I like that test. If the AI couldn't find a measuring spoon it would need to grab any spoon it could and just "eyeball it". Also, if there wasn't an actual mug then maybe a glass will work (but not a pint glass) and certainly not a plastic cup. When delivered it would have to know to say "couldn't find a mug so i grabbed a glass". there's other things too, can't find regular coffee but it found some instant coffee? The AI would need to decide if that will work or should it ask first. All of those things are petty easy for a human.