Still, there are some ways to eliminate the escrows like implementing smart contracts but those are much more complex and highly limited when it comes to Bitcoin's network so the escrow business may last for a while but not forever. Surely there will be an advancement in the field where we can trade Fiat/BTC and vice versa in a trustless manner.
I agree with stompix here. In addition, if your bank offers a smart contract on some blockchain to buy BTC (for example using wrapped Bitcoin), they could in reality offer you BTC directly. Or you don't need the smart contract at all. I remember a bank which had Ripple gateways. But then the fiat-to-crypto part is already solved, as XRP/BTC can be done with an atomic swap or on an escrow/P2P exchange without any risk (other than an 51% attack or so) because all transfers can be proven.
CEX are here to stay and in this age where people want to solve everything with two taps it's DEX who are at risk.
I agree somewhat regarding CEX. For a certain public they're very useful if they're responsibly and securely managed. I think this business model also will stay, But on the other hand I can well imagine that the DEX/escrow model can grow too. I think we're moving into a scenario where some big CEXes will stay and continue to grow, but many small CEXes will be replaced by DEXes, because they offer better terms regarding KYC (small exchanges always are prone to be hacked, and your KYC data is then in danger to be sold to criminals), and a complete exit scam is also more likely.
So in the CEX scenario, you have your bank, the exchange, and the transfer between them.
In a DEX scenario you have the bank, the other guy, the transactions between you and him.
You still haven't gotten rid of the bank or PayPal or Wu or whatever, all you have avoided is not sending your id to a CEX but instead your name and bank account to a complete stranger on the internet.
I can see where you're coming from. And I agree that there's always a middleman if you use a bank account or centralized wallet service, and if not, then all you can do is a person-to-person trade currently (a trade sending cash through a postal service seems even more risky).
But on platforms like Bisq you can see how old the accounts are (and as they tie account names to a bank account hash, old accounts are extremely unlikely to have used stolen bank accounts), and thus you can manage the risk a bit.
Moreover, while you have to send your name and bank account number to the counterparty, CEX verification takes currently much more data (at least photos of yourself, more likely videos), which in Deepfake times can be very dangerous if they get in the wrong hands. Names are public everywhere, bank account data in the wrong hands can be a risk but much less than a photo/name/phone number combination, which is what criminals will get if they hack a CEX.
The DEX proposal I linked to above was proposing to divide buy and sell orders in very small parts which are filled by different buyers and sellers to make escrowers unnecessary, so malicious parties lose more (reputation) than what they're able to scam, use LN for the Bitcoin side to be able to process micropayments, and a bank API for the other side (which is something not all banks will offer, and has additional risks). It's not a completely bad idea, but I guess this model could also be exploited to gather personal data, and of more people than with the traditional DEX model.