abadidea
0xabad1dea@infosec.exchange
So Neuralink has an open technical challenge – which is fine – but what they want is comically absurd (1:200 lossless compression of a signal that looks like TV static, that runs very quickly, on an extremely low power chip) and they don't even promise any sort of specific reward for a working solution that only needs to break physics, disregard information theory, and summon six unicorns https://content.neuralink.com/compression-challenge/README.html
7
1
1
Just tell them an AI should predict the correct result instead of using the actual input.
0
0
0
@0xabad1dea On the bright side most likely if you can actually solve their challenge they will offer you a job on the basis of your impressive technical ability, then mistreat you until you quit
0
1
0
Sebastian Krzyszkowiak
dos@librem.one
@0xabad1dea Based on given requirements, a script that hashes the content, uploads it to some online content-addressable storage and then downloads it back on "decompression" would probably be unbeatable 😜
0
0
0
Pavel Machek
pavel
2
0
1
abadidea
0xabad1dea@infosec.exchange@pavel I have not misunderstood the challenge; it's to get someone to do millions of dollars of unheard-of-levels-of-brilliant engineering work for free and then submit it as a zip by email after no promises of any sort of compensation have been made
0
0
0
@0xabad1dea Pretty sure it would be easier to invent a time machine, take Elon back to 1940, and have Claude Shannon explain to his face how impossible this is.
0
0
0
Manawyrm | Sarah
manawyrm@chaos.social@pavel @0xabad1dea that might be -- but this is real world data with sensor noise around it -- which has to be reproduced bit perfectly after decompression.
saving actual random data (from noisy ADCs and the environment) is the "challenge" here. that's not going to happen 😹
0
0
1
followup: it’s been months and the leaderboard is still… zip 🤐
1
0
0
@0xabad1dea who would have guessed 🤣
0
0
0