I edited a transcript of a 1.5h conversation, and it was 10.000 words. That's roughly 1/8 of a book. Realising that a day of conversations can have more verbal content than half a book(!) seems to say something about how useless many conferences are, and how incredibly useful they could be.
In conversation, you can adjust your speech to your partner. Don't waste time explaining things they already know; but don't skip over things they don't. Find examples and metaphors relevant to their work. Etc.
When talking to a group of people, everyone is at a different place, so no matter how you talk, it will be suboptimal for some of them. Some will be bored, some will misunderstand parts. Some parts will be uninteresting for some of them.
7Bird Concept
Yup. I think it follows that when people do talk to groups in that way, they're not primarily aiming to communicate useful information in an optimal manner (if they were, there'd be better ways). They must be doing something else.
Plausible candidates are: building common knowledge and signalling of various kinds .
Just because talking to a group is not optimal for an average group member, it could still be close to optimal for some of them (the audience that the speaker had in mind when preparing the speech), and could maximize the total knowledge -- as a toy model, instead of one person getting 10 points of knowledge, there are ten people getting 5, 4, 3, 3, 2, 2, 2, 1, 1, 1 points of knowledge, each number smaller than 10 but their sum 24 is greater than 10; in other words, we are not optimizing for the individual's input, but for the speaker's output.
What it says on the tin.