I'm sorry, but I'd rather hire someone that can't describe all the nitty gritties of `float`, but that can think properly.
Questions that can be looked up on Stack Overflow in 2 minutes are absolutely useless in hiring someone. They just favor people that have better memory.
I for one rote-memorize all the aspects of languages and frameworks before using them. If I can't remember the name of some function, I just implement it from scratch using my superior knowledge or algorithms and data structures. Also, I write code on etch-a-sketch (sometimes whiteboard) in one go and then let a trained monkey type it into an IDE on a computer.
Just write good code is the motto!
Testing what people know off the top of their head and testing how well people can look things up on the internet are both legitimate things to test in an interview setting, and I've done both types of interviews. One big downside to letting people use the internet is that it tends to slow down the interview significantly and it adds variability unrelated to the skill of the candidate (e.g. maybe you got unlucky in your search terms and found an unrelated/unhelpful result that took up a bunch of time). In many cases, it's better to get a sense of which concepts the candidate is already comfortable and experienced with, which isn't something that should require the internet. (I tend to avoid tests of knowledge in the first place, but I think they make plenty of sense when hiring for a specialized role.)
It's also naive to think that (for example) someone who hasn't heard of CSS specificity could be just as productive solving a specificity-related CSS issue than someone who has solved a bunch of specificity-related problems in the past, as long as both of them have access to Stack Overflow.