[...] It was an interesting review until he got to the part where he was talking about the disadvantages, and I was curious why adoption was associated with centralization. It will be disappointing if transaction throughput can only be scaled in a centralized manner, raising the issue of intermediaries in the Bitcoin Network.
Some nodes are better connected than others since they exchange coins with more users (i.e. an exchange is dealing with more counterparties than you and me for example), however that does not necessarily make them intermediaries in the classical sense. That is, even major nodes can't block specific transactions (as neither origin nor recipient are known to them) like, say, a bank could. I'm not sure how robust the network
currently would be if multiple major nodes were to go offline, but overall robustness should increase as nodes join the network and routing algorithms gets improved.
The concept of watchtowers is often critized (for example in the video you linked), but they are not necessary for using the lightning network. They do shine a light on a bigger issue with lightning network though, which is ease of use compared to on-chain transactions (i.e. watchtowers are mostly meant to help with that). Nonetheless I do believe usability will get better in time, just as is the case with any software.
About the alternatives: On-chain approaches like sharding or the good old big block are unfortunately largely prone to centralization and stand to offer a lesser increase in transaction throughput than lightning network does. Opinions on this matter widely differ of course, but that's a different can of worms.
