Just close your eyes and swing as hard as you can! (I have not morticed anything since primary school)
Posts by Ada ๐ณ๏ธโโง๏ธ
I saw a company refer to FP16 as "full precision floating point" and FP8 as "half precision floating point" and I can't take this anymore
Is it Riitakerttu again ๐
And I mean, JK Rowling has publically shown her support for Riitakerttu, a doctor whose job it is supposedly to help trans kids. Riitakerttu has been very public about her hate towards trans children, and her weird focus on this sexual behavior of children.
It is important to understand how small Finland is: we have 2 trans polyclinics, one at TAYS and one in HUS, and their adjacent child trans clinics. I don't know why these are separate. In total, they employ maybe a dozen doctors. One of those doctors is Riitakerttu, our version of Rowling I guess.
The youth trans clinic at TAYS employes one of the worst transphobes I have ever heard of. The "real" trans polyclinic in the adult psychiatric section is in my opinion very chill, although severly underfunded.
๐ฆ BIG NEWS, Rustaceans!
Tickets are on sale and our speakers & schedule are live for #rustconf 2026!
๐
September 8โ11 | Montreal + Online
A thread on what to expect...
rustconf.com/register?utm...
it is time for the imposter syndrome, again
I have no idea if this makes sense btw, I don't really know english
but what about the food that I sauce?
olispa joku termi joka kuvaa tuollaista siviilisaation, tai kansan, tuhoamista, tai murhaa
Generating misinformation with AI is a major issue! And even if the AI water usage is blown out of context and misunderstood, datacenters have caused real issues to real people! And as an artist, he is directly affected by the widespread use of image models! He is right to be angry.
Few days ago I saw a close friend of mine post about how upset he was that people were wasting water to generate April Fools slop. I had a strong urge to go "umm actually" and explain how the water usage of AI is misunderstood. But at the same time he is right to be angry!
This is something that I have struggled with a lot, mostly because I haven't properly internalized that I am in fact an expert in this field myself. And I guess I am autistic too, which probably explains why I get so annoyed when people get angry at The Wrong Thing. But I am learning ๐
This just in: pollution causes higher ISO settings during the night. You might be eligible for exposure compensation.
A finger curls on a slightly less ominous monkey's paw: you now also have problems with Outlook.
I asked copilot to implement an algorithm I have been struggling with for 8 years and it just... did it. I feel extremely annoyed by this.
just wait until they hear how much water it takes to make a book
life is so beautiful
Far more vibrant picture of the sky, showing a beautiful aurora that alternates between green and red glow. The foreground shows a lot more trees, and also the ground, my fence and sheds. There is an off-scene light that is partially lighting some of the trees, and the picture has impressive, almost unreal dynamic range.
In comparison, this is what my phone saw. I am sure that is used some kind of "AI enhancement" to boost the aurora, since none of my captures have that reddish hue and I also did not see it by eye.
Faint aurora borealis in the night sky, and some stars, with few trees in the foreground. The aurora is making the sky glow green, and you can barely make out the wave aurora pattern. Everything is blurry and noisy and I domt know how to use my camera :(
Lost my tripod base plate(?) so the only lens I could use was the massive Sigma 80-400mm. Even at 80mm the fov was way to narrow to properly capture the aurora. I want to try again today but I have no idea how to set up my camera with the 8mm lens since I can't use the tripod.
A huge part of the AI hype is tech bros and CEOs experiencing genuine joy for the first time. Idk what that says about the world.
you really hammered that thick wood deep into those gaping holes huh
I am not woke enough for bluesky ๐
Being labeled an AI enshittification advancer on the same week I quit my job in the AI enshittification factory hurts a bit ๐ญ
time to update your threat model I guess
I am mostly just upset about Opus 4.6 feeling like a big downgrade from 4.5 I guess. Sorry I was more rambling about thoughts I've had lately, and kind of missed your original point.
so that they can claim their newest model is the best in the world again. I think this leads to a more kiki frontier, with sharp peaks in knowledge/skills related to specific benchmarks and degraded performance in other areas.
So companies claim models get better, while in practise they get worse.
It is very hard to get a complete picture of current capabilities of LLMs, mostly because we don't really have any good way of evaluationg general performance, or even how to define it. At the same time, the companies training these models can cherry-pick one or two benchmarks to train against,
I think it is mostly due to the jagged frontier thing. LLMs are great at some very specific tasks, okay at some, and horrible at others. CEOs and investors like to focus on the tasks LLMs are good at and extrapolate from there, and AI critics extrapolate from the other end of spectrum.