Semantic Interoperability and the Future of AI
Jan 02, 2021
This is super-interesting but also super-difficult, especially when compressed into four longish paragraphs as in this post. But here's the gist: "symbolic AI codes human knowledge in the form of networks of relationships between concepts accompanied by rules (models, ontologies)," while by contrast a neural network "induces algorithms to recognize visual, linguistic or other forms from large masses of data." In 2021 these are beginning to merge, but there is no consistency on the symbolic side; "there is currently no way to code linguistic meaning in a uniform and computable way." So Pierre Levy proposes Information Economy MetaLanguage (IEML) "to translate any ontology and to connect all categories" into expressions called USLs (Uniform Semantic Locators). What that means in practice is introduced in this post as the Intlekt Editor for creating these expressions in IEML.
Today: 1 Total: 105 [Direct link] [Share]
