Yes, that setting (which does not need a transaction) means that all tables are read without shared locks. There are still schema stability locks, modifications still take locks. Read Uncommitted allows both dirty reads and duplicate reads and missing rows.

If you're going to do this, make sure that it's in a scenario where if the query results are wrong a fair portion of the time there will not be business consequences. Basically, make sure before doing this that the exact values returned aren't important as they can be wrong if there is *any* concurrent inserts or updates which increase the row size on the table, even if it's not rows that the query will affect.

In a test I did, under read uncommitted, 60% of queries which counted the rows in a table returned an incorrect row count.