Eight Questions About Minds, Brains, and AI

Summary

Echohive's film and reading companion examine eight questions linking everyday mental life with artificial intelligence, while separating evidence, hypotheses, and explanatory analogies. A March 2026 fMRI experiment with 32 participants found that context changes affected remembered temporal distance, supporting a memory mechanism but not a complete explanation of why time seems to accelerate with age. The discussion of endless feeds presents reward learning and frictionless continuation as useful parts of the account, while warning that a slot-machine analogy is not a diagnosis of doomscrolling. It also uses the brain's often-cited approximately 20-watt metabolic estimate to argue that efficiency comparisons with AI data centers require matched tasks, accuracy, speed, users, and energy boundaries. On learning, the companion distinguