There's a reason why W3C didn't just replace <b> and <i> with <strong> and <em>. A lot of people think they're just equivalent, but a lot of other people disagree. "Please display this in italic" arguably sends a different message from "Please emphasize this", and in some settings they may not be interchangeable.

The classic example is screen readers. A phrase that’s italicized because it’s important should probably be slightly louder, but a phrase that’s italicized because it’s foreign should probably be pronounced slightly slower. Those are entirely orthogonal things, and it’s useful to distinguish them.

Semantic HTML + CSS can be taken too far, but having separate layers gives you a lot more flexibility than presentational HTML alone, which tends to be very brittle.

If <i> means something different from <em>, you can keep them distinct while keeping everything "semantic". In fact, you could have a dozen different HTML elements for each use case of italics, e.g. <foreign> for foreign words and <t> for titles of books and movies, and it will be a mess, but it will still be "semantic". So the semantic-vs-presentational argument doesn't even apply here.

Meanwhile, the whole concept of using italics for emphasis is a Western idea. CJK languages, for example, don't use italics for emphasis, and most of their italic fonts actually look terrible on screen.

I don’t understand what you’re arguing for or against here.

CJK = Chinese Japanese Korean I believe. There is no concept of italics in those languages.

Yes, but if anything that seems to contradict the gist of the first paragraph of the comment.

Just to expand on kijin's reply...

In HTML5, <i> and <b> now have the deliberately vague semantic definitions of "alternate voice" and "stylistically offset" – this could cover emphasis, book/film names, key words, foreign phrases, etc. Incorrect use of the more specific semantic tags (eg. generating <em> when the user sets a film title to "italic") is worse than a lack of semantic detail.

HTML5 was designed with the understanding that not all web content can be authored by someone who knows the correct use of semantic HTML tags like <cite>, <em>, <strong>, <mark>, etc. Application-level copy can use those tags, but a rich text editor (for normal people) can't know which one is appropriate, so it should always default to the vague ones, <b> and <i>.

Why default to those? You dont justify that, you only justify the existence of a default.

The b element is back in HTML5.