Bitcoin users are an even more privileged class of the already privileged population. They have regular access to a private computer, and robust internet access. That is not the case for most people in the world.
Lots of people in developing countries have access to cheap Android smart phones but no computer these days. Fortunately that's all you need to use Bitcoin.
Only with a thin client, defeating the purpose of Bitcoin. The Blockchain is currently over 18 GB, more storage than any cheap phone I've ever seen, and takes days to download.
Even Satoshi said it is likely most people will use lightweight clients. Even the original paper describes how to make a lightweight wallet that don't need the full history.
I think fees will get cheaper as times goes by and the bitcoin network and market matures but it is very hard to say right now with the current transaction size limitations.
Bitcoin is still in early development and very niche, so far the incentives have been working well enough but once the block rewards halves is yet to be seen if fees will be competitive enough and whether the price for them will go up. We also have to hope that the block size and the network capacity can grow fast enough to support its own weight.
In general I think that as more players come in the industry the fees of exchanges, atms, debit cards etc will decrease.
I wish some core developer could also contribute to answer your question and provide more insight than i can offer.
If the wild success of M-Pesa is any indication, then the traditional finance industry will reach the next billion consumers well before an obscure technology like Bitcoin will.
Given that M-PESA is run by a telephone company, is it the traditional financial industry or not? It can be argued both ways. SafariCom effectively runs a bank, but it is nonetheless primarily a technology company.
