ulyssessword
No bio...
User ID: 308
ulyssessword
9hr ago
Is your UI the same as mine? Can you tell if I visited the comment section in that second link in the screenshot?
ulyssessword
9hr ago
Feature request: Some way to differentiate between "X comments, and you've seen all of them" and "X comments, and you've seen none of them".
ulyssessword
9hr ago
What does that mean? I'm pretty picky, and often do look at Youtube or Instagram, see that there's nothing interesting there, and then close the tabs and go to bed or sit under a tree with a physical book. Maybe I'm a bit odd. If I'm feeling... stressed? I'm not sure what the state is... I'll refresh The Motte or something over and over for a while, and yeah that's dumb, I shouldn't do that. I should probably take a nap at that point.
I'm right there with you, but I don't think it's universal. I feel like I won the lottery of fascinations because the supposedly-addictive (and also useless) content that gets spread around is just boring to me, so I don't get sucked into those holes. It takes zero effort whatsoever on my part.
ulyssessword
10hr ago
...but calling this "screens" feels like calling slot machines "levers" or something.
I'd go much, much broader than that. It's like calling slot machines, toasters, forklifts, and home gyms "levers". Sure, they all have the same basic interface, but they're wildly different than each other in every way that matters.
"Screens" covers direct communication with IRL friends, pseudonymous (or real-name-but-it-doesn't-matter) social media like Twitter/Discord, longform content like ebooks/movies/TV/podcasts, shortform content like news articles/memes/alerts, official interactions with the government or other institutions, ads, games, work, etc.
Being on screens all day is probably worse than pulling levers all day, but it's still wildly underspecified. Even if they define it better in the actual podcast than your comment, it still feels like they're over-reaching with the label.
(unless they're talking about eyestrain and neck problems, but I have a feeling that never came up)
Anyway, is there anyone out there who has an actually useful way of discussing "screens," especially in respect to children, but also in general?
I'm keeping an eye out, but I haven't really seen it. One step better is people talking about "algorithms", and how there's a race to the bottom as genuine value loses out to virality.
ulyssessword
1d ago
Fake News You Can Trust lives up to its name once again. Their article correctly identifies that Mamdani condemned the right-wing victims, and that the perpetrators were left-aligned (I can't confirm "ISIS-inspired Muslims"). The other 95% is fake, of course, but they got the key points.
ulyssessword
2d ago
I doubt that most Iranians are ever in a position that a 1 hour phone call could guarantee the safety of Americans from a would-be terrorist attack.
I think it's a more accurate measurement of sentiment, even if it isn't practically useful.
Just turning the question around. If the information about a terrorist attack in Russia next week fell into your lap, how much time would you estimate it would take you to ensure that the right people got that that information, and how sure are you that your effort would actually prevent the terrorist attack?
Informing the Canadian authorities would probably take a couple hours up front, then a day or two in followups. I don't think they would be very invested in stopping it nowadays, so it probably wouldn't be prevented. Regardless, I did my duty.
Similarly, I'd hope that an American (even one who chants bad slogans) would inform the American authorities, and an Iranian would inform the Iranian authorities (assuming they suspect rogue actions instead of government ones). Those reports would have very different results, of course.
For the sake of argument, let's say a one-hour talk with 911 and police, after observing some strongly-suspected imminent terrorist preparations. That's about as small as you can go and still have it be a genuine inconvenience.
I think plain civic duty would get you to 75% (EDIT: among Western allies), with most of the remainder being indecision and passivity, not active hostility.
ulyssessword
2d ago
Not sure if I should seek out the Elite enemies on the map...?
You've gotta.
I haven't played 2 yet, but in StS1 (and Inscryption, Astrea, etc.) the long-term benefits of high-power fights are nearly required to beat the end bosses. If you fight two elites per Act, you'll reach the end boss with six (non-boss) relics. If you fight four, you'll have twelve. If you play it safe, then you'll be underpowered by the end, and almost certainly lose.
ulyssessword
3d ago
Same, and I'm also more optimistic about Claude than ChatGPT. Alignment and safety did turn into effectively usable capabilities.
ulyssessword
5d ago
you have no money (or negative money!) when it's time to contribute to the public good.
Just think of what you could do with billions of dollars. You could build an online shopping, shipping, book publishing, and cloud computing service that improves countless millions of lives. You could revolutionize communication and information flow or connect rural and remote people to the World Wide Web. Heck, you could just make a fun game.
