@abreaction @failedLyndonLaRouchite @inthehands That is a much harder, more expensive problem than "improvements in AI" - the flaws in AI often boil down to "failure to correctly ground the concept in the real world".

The model hallucinates because without grounding that understands the concepts of "food" and "color" as subjective experiences, "blue" and "blueberry" are almost the same.

Robots *require* grounding to connect their actions to their task.