. While the coursework was trivial (freshman CS) the profs were hardasses about attendance, so I dropped out after two semesters. Moved back in with my mom to her great disappointment and did odd jobs to make my student loan payments and help with rent. (Picking apples in autumn and a part-time gig at the butcher's shop she worked at for the rest of the year, mostly. I didn't have trouble with showing up to these jobs for some reason.)
If discipline (conscientiousness?) is a Real Thing then I'd wager I'm <1st percentile. Whatever standard script typically engenders "work ethic" in people was completely ineffective on me; the only thing that's motivated me to do things I actively don't want to do is an overwhelming sense of panic and imminent fear of disaster. This is a pretty severe character flaw, there's no sugar-coating it, and I haven't been able to overcome it except in brief spurts. I've tried a friend's Vyvanse prescription: it seemed to make it easier to initiate annoying tasks, but I wasn't in school or working full-time, so I have no idea if it would've been effective in those scenarios, and it may have been placebo to begin with.
I'm 29 now, and unfortunately I never found a satisfactory solution to this problem; I gave up on the "standard" normie wageslave life a long time ago. Post-COVID, online courses and remote work might be a viable means to cope with the issue -- when I was in college, the idea of an online class being remotely equivalent to a "real" one was pure fantasy. You can, apparently, even get student loans for them now. I found my own personal success (such as it is) in other ways: I made a decent chunk of money in crypto early on (ironically using some of the student loan money which I'm still making the minimum payments on -- the interest rate is so low, it's the only Rational(tm) choice) and multiplied it with some Wallstreetbets-tier investments. The COVID years were particularly kind, and Nvidia secured the bag, so to speak. This was enough to live independently and comfortably (though not lavishly) for the foreseeable future.
This came out as more of a blackpilling post then I expected; I don't know that I have any advice per se, as "get lucky with crypto 8+ years ago" is not exactly actionable. That said, if you aren't able to solve the root of your issue (as I wasn't) it's worth considering more unconventional coping methods, e.g., finding some way to make enough money to achieve your goal of living independently. You are clearly very literate, fluent in English, intelligent, and an American citizen: this combination alone puts you ahead of a lot of people, and there has never been a better time to make a living on the internet. If your morals are at all flexible and you have a bit of risk tolerance (and if the alternative is "getting kicked out and eventually living on the streets", well...) there are many roads available.
Probably not the answer you were looking for, but (personally) the NEET life is great, if you can find a way to make it happen.
My response to this doesn't quite rise to the level of a gigachad, but it's close. I am very much not pro "CHOP/CHAZ"-adjacent behavior, and if it ever reaches that point, with youth gangs permanently occupying large public spaces and accosting anyone who passes near, that is obviously bad. But I think groups of youths having a place they feel is "their spot" and yelling at anyone who comes too close to "fuck off" is just teens doing dumb teen shit, and that's what this tends to look like in practice. My experience is they're also likely to try to bum some smokes or get you to buy them some booze. Antisocial, sure, but I wouldn't say it's turning the neighborhood into a ghetto. I think this kind of low-level dysfunction is pretty close to unavoidable in urban areas without extreme measures.
More options
Context Copy link