Re: What's our Future in 2025? - Semantic Web: Past, Present, and Future

po 23. 12. 2024 v 21:47 odesílatel Ansgar Scherp <mail@ansgarscherp.net>
napsal:

> Dear Semantic Web community
>
> As the year ends, the question remains: How will the Semantic Web look
> in 2025?
>
> Together with Katja Hose, Maria-Esther Vidal, Gerd Groener, and Petr
> Škoda, we contributed this year an article titled "Semantic Web: Past,
> Present, and Future", see
> https://drops.dagstuhl.de/entities/document/10.4230/TGDK.2.1.3 to the
> new Diamond Open Access journal on Transactions on Graph Data and
> Knowledge (TGDK, https://tgdk.org/). This primer has been a living
> document for 13 years; see the link for details!
>
> As many of us cannot resist checking emails over the break, I ask you to
> consider the question: What is our future in 2025?
>
> Many "classical" Semantic Web researchers have witnessed the rise of
> Linked Open Data, Freebase being bought by Google (largely seen as a big
> win), and the beginning and success of the Knowledge Graph era. Many of
> us have moved on to or added topics like graph representation learning,
> graph neural networks, language models, etc., to our research portfolio.
> So, how much of the classical topics are left? How much will come back?
> Should we work more on knowledge graph embeddings obeying OWL axioms?
> Shall we have federated queries not only over multiple data sources but
> also include hybrid queries using similarities in graph embeddings?
>
> In the current 2024 version of the primer, we include the latest W3C
> standards developed by the Semantic Web community. But we also explain
> the journey from the famous Linked Data principles via Knowledge Graphs
> to the hot topic of machine learning on graphs!
>
> *The article linked above is an invitation to contribute. Contact me if
> you are interested!* It will be updated from time to time. We like to
> receive your feedback, and perhaps you would like to contribute to a
> future version.
>
> Best wishes and happy holidays,
>

2025 will be the year of Agentic AI.

Does the semantic web offer advantages in agentic AI?  Yes and no.  While
agents can parse structured data and unstructured data pretty well, does
RDF offer any significant advantages over JSON.  Probably not since agents
will be able to figure out what URIs are.  Is RDF cheaper in tokens, or
have fewer HALs, again probably not.

I think the areas of the sem web that will do well are those that embrace
the new agentic future.  Ultimately that is where most of the finding will
be, and so, those that can make their case will get new funding, as old
funding slowly decreases.  I think sem web in the enterprise will be much
less valuable as competition from giants will be able to do more with
less, and the moat is small.  Perhaps things like Agent safety and shields
will be a good area to focus on.

Sem Web is IMHO a good environment for data and agents.  I think it will be
important to remember that the important word in "Semantic Web" is "Web"
and that the important word in "Linked Data" is "Data".


>
> Ansgar
>
> PS: For 2025, there is a plan to update the German version of the
> article in an introductory textbook on artificial intelligence. It will
> add Shallow Graph Embeddings, Graph Neural Networks, and the interplay
> of Knowledge Graphs and Language Models. Once ready, I plan to ping this
> back to the English version.
>
>
>
>
>
>
>
>
>
>
>