Nobody is sure about the details about what a good "observer" is, but probably big (like a few millions atoms) and hot (like not cooled in liquid helium) is enough. The idea is that the system has too many internal states that are difficult to track, that it can be approximated like a classical system and hide the weird parts of the quantum effects and get a result that is like if the wave function has collapsed.
Nobody is sure, but check something-something-decoherence in a hundred years, because I guess the people working in that that field will solve your question with an explanation like the one I wrote in the previos paragraph. (If you are interested, search in google just "decoherence". I call it "something-something-decoherence" so it's super clear that there are still too many problems to solve and nobody is sure that it's the correct explanation.)