AI Systems Learn Patterns, But Lack Human Understanding From Experience
• Generative AI systems like Dall-E and ChatGPT generate new data by learning patterns, unlike human minds which learn to predict sensations and use predictions to guide actions.
• Human minds learn patterns in a context of survival and action, allowing us to distinguish causation from correlation.
• Current AIs lack ability to actively intervene, so can't test and improve their world models like humans do through action.
• Humans also make internal predictions about physiological states like hunger to maintain bodily integrity.
• Humans benefit from culture and science to collectively improve our models; we also self-reflect on our own knowledge, unlike current AIs.