I think this could have had better examples which illustrated better what the message was attempting to convey. The general concept I can largely agree with the simple manner it was described, however the ‘examples’ are basically demonstrative of the lack of cause-effect relationship. the ‘causation doesn’t always equal correlation’ type of argument. and I had the impression there was more to ‘the onion layer’ conception than the examples were able to illustrate or demonstrate further. When I test digital systems, LLMs, for their ability to reason in layers or ‘keep secrets’, I consider things similar to the general points of your shared hypothesis. However I likely would have focused on providing more “positive” examples and/or offering deeper clarity as to why the “negative” examples received those scores in specific in terms of failing to meet the defined criteria of the onion layer cohesion between layers. for my example, of the above examples, what does a car dealership that also sells illegal drugs have anything to actually do with private medical information? those seem like erroneously classified data points rather than a failure of the onion layers to have cohesion, in and of themselves. perhaps something this concept attempted to demonstrate is that data organization and categorization are well worth anyone to further contemplate or evaluate if their own internal (or external digital) system is accurately classifying and sorting through it’s own data points, or if disorganization is hidden beneath some obfuscation of another onion layer or multiple layers inwards/forwards.
Thank you for the share, very great for a brief 3 minute read. Lots here to contemplate and consider, I know a lot of these or similar ideas also frequent my own mind.
I think this could have had better examples which illustrated better what the message was attempting to convey. The general concept I can largely agree with the simple manner it was described, however the ‘examples’ are basically demonstrative of the lack of cause-effect relationship. the ‘causation doesn’t always equal correlation’ type of argument. and I had the impression there was more to ‘the onion layer’ conception than the examples were able to illustrate or demonstrate further. When I test digital systems, LLMs, for their ability to reason in layers or ‘keep secrets’, I consider things similar to the general points of your shared hypothesis. However I likely would have focused on providing more “positive” examples and/or offering deeper clarity as to why the “negative” examples received those scores in specific in terms of failing to meet the defined criteria of the onion layer cohesion between layers. for my example, of the above examples, what does a car dealership that also sells illegal drugs have anything to actually do with private medical information? those seem like erroneously classified data points rather than a failure of the onion layers to have cohesion, in and of themselves. perhaps something this concept attempted to demonstrate is that data organization and categorization are well worth anyone to further contemplate or evaluate if their own internal (or external digital) system is accurately classifying and sorting through it’s own data points, or if disorganization is hidden beneath some obfuscation of another onion layer or multiple layers inwards/forwards.
Thank you for the share, very great for a brief 3 minute read. Lots here to contemplate and consider, I know a lot of these or similar ideas also frequent my own mind.