SoylentNews is people
SoylentNews
A machine learning algorithm has created tiny (64×64 pixels) 32-frame videos based on text descriptions:
The researchers trained the algorithm on 10 types of scenes, including "playing golf on grass," and "kitesurfing on the sea," which it then roughly reproduced. Picture grainy VHS footage. Nevertheless, a simple classification algorithm correctly guessed the intended action among six choices about half the time. (Sailing and kitesurfing were often mistaken for each other.) What's more, the network could also generate videos for nonsensical actions, such as "sailing on snow," and "playing golf at swimming pool," the team reported this month at a meeting of the Association for the Advancement of Artificial Intelligence in New Orleans, Louisiana.
[...] Currently, the videos are only 32 frames long—lasting about 1 second—and the size of a U.S. postage stamp, 64 by 64 pixels. Anything larger reduces accuracy, says Yitong Li, a computer scientist at Duke University in Durham, North Carolina, and the paper's first author. Because people often appear as distorted figures, a next step, he says, is using human skeletal models to improve movement.
Tuytelaars also sees applications beyond Hollywood. Video generation could lead to better compression if a movie can be stored as nothing but a brief description. It could also generate training data for other machine learning algorithms. For example, realistic video clips might help autonomous cars prepare for dangerous situations they would not frequently encounter. And programs that deeply understand the visual world could spin off useful applications in everything from refereeing to surveillance. They could help a self-driving car predict where a motorbike will go, for example, or train a household robot to open a fridge, Pirsiavash says.
An AI-generated Hollywood blockbuster may still be beyond the horizon, but in the meantime, we finally know what "kitesurfing on grass" looks like.
(Score: 4, Touché) by requerdanos on Sunday February 25 2018, @02:03AM (2 children)
A pixel is not a measure of width, depth, length, or any other "size" measurement. It indicates that however large a given image--the size of a hydrogen atom, or the size of the known universe--that image has been divided into elements of usually equal size.
Thus, the videos are either the size of a hydrogen atom or smaller, the size of the known universe or larger, or somewhere in between. Thanks.
For a better comparison, they are about the size of an elephant: link to 64x64 picture of elephant [freworld.info]. Photo license CC BY-SA 4.0, attribution "Andrew Shiva"
They are about the size of a school bus: link to 64x64 picture of school bus [freworld.info]. Photo license Creative Commons Attribution-Share Alike 3.0 Unported, attribution "Marcin Szala"
They are about the size of the Eiffel Tower: link to 64x64 picture of Eiffel Tower [freworld.info]. Photo license Creative Commons Attribution-Share Alike 3.0 Unported, attribution "Julie Anne Workman".
They are about the size of--wait for it--The Library of Congress: link to 64x64 picture of LoC as seen from the North [freworld.info]. Photo license Creative Commons Attribution-Share Alike 3.0 Unported, attribution, some Wikipedia user named Hugaholic (I didn't ask).
Note: Because the films are 32 frames, you must look at any of the above photos about 32 times to best get the best feel for the "size".
(Score: 2) by VLM on Sunday February 25 2018, @02:35PM (1 child)
Its also worth pointing out no one would collect postage stamps if they were that cruddy looking. 200 to well over 600 dpi equivalent is more like it. I remember my father collected stamps for awhile and when I was a kid he pointed out a stamp tangentially showing the instrument panel of an aircraft and using a magnifying glass you could read the needles on the visible instruments, and I was impressed. Not sure if thats normal or unusual. Looking over his collection together they all looked pretty high res in general. It seems unusual to issue stamps that are intentionally ugly, also. In like the ancient boomer era, stamps were kind of a nation-level 4chan meme factory so if you wondered what meme the nation of Brunei found interesting, my dad probably had one of their stamps. Ebay was very hot for stamp trading around the turn of the century, he sold them all.
An even worse would be a comparison to coin currency. If you have really good vision or a camera macro lens or a triplet, the oldest wheat pennies have the artist's initials along the bottom border, the lincoln memorial obverse side (up till '09 or so) if you looked real close you could see lincoln's statue inside his memorial. IIRC you can count 34 individual kernels of wheat on the obverse of a wheat penny. In the old days before smart phones everywhere, if you were bored enough you could entertain yourself for awhile with a pocket of change and a magnifying glass or geologists triplet, although a guidebook to coin collecting helped.
I find the mint's behavior over the last decade or two to be pretty annoying and money grubbing. It was cool having the same currency for decades and you got to know its little details pretty well, now it seems every year we have all new paper and/or coinage. Perhaps trying to cover up massive nation-state level counterfeiting rings. "Oh you went to the expense to make indistinguishable copies of our $20? F you we'll release a new one seemingly every year just to F with you" You can't win that battle but you can put up a fight, perhaps that is what our mint is doing. It is kinda obvious, though, last century they messed with at most one individual major currency item per decade, lately they mess with every piece of currency every decade.
(Score: 2) by requerdanos on Sunday February 25 2018, @09:49PM
Makes me wonder, in an idle sort of way, what area of a postage stamp 64x64px would be able to reproduce faithfully. A mm2, maybe?
Every $20 ever issued is still valid currency. Don't counterfeit them, but if you do--counterfeit whichever one(s) of them you want. They're all valid.