Fediverse is now integrated with Threads, and Threads is bigger than Twitter. With any luck, Fediverse becomes the open source, Threads kills X, and news media continues to die the same slow death it would have anyway.
If you haven't noticed, the kids don't read, they like to watch.
Related story for people who haven't heard about it: At one point CERN had problems which seemed to be related to the phase of the moon. After much head scratching, it turned out that the problem -- with an underground particle accelerator -- was indeed related to the moon phase, since the moon's gravitational pull was deforming the path of the particle beam.
I'm guessing it's more to do with the interaction of the gravitational vectors of sun and moon. During a half-moon it is pulling at a right angle to the sun, during a new-moon with the sun, and during a full-moon against the sun.
Probably the same cause of tide supercycles - when moon and sun line up, the pull is additive. When they oppose, the moon detracts from the sun’s pull by a bit.
Writing that out I’m not sure it makes sense, but someone smarter will be along to correct me shortly.
When the moon is full it is opposite the sun and reduces the effect of the sun’s gravity. When it is new it is in the same direction as the sun and increases the pull toward the sun.
I want to agree, except that I believe when people authenticate to a particular site and post something, they're only really concerned with readership by other people who authenticate to that same site. Like if I have a car problem and post about it on a car forum, the ability for anonymous users to read my post is of little concern because they're not nearly as likely to reply to me as someone who's already logged in. If my goal was to spread information more widely instead of soliciting in-band participation, then in that specific scenario yes of course I need to find some other way to do that, but I don't think that's a typical goal with sites that strongly favor authentication; the goal is to get likes, replies, and so forth.
Tiktok has stopped allowing me to view videos when not registered and logged in. I only view them when they are sent to me. I’m never going to register but they’ve made it less useful to the registered people who send me links.
I submitted this link yesterday, I hesitated on the title because the nethack reference would have been a good one but decided against it to avoid spoiling the bug, but as a result it didn't get traction. Too bad for HN readers that the title which worked here is the one with the spoiler! Still a fun story though, glad it made it to the front page one way or another :).
A few years ago I added these banners to my shell prompt. It shows like the MOTD when opening a terminal on a full moon, new moon, or Friday the 13th (these are the special dates used in Nethack). https://github.com/CGamesPlay/dotfiles/blob/master/files/.co...
Correct me if I'm wrong, but didn't a friend of mine fisted/fstd/Timo "solve" this game years ago and make a bot to raep the highscore?
Not saying that this isn't a cool project and information, but my understanding is that the game has bugs/"working as intended" which allow anyone to basically auto click to rack up thousands of points.
P.S Timo if you're reading this I love you with all my collection))) <3
I remember reading about a bot that completed a NetHack ascension speedrun by hacking the random number generator, but it's probably not the one you were thinking of:
I guess I'm not entirely clear on the background/context here. I gather that the tweet's author isn't a serious Nethack player himself, but he is trying to train a neural net to play the game, and his training system somehow takes a model created by someone else as a baseline and tries to fine tune it somehow? But despite Nethack being based on randomly generated dungeons, the other model gets a consistent score every time, somehow? But even though the reference system reliably gets the same score through all the randomization of the dungeon, the game's full moon mechanic somehow throws it off significantly.
I feel like I mostly understand most of the pieces of this story when taken individually, but I'm having trouble assembling most of them into a coherent whole.
Their model is trained only on runs it has seen before which didn't include sufficient full moon runs. So its performance degraded when it encountered a sufficiently novel variant.
Which is where the I part of AI always falls down, input that sufficiently differs.
E.g., train facial recognition on a corpus of predominately white American faces, African Americans suffer a horribly high false positive rate when the cops use your model on surveillance footage.
I don't know exactly about this case, but when training models like this I tend to fix a small set, or even just one, seeds to use as a baseline 'quality measure' -- while this has the risk of over-tuning, always measuring quality using random seeds means you can misjudge a model's quality because you get particularly lucky, or unlucky, seeds.
However (and again I've hit this), sometimes you don't fix everything enough, and still have some unexpected variation, like in this case.
The score they're reporting is almost definitely the average over a set of fixed seeds I imagine. They just didn't realize that the seed is not sufficient to establish the play experience, the system clock is a factor too.
Readit News
I wish people would stop using twitter as the prime host for these kinds of things.
"Thanks, multi-part Twitter links are now held hostage behind a registration wall, making them unreadable and lost for most."
If you haven't noticed, the kids don't read, they like to watch.
Dead Comment
The story being that king tides were getting a connection wet near the shore and causing problems.
Writing that out I’m not sure it makes sense, but someone smarter will be along to correct me shortly.
Not saying that this isn't a cool project and information, but my understanding is that the game has bugs/"working as intended" which allow anyone to basically auto click to rack up thousands of points.
P.S Timo if you're reading this I love you with all my collection))) <3
https://pellsson.github.io/
I feel like I mostly understand most of the pieces of this story when taken individually, but I'm having trouble assembling most of them into a coherent whole.
Which is where the I part of AI always falls down, input that sufficiently differs.
E.g., train facial recognition on a corpus of predominately white American faces, African Americans suffer a horribly high false positive rate when the cops use your model on surveillance footage.
However (and again I've hit this), sometimes you don't fix everything enough, and still have some unexpected variation, like in this case.