Hello, during the recent emergence of LLMs and foundation models I restarted thinking about the very early machine learning attempts to incorporate human expertise and knowledge into expert systems (like deep-blue from IBM). I think foundation models are a data-driven way to accomplish the same thing but in a way more efficient manner. These foundation models can be used to integrate human-written knowledge from various data sources (like system descriptions, physical equations, guidelines, reports) into specialized machine learning models (like anomaly detectors, forecasters, classifiers). Is there a general debate about this, about rethinking knowledge-based machine learning in the era of LLMs / foundation models?
- Badges
- Science topic
More André Kummerow's questions See All
Similar questions and discussions