Oh wait, those were all private billionaires' projects. The "public good" doesn't count unless it's filtered through the government first.
ulyssessword
8d ago
I had to reread that one so many times to figure out how good it was.
ulyssessword
9d ago
FYI, the wikipedia page doesn't contain the quote you're referencing.
ulyssessword
10d ago
Look - if Glock can't sell guns to the government while saying - you can't shoot black people because you have problems with racism, why should Anthropic be able to do so?
A toolmaker should have no say in how his tools are used once bought.
They can certainly offer to sell guns to the government under those terms, and the government can tell them to pound sand.
Similarly, Anthropic can offer to sell Claude without mass domestic surveillance or autonomous kill capacity, and the government can...agree, go back on their decision, and blacklist them from their entire supply chain. Apparently.
ulyssessword
10d ago
But The Associated Press Associates of Pressiness might be out to get you.
ulyssessword
10d ago
OpenAI just agreed to do what Anthropic would not do.
Source? The Altman tweet announcing it said that he (and the DoW!) agreed to do what Anthropic was punished for.
ulyssessword
10d ago
if they only have one drink a night
"One drink" is often standardized as 0.6 ounces of alcohol. There isn't any indication as to the quantity, but if it was a cup of vodka, then it would be about six standard drinks, which is certainly enough to count.
ulyssessword
10d ago
He hates Trump though and always encouraged people to vote against Trump?
He posts about 95% non-Trump content (by a broad definition, or 99% by a narrow one), so I'd still call it "rarely". And while we're posting 2016 articles, I'll highlight You are Still Crying Wolf.
He's certainly anti-Trump, but he's not a TDS-suffering obsessive.
EDIT: hat tip to ControlsFreak: This entire comment chain is about an anecdote from his past, not the main thrust of the paper. He described one somewhat-bad experience with GPT-5 Pro (presumably actually 5, not 5.1 or 5.2), but the rest was about 5.2. Ctrl+F "5" in the article to see that the mention is unique. It might be worth mentioning GPT 5.3 now (he mentions using Codex, so the restrictions don't completely lock him out), but even I think being three weeks behind the state of the art is fine.
I'm not sure about 14.6% vs. 31.3% pass rate on research-level math questions being a huge difference, but it's definitely noticeable.
Also, using a six-month-old model is better than usual for Science. If they had been 12 months behind 5.2 Pro (itself two months old by now) instead of four, then they would be dealing with a zero percent pass rate as o3 wouldn't have been released yet.
ulyssessword
16d ago
A downside is that the model is hard-wired into the chip, allegedly two months from model to production.
That's (almost) the difference between Claude Opus 4.1 and 4.6 (skipping over 4.5), or GPT 5.1 vs. GPT 5.3 (kind of, since it's a restricted release).
There's probably a niche for it, but it probably won't become a core piece of the landscape even discounting the cost.
ulyssessword
17d ago
Microsoft Excel is better than other spreadsheet software.
ulyssessword
20d ago
Yes, they do, that's why there's that silly loop.
That loop is still well-defined though. The example picture on Wikipedia shows the line spiralling around the pole, and ending near the same meridian as its starting point. A different line with the same length and starting point but a slightly shallower angle would end a bit northwest of there, and so on for the rest of the angles which would trace the loop.
If it reaches the pole after spiralling inwards for an infinite number of turns (but finite distance), then it's not well-defined. Otherwise it is.
These are different effects. One is the distortion caused by curvature, and the other is the fact that the surface is finite.
A 1% distortion is one effect, but a 200% distortion is another? What's the threshold between the two, because it sure looks continuous to me.
You get silly answers if you ask for great circle distances longer than the circumference, too.
It's simply not defined. Much like your definition is only defined for a subset of angles near the poles (which is why those points don't appear on the curve).
ulyssessword
20d ago
ulyssessword
20d ago
going 850 miles in a wide range of northerly directions isn't well-defined there.
That's okay, those points don't appear on the curve. From eyeballing it, East-North-East is about as close as you can get to North before that becomes an issue.
Pick a better-behaved set of numbers and the "rhumb ring" (which doesn't seem to be a standard term) will touch the great circle ring in both the North and South directions.
Nah. You pick a better-behaved function, and it won't be an issue. The same distortion that causes you to measure a 300 mile distance as a 302 mile arc causes you to measure a 250 mile distance as an 850 mile spiral.
ulyssessword
21d ago
see below. distance-direction vs. direction-distance.
Sorry, that was the wrong screenshot, and I'm also on desktop. Here's the actual screenshot (but mobile is the same AFAICT):
/images/17731314194047902.webp
More options
Context Copy link