How can you code know what data to expect without typed field? Your code size can grow up even 10x if you need to assume that any record can have any shape...
You could have more complex and user defined types in an ideal super-SQL, like "int or map:string->bool" etc., but you WANT types. They reduce complexity at all levels! You want them in you code, you want them in your data!
Sure, go ahead, just use MongoDB with no constraints on collections structure... You'll end up with 5x more code in you application and 5x more bugs.
SQL and the relational model help you think more, code less (and slower - but this is an advantage) and have less bugs with less effort! If you're lazy and prefer thinking deeper to "working harder", you'll always prefer properly typed SQL :P (Unless you're working with stuff that's naturally shaped as graphs or tensors.)