3 ms·
You start with a semantic knowledge graph. Then you program agents that manipulate that graph. Think semantic web. Nothing is centralized. The challenge is the
by miguelrochefort 11y ago
You start with a semantic knowledge graph. Then you program agents that manipulate that graph. Think semantic web. Nothing is centralized.
The challenge is the UI. We need some kind of AI that understands the context and user's preference in order to dynamically generate the best interface, taking into account both data consumption and input.
- walterbell 11y agoIf the graph is not centralized, does that imply it is public? Or could segments of the graph be private to individuals, groups, companies, nations, etc? E.g. the industry currently has FreeBase/Wikidata (public) and Google/Facebook/Microsoft semantic graphs (private). Would client endpoints arbitrate between these non-intersecting graphs?
- miguelrochefort 11y agoIdeally, everything would be public. Not only does it make the design much simpler, but it also happens to fit my total transparency ideal (I'm not a big fan of privacy). I have not given any thought about the co-existence of public and private graph segments.
- walterbell 11y agoThere's the small problem of the public graph being less complete than the private graph, due to corporate investment in the private graph.
- miguelrochefort 11y agoI don't see a problem.
- walterbell 11y agoThe problem is that users will have less functionality on the public graph than a competing client which can access both the private and public graphs. Less functionality means less adoption means less resources to improve the public graph means less functionality ...
- miguelrochefort 11y agoI'm confident there won't be an incentive to introduce fragmentation. In any case, users will decide.