In all honesty, llms are probably going to make all this entirely redundant.
As such semantic web was not a natural follower to what we had before, and not web 3.0.
As such semantic web was not a natural follower to what we had before, and not web 3.0.
> It would of course be possible to sic Chatty-Jeeps on the raw markup and have it extract all of this stuff automatically. But there are some good reasons why not. > > The first is that large language models (LLMs) routinely get stuff wrong. If you want bots to get it right, provide the metadata to ensure that they do. > > The second is that requiring an LLM to read the web is throughly disproportionate and exclusionary. Everyone parsing the web would need to be paying for pricy GPU time to parse out the meaning of the web. It would feel bizarre if "technological progress" meant that fat GPUs were required for computers to read web pages.
The second point is silly, because there is no reason for everyone to train their own LLMs on the raw web. You'd have a few companies or projects that handle the LLM training, and everyone else uses those LLMs.
I'm not a big fan of LLMs, and not even a big believer in their future, but I still think they have a much better chance of being useful for these types of tasks than the semantic web. Semantic web is a dead idea, people should really allow it to rest.
In 5 years resource price is likely negligible and accuracy is high enough that you just trust it.
What makes you think that I am not an expert btw?
It indeed seems like you appear to believ that what's written on the internet is true. So if someone writes that LLMs are not a contester to semantic web - then it might be true.
Could it be, that I merely challenge that author of the blog article and don't take his predictions for granted?
To think a turd such as JSON-LD can save the "SemWeb" (which doesn't really exist), and even add CSV as yet another RDF format to appease "JSON scientists" lol seems beyond absurd. Also, Facebook's Open Graph annotations in HTML meta-links are/were probably the most widespread (trivial) implementation of SemWeb. SemWeb isn't terrible but is entirely driven by TBL's long-standing enthusiasm for edge-labelled graph-like databases (predating even his WWW efforts eg [1]), plus academia's need for topics to produce papers on. It's a good thing to let it go in the last decade and re-focus on other/classic logic apps such as Prolog and SAT solvers.