Your idea of baking some logic that makes your node open some channels to with random nodes with few channels is not really a good one, let me tell you why.
As a node operator, you have 2 incentives to locking up your funds in payment channels:
1- So that you can save in fees while making your own transactions
2- So that you can earn some fees while routing payments
The only way of maximizing both things above is to connect to nodes with plenty of connections. By connecting to a couple of central hubs right now your ability to reach most of the network; your "coverage" if you will, is maximized. Also you are highly incentivized to connect to central hubs since that is more likely to increase your routing volume, in case you are capable of routing.
You have to also consider that opening a channel with a distant and lonely node will more likely just make you spend more money when making payments, since you'll probably just have to perform more "hops" in order to reach your destination when using them.
All in all, I think there's no way around it, hubs will arise. But I don't think that's a problem because the barriers to become a hub are really not that high. It's basically just get some capital and have (or hire) the technical know-how. Now compare that to the banking world where you have to spend probably millons of dollars and be specially well politically connected in order to open your own bank.
Also the thing that you say about KYC and AML stuff doesn't quite apply because AFAIK payment routes are using the onion protocol, so even though your payment passes by a central hub, it will probably not be able to learn much about it as you'd think.
I agree. There Are Now 3 Million Data Centers in the U.S., and Climbing. Low barrier to entry. Centralisation will occur but is balanced based on efficiency. All could run lightning or EOS. opportunity cost.