Comments on: Giving Cloud Data Warehouses A Relational Knowledge Graph Overlay https://www.nextplatform.com/2023/07/11/giving-cloud-data-warehouses-a-relational-knowledge-graph-overlay/ In-depth coverage of high-end computing at large enterprises, supercomputing centers, hyperscale data centers, and public clouds. Wed, 19 Jul 2023 19:15:09 +0000 hourly 1 https://wordpress.org/?v=6.5.5 By: Hubert https://www.nextplatform.com/2023/07/11/giving-cloud-data-warehouses-a-relational-knowledge-graph-overlay/#comment-211020 Wed, 12 Jul 2023 01:12:18 +0000 https://www.nextplatform.com/?p=142641#comment-211020 That last statement is a bit intriguing to me (essentially): “Now language models can [build] semantic layers and knowledge graphs [using] data assets in one database”. National Lab folks seem (to me also) to want to do something similar with, for example, climate change sub-models (eg. an enhanced “Penman-Monteith” based on AI re-analysis of field data?). I’ve seen AI’s outputs in terms of stochastic auto-correlated next item predictions (time-series), but not in the more evovled form of AI-derived process-model equations, or semantic graphs (as suggested by the statement). Are such outputs actually a thing today, or is it meant that we should presume, from AI’s outputs, that such high-level representations exist implictly (within the weights of the trained network), but that their explicit forms are as yet to remain unknown because of the unscrutable nature of biomimetic data processing?

A couple of lit. refs. would help if anyone has those … (or a sample output from a software product that derives such representations).

]]>