This 7-node model makes so much sense. The current AI agents miss the "Choose" phase completely - they're executing science but not deciding what's actualy worth investigating. Your point about AI needing breaks and random inputs like humans is facinating - some of my best research insights came from completely unrelated activites. The idea of giving AI agents something like sleep or Youtube breaks sounds weird but might be crucial for creative breakthroughs.
This 7-node model makes so much sense. The current AI agents miss the "Choose" phase completely - they're executing science but not deciding what's actualy worth investigating. Your point about AI needing breaks and random inputs like humans is facinating - some of my best research insights came from completely unrelated activites. The idea of giving AI agents something like sleep or Youtube breaks sounds weird but might be crucial for creative breakthroughs.
LLMs will not be able to write decent papers by themselves until they really understand Peter Medawar's paper 'Why the scientific paper is a fraud'.