3 ms·
Kind of ironic, given how small the sources are.
by laverya 6y ago
Kind of ironic, given how small the sources are.
- toxik 6y agoInteresting challenge Smallest demo that compressed to the largest video file
- makapuf 6y agoCat /dev/random?
- timClicks 6y agoSurely dd would be faster than that.
- CodesInChaos 6y ago/dev/random is super slow and even /dev/urandom is surprisingly slow. If you care about performance, something like round reduced AES or ChaCha would be a decent choice.
- IcePic 6y agoOpenBSD /dev/(u)random uses ChaCha20 and gives me a few hundred MB/s on recent hw.
- vanviegen 6y agoThat would actually be really easy to replicate in such a way that no human could tell the difference.
- toxik 6y agoSomewhat lacking in the aesthetic department
- mckirk 6y agoThat actually nicely illustrates the parallel between AI and compression. An 'almighty' AI would be able to reverse engineer the source code from what it sees and compress it back down to that size. But then it might determine that the best compression would be no transmission at all, and decide to eradicate any potential viewers. Those damn pesky almighty AIs.