I am a habitual watcher of media in languages I don't know at all, via subtitles. If you do this as a habit, you get used to taking in the subtitles without making a conscious effort to focus on them and read them, and it becomes lot less distracting from the action than you would think.

exorcise the demon possessing the baby

what

The scope is insane and there is zero chance it would ever make its budget back. Somebody should give me loads of money so I can make it

There are 5 endings deciding the fate of humanity, in the spirit of Deus Ex, though they have to do with which faction the player aligns themselves with during the story rather than a single choice at the very end. It all revolves around humanity's relationship with technologically superior aliens

Can I have Deus Ex

I have the beginnings of a lorebook and design doc for a spiritual successor sitting around on my hard drive that I revisit now and then. The player character is forced into cybernetic augmentation, makes hard choices about their allegiance, and travels to the moon

Heck yeah :D

:O I would love to! My favorite city in Kanto is Celadon, and it has my favorite gym too. Have you done Celadon yet? Can you please draw an Oddish somewhere in the area near Erika's gym??

In the case of LLMs, i.e. generative text models, that math equation you're optimizing is one that essentially remembers and can recite every bit of text in the training data, but does so in just the right lossy way that it can also give useful outputs for novel inputs that weren't seen in training.

But LLMs as they exist right now are not "AI" by any meaningful or widely understood definition. It's a reappropriated buzzword for branding & for confounding laypeople. What we call "learning" would be better understood as optimizing, tuning the dials on a long but conceptually simple math equation

But generative models are still probably best understood by laypeople as a very special JPEG compression algorithm. It turns input data into model weights during training, then outputs a JPEG-ified amalgamated version of that input data during inference. (...Or sometimes just verbatim.)

This is not to undersell how useful this can be. It turns out that if you're very clever and very careful, a generative ML model can be trained in just the right way that you kept the general patterns of the input and lost everything else, and you get LLMs or stable diffusion models.

Generative ML models are fundamentally a kind of lossy compression of their input, like JPEGs are lossy compression of images. This is why even before we had models reproducing whole passages of novels or news articles, we already had standards in ML about getting permission for your training data.

This is bad and disingenuous framing that is encouraged by ML companies trying really really hard to make you think LLMs are "AI" with minds that learn like humans do. They are not. The better analogy is that training LLMs on a corpus is like saving an image with a high level of JPEG compression.

Yes!!!!!

What prize did I win

More Gen 1 resprites

Unless my memory of the book fails me, this "loser guy" was the expert scientist who, while methodically studying astrophage cells, was the first one to discover how they reproduce and to explain their lifecycle and why they transited between Venus/the Sun. Doesn't seem like an apt description.

I'm waiting until I can see the movie in the comfort of my own home so I can't comment on the adaptation. And I don't agree even a little bit with Weir, claiming his work isn't political. But I take umbrage with describing the book as "Loser guy gets drafted into service, excels, saves the world"

And because he isn't a "loser guy", but a competent scientist and teacher and generalist who participated in all the same astronaut training as the other crew members, he's able to carry out his mission with the help of a similarly excellent engineer.

Then he has a moment of weakness when thrust suddenly and unexpectedly into the role of crew member, never being told and never realizing he was the tertiary backup.

He is initially frustrated after he wakes up and realizes. Then he moves past it and stays focused on the mission as always.

I haven't seen the movie. I enjoyed the book and have read it a couple of times. I'm not on board with calling Ryland a "loser guy". He's an expert in the relevant field of exobiology who enters and stays with the project enthusiastically, and he makes significant contributions...

Looks like Pewter to me

There is a very serious and worryingly competent propaganda push going on in recent months to inflate fears about so-called "AI" (i.e. chatbots). I do not like it and I do not trust it. It is revolting how it masquerades as educational/essay content while being so ready to lie and misrepresent.

Screenshot of the "Pindex" YouTube channel page, showing a list of recently published videos. Past a point, they're all about "AI" doom.

This channel recently pivoted hard into chatbot doomerism content.

More like the aforementioned thing. This video is a chatbot doomer propaganda segment which gives far too much credulity to marketing bullshit, riding in on some clickbaitly-titled but ultimately shallow commentary on US politics like a Trojan horse. This time with the involvement of Stephen Fry.

I think folks like this need to ask what their SBMM systems are really optimizing for. Is it really that players *in general* are only happy when they get close wins? Or is it that you've engineered a system that oppresses and pushes out everybody else, and you got stuck with the Gamers (TM)?

(To add: My passion for Smash Bros was ignited by a friend who was way better at it than me, and wouldn't take it easy on me. I played against him all the time, got stomped again and again, got better and better until I was finally his equal...and he insisted we banned Metaknight. Shoutout to Zach.)

I think folks like this need to ask what their SBMM systems are really optimizing for. Is it really that players *in general* are only happy when they get close wins? Or is it that you've engineered a system that oppresses and pushes out everybody else, and you got stuck with the Gamers (TM)?

The whole setup of DRG's multiplayer encourages this! The devs built their famously positive and not-toxic community by building a multiplayer system that mixes everyone together, regardless of skill level. And it's worked *very* well for them. DRG is a smol indie game with more players than FO76.

Some of my favorite experiences with this game are me and my frequent greybeard co-op partner hazing a couple of enthusiastic greenbeards who joined our party by cranking up the difficulty and doing our best to carry them, and it being basically a coin flip whether we'll be able to succeed or not