Why do we still use Ethernet MAC addresses?

I'm sitting here trying to think of reasons we really need MAC addresses and the ARP and NDP protocols.

Wouldn't it work just as well for each NIC to have only one address (the IP address), and for the switches to follow the same rules they do for MAC addresses, but for IP addresses? Put the source IP of a packet into the CAM table, age it out if it's not heard in a while, flood broadcast packets, and unknown unicast, etc. etc..? It seems like there's opportunity here to eliminate some overhead and complexity.

I understand the historical progression from Thicknet, to hubs, to bridges, and finally to switches, and I understand that back in the day there were many competing L3 protocols, not just IP, so in that context it makes sense how Ethernet came to be. But besides just historic and network-effect reasons, is there any technical advantage to having the two separate addressing schemes in modern networks?