> With the advent of large language model-based artificial intelligence, semantic HTML is more important now than ever.
It's odd: I remember seeing an argument recently, though can't remember where exactly (perhaps [0]?), that LLMs make semantic HTML obsolete, because they "understand" the text anyway.
After all, humans didn't need html to be semantic in order to be able to read it — machines did. And if machines are approaching humans' capability to read texts, then doesn't this put the whole semantic html exercise into question?
Humans interpret text as rendered by the browser. That includes a lot of visual information and even text that only becomes visible in response to user interaction.
For an LLM (or an AI in general) to do what humans do, it would have to process the rendered output and interact with the site like we do (such as clicking on a dropdown to make the available options visible).
The same information represented as semantic HTML (i.e text) is far cheaper to process for an AI. I think cost is a key consideration in everything AI. If it's too expensive then it won't happen, even if it could theoretically be done.
It's odd: I remember seeing an argument recently, though can't remember where exactly (perhaps [0]?), that LLMs make semantic HTML obsolete, because they "understand" the text anyway.
After all, humans didn't need html to be semantic in order to be able to read it — machines did. And if machines are approaching humans' capability to read texts, then doesn't this put the whole semantic html exercise into question?
0 – https://shkspr.mobi/blog/2023/05/does-ai-mean-we-dont-need-t...