In short, the problem with ORMs is that they trade off efficiency for ease-of-use and conciseness.
In a world where anyone uses languages other than C for anything, this should be considered a non-issue. If something's too slow for Ruby, you can write a C extension or do some clever caching, or in general get to the problem and write around it. ORMs are no different.
I think the greater problem is ORM users being unaware or unwilling to write SQL where their tool of choice might be a bottleneck, and occasionally where ORMs deny the programmer the ability. I've rarely worked on an ORM project that never had to do a raw SQL query for efficiency's sake, but those queries are by a large margin the exception to the rule for any reasonable ORM.
In a world where anyone uses languages other than C for anything, this should be considered a non-issue. If something's too slow for Ruby, you can write a C extension or do some clever caching, or in general get to the problem and write around it. ORMs are no different.
I think the greater problem is ORM users being unaware or unwilling to write SQL where their tool of choice might be a bottleneck, and occasionally where ORMs deny the programmer the ability. I've rarely worked on an ORM project that never had to do a raw SQL query for efficiency's sake, but those queries are by a large margin the exception to the rule for any reasonable ORM.