always and always and always and always in need

sports is really like is this 16 year old we abducted from his schooling the next coming of christ

hanif abdurraqib, there’s always this year: on basketball and ascension

re: prev, @thermocrying: they saw f1 overtaking (pun intended) hockey on ao3 so knew they needed to do something to boost engagement

i said what i said!

The conversation around AI is going to get away from us quickly because people lack the language to distinguish types of AI--and it's not their fault. Companies love to slap "AI" on anything they believe can pass for something "intelligent" a computer program is doing. And this muddies the waters when people want to talk about AI when the exact same word covers a wide umbrella and they themselves don't know how to qualify the distinctions within.

I'm a software engineer and not a data scientist, so I'm not exactly at the level of domain expert. But I work with data scientists, and I have at least rudimentary college-level knowledge of machine learning and linear algebra from my CS degree. So I want to give some quick guidance.

What is AI? And what is not AI?

So what's the difference between just a computer program, and an "AI" program? Computers can do a lot of smart things, and companies love the idea of calling anything that seems smart enough "AI", but industry-wise the question of "how smart" a program is has nothing to do with whether it is AI.

A regular, non-AI computer program is procedural, and rigidly defined. I could "program" traffic light behavior that essentially goes { if(light === green) { go(); } else { stop();} }. I've told it in simple and rigid terms what condition to check, and how to behave based on that check. (A better program would have a lot more to check for, like signs and road conditions and pedestrians in the street, and those things will still need to be spelled out.)

An AI traffic light behavior is generated by machine-learning, which simplistically is a huge cranking machine of linear algebra which you feed training data into and it "learns" from. By "learning" I mean it's developing a complex and opaque model of parameters to fit the training data (but not over-fit). In this case the training data probably includes thousands of videos of car behavior at traffic intersections. Through parameter tweaking and model adjustment, data scientists will turn this crank over and over adjusting it to create something which, in very opaque terms, has developed a model that will guess the right behavioral output for any future scenario.

A well-trained model would be fed a green light and know to go, and a red light and know to stop, and 'green but there's a kid in the road' and know to stop. A very very well-trained model can probably do this better than my program above, because it has the capacity to be more adaptive than my rigidly-defined thing if the rigidly-defined program is missing some considerations. But if the AI model makes a wrong choice, it is significantly harder to trace down why exactly it did that.

Because again, the reason it's making this decision may be very opaque. It's like engineering a very specific plinko machine which gets tweaked to be very good at taking a road input and giving the right output. But like if that plinko machine contained millions of pegs and none of them necessarily correlated to anything to do with the road. There's possibly no "if green, go, else stop" to look for. (Maybe there is, for traffic light specifically as that is intentionally very simplistic. But a model trained to recognize written numbers for example likely contains no parameters at all that you could map to ideas a human has like "look for a rigid line in the number". The parameters may be all, to humans, meaningless.)

So, that's basics. Here are some categories of things which get called AI:

There's plenty of software that follows a normal, procedural program defined rigidly, with no linear algebra model training, that companies would love to brand as "AI" because it sounds cool.

Something like motion detection/tracking might be sold as artificially intelligent. But under the covers that can be done as simply as "if some range of pixels changes color by a certain amount, flag as motion"

2. AI which IS genuinely AI, but is not the kind of AI everyone is talking about right now

"AI", by which I mean machine learning using linear algebra, is very good at being fed a lot of training data, and then coming up with an ability to go and categorize real information.

The AI technology that looks at cells and determines whether they're cancer or not, that is using this technology. OCR (Optical Character Recognition) is the technology that can take an image of hand-written text and transcribe it. Again, it's using linear algebra, so yes it's AI.

Many other such examples exist, and have been around for quite a good number of years. They share the genre of technology, which is machine learning models, but these are not the Large Language Model Generative AI that is all over the media. Criticizing these would be like criticizing airplanes when you're actually mad at military drones. It's the same "makes fly in the air" technology but their impact is very different.

3. The AI we ARE talking about. "Chat-gpt" type of Generative AI which uses LLMs ("Large Language Models")

If there was one word I wish people would know in all this, it's LLM (Large Language Model). This describes the KIND of machine learning model that Chat-GPT/midjourney/stablediffusion are fueled by. They're so extremely powerfully trained on human language that they can take an input of conversational language and create a predictive output that is human coherent. (I am less certain what additional technology fuels art-creation, specifically, but considering the AI art generation has risen hand-in-hand with the advent of powerful LLM, I'm at least confident in saying it is still corely LLM).

This technology isn't exactly brand new (predictive text has been using it, but more like the mostly innocent and much less successful older sibling of some celebrity, who no one really thinks about.) But the scale and power of LLM-based AI technology is what is new with Chat-GPT.

This is the generative AI, and even better, the large language model generative AI.

(Data scientists, feel free to add on or correct anything.)

anyway I'll never forgive arrival (2016) for taking a story written by a chinese american author and making the chinese military the villains in the movie when there's literally no reference to that plotline in the original work

there are no other asians in that movie, other than the Evil Chinese Commie Military Commanders

They've done it again. How do they keep doing this? If these examples are representative they've now done the same thing for the short video/clip landscape that Dall-e did for images a while back.

