Okay, so I get obsessed with concepts easily. I admit that now.
In 2021, I had the idea for an algorithm to ‘compress random data’ - and I spent about a year working on it. I literally built a super computer in my lounge room [for this algorithm.] I had additional ‘circuits’ installed in my house to run the super computer. The noise of starting the machine (which was in fact an old Dell m1000e chassis stuffed to the brim with high-core servers) was so loud it made me cringe and step outside the room. The fans sounded like a jet engine. The exhaust was so forceful it was enough to violently shake my blinds from meters away.
Later on, my boss let me use thousands of dollars of compute power to test out my ideas (thanks for the startup credits Google!) I did all of this to try answer the question I had become obsessed with: ‘can you compress random data?’ I wasn’t successful, and I’d be very surprised if anyone else will be too. There’s a lot of problems with trying to compress random data. If you have a small quantity of information then the chances of finding some kind of function that produces the data that’s smaller than the output is lower because you have less room to manoeuvre.
On the other hand: with more data the number of possibilities for a function goes up exponentially. Every single bit of data literally doubles how many numbers can be stored. It’s also not like maths is of much use here. Maths is about order and making sense out of chaos. Randomness is like the opposite to maths. For this reason I thought the solution likely relied on brute force but never in my results did I see anything that resembled compression. So either my approach was flawed or the idea is just so improbable that almost no one is going to find a solution.
In any case: I do think that a general-purpose compressor is impossible. I would also advise others who might be interested in random data compression to choose to work on something that is a little more practical. There just isn’t enough room here to manoeuvre here, and whatever ‘success’ might look like is likely to be very disappointing and almost certainly impractical.