I do not think so. Most of my ipfs bookmarks from 2 years ago are offline now, everything gone because no peer has the content.

FileCoin just recently launched and it aims to fix this exact problem.

How does adding crypto fix this problem? I presume it's to incentivize peers into retaining and serving the content, but who's going to pay? The author? The viewer? How is either going to work for long tail content that's been abandoned by the author?

Wait, hmm... maybe if we use the blockchain here we can solve those problems too?

Or machine learning?

It doesn't fix the problem yet, it attempts to solve the lack of incentive to contribute to networks like ipfs/tor. All blockchain with reward based storage solutions I've tried so far don't meet expectations.

To this day, the Internet archive project is the most contributing solution to persistent Web information.

Any system that is that resistant to censorship will have to deal with it being used for files that the government most works to censor. That sort of relation will drive other people away and you'll find government increasing their power to be able to take it down (or at least identifying those who access it and take them down). Until society is willing to tolerate such material as the cost of a system that cannot be censored, the majority will continue to go for systems that give government the ability to censor what the people cannot tolerate, under the pinky promise that government won't use their powers for evil.

Where by “impossible” you meant “quite easy”, right? There isn’t some technical exception to following the law. IPFS can potentially help with the problem of someone dying or forgetting their hosting bill but it is subject to legal requirements like everything else and the hash mechanism makes both detection and compliance easy: if you host other people’s content, you’re either blocking certain values or getting fined/jailed on their behalf.