OpenAI are on another level, clearly, but it is also funny and sort of wearying how every new model they release is like an AI-critical guy's worst nightmare

in short actually what the fuck

I'm sure sometime in the last few months I must have said something like 'yeah they will probably solve the spatial and temporal coherence issues that hamstring every instance of AI rotoscoping sooner or later'. I didn't really expect it to be this soon though??

like you've got 3D rotations of complex objects. detailed multilayer forests and snow and shit like that with complex camera moves. cloth physics. hair physics. movement in and out of depth. complex multiplane camera rotations with water physics. photorealistic humans at all kinds of angles and poses. stylised characters. crowd scenes. pictures in pictures. i'm going over stuff that's complicated and time consuming for a human animator, I don't know what a diffusion/transformer model finds difficult.

but honestly what's really throwing me is like, the types of jank? yeah, here and there you get a little bit of classic AI jank like shapes morphing into other shapes, objects vanishing behind other objects, that kind of thing. a big tell is the failure of object permanence, so if something walks out of view, it will likely disappear entirely, and something else can appear out of nowhere. (though not always - it's improved here as well.) that's the sort of jank I expected.

but a lot of the examples, the jank involves weird levels of coherence, like the one with the basketball - the basketball generates a second basketball which clips through the hoop, sure, but how the hell is the AI's model of 3D space good enough to figure out that a basketball there would have to intersect with a hoop in the same location? sure, if you frame through the video, you can see that the hoop kinda dissolves through the ball... but at full speed it really looks like a 3D render with clipping.

further flaws I notice - the videos notably don't include very many examples of fast motion. a lot of them have a bit of a floaty, slow-mo feel to it. I suspect that has to do with how it propagates the motion through time. it also tends to have that kind of glossy advert-like quality to the photography which I've come to associate with AI - everything is perfectly studio lit. in general I'd say it's better at photorealism than stylisation - the stylised images (mostly furry 3DCG-looking characters) tend to look a bit creepy, with too-wide staring eyes.

anyway in classic OpenAI fashion, they post a lot of glossy pictures but they're kinda cagey about how the thing actually works. it splits images into spatiotemporal chunks and does ~transformer magic~ on them and apparently that just gives you a really detailed world sim. besides that, they used a recent technique where you get another AI to label the training data to feed into the training of their main AI - you'd think it would collapse in on itself but no that works apparently?

on some level I guess it makes sense. transformer models scale well with data and like, video gives you thousands of closely related pictures to train off of and discern underlying patterns.

but also it can just straight up render a convincing simulacrum of minecraft. minecraft!! though probably not in realtime lmao. the pages here don't mention how long it takes to generate a video and what sort of hardware you need to throw at it to get these kinds of results.

until I saw this I would have said that 'AI-generated animated film' is years away at least. like the 'Animate Anything' model from that team in China earlier this year was impressive but clearly limited to fairly specific scenarios. this one... this one you could probably make an entire AI-animated feature-length film and have it not look like shit. it wouldn't be as striking and intentional as a really good human animator, you'd have to throw out a lot of bad shots along the way and have a pretty specific vision of what you're aiming at. I suspect the first good AI films will be made by people who are experienced making films using other techniques, and the best uses of the tech will combine it with other techniques to get a 'best of multiple worlds' situation. but still...

I'm too tired to figure out what I feel about this. future's gonna be fucking weird.

sports is really like is this 16 year old we abducted from his schooling the next coming of christ

The Palestinian players each match have the number of days they’ve been getting attacked, bombed, and murdered written on their arms and today they had 110. Mohammed Saleh (in the photo), whose family live in Gaza, has been very open about the situation and talked about how he hasn’t heard from his family in weeks and he’s very concerned. While we’re all so proud of the national team for making the round of 16 for the first time ever, it’s bittersweet that it comes at a time like this where no one is able to fully enjoy it at all.

“In the war film, a soldier can hold his buddy—as long as his buddy is dying on the battlefield. In the western, Butch Cassidy can wash the Sundance Kid’s naked flesh—as long as it is wounded. In the boxing film, a trainer can rub the well-developed torso and sinewy back of his protege—as long as it is bruised. In the crime film, a mob lieutenant can embrace his boss like a lover—as long as he is riddled with bullets. 

Violence makes the homo-eroticism of many “male” genres invisible; it is a structural mechanism of plausible deniability.”

–Tarantino’s Incarnational Theology: Reservoir Dogs, Crucifixions, and Spectacular Violence. Kent L. Brintnall.

well this isnt setting a scary precedent 👁️👄👁️

u write some bad poetry in highschool that alludes to a miscarriage and its used to put u in prison u sing freely about politics and against genocide now its seen as a credible threat to the state now it can be brought up to a bunch of pierogi looking idiots who believe in charcoal smoothies and they get to weigh ur art as if it was as heavy as a gun i mean bro

The video camera is plugged into the VCR, allowing it to record itself being poked and prodded.

GIF’d version of Vide-Uhhh! (2005)

you are literally so so so right you’re the most correct ever

x

“The book that I did, the lectures, ‘Playing In The Dark’, was an examination of how Melville and Twain, and Hemingway and Flannery O’Connor, and all of these people got - brilliantly or not - alot of mileage out of wrestling with those questions about the presence of african-americans in this country.”

TONI MORRISON, “JAZZ” INTERVIEW (1993).

The Tiger Poem in Classical Maya!

Literal translation:

Transliteration:

Character Transliteration (ALL CAPS are characters that stand for full words, lower case are syllabic):