@doescience We have 3 types of Narrow AI used today.

Non-symbolic AI for Learning w/inductive reasoning supports Discovery Science

Symbolic AI for Understanding w/deductive reasoning support Hypothesis-based Science

Hybrid AI = Learning & Understanding. https://medium.com/@shawn.p.riley/a-basic-guide-to-artificial-intelligence-2020-edition-49cd32612814

@doescience if you are going to work with non-symbolic AI, symbolic AI, and hybrid AI for the scientific method or complex systems you also need to really understand the role of interoperability in AI and automation.
https://www.linkedin.com/pulse/understanding-how-increased-interoperability-enables-use-shawn-riley/

@doescience you might also be interested in how @ENERGY invented symbolic AI @PNNLab & Tech Transferred to Industry supports automation of scientific frameworks like Claim Evidence Reasoning (CER) & Argument Driven Inquiry (ADI) w/explainable results. https://www.linkedin.com/pulse/automating-argument-driven-inquiry-shareable-cognitive-shawn-riley/

@ScienceUnderSec you might be interested in reading this thread since DOE invented symbolic AI @PNNLab used to accelerate, augment, & automate hypothesis-based science and is used in DHS & NSA's IACD for sense-making and decision-making. No sense paying for what you already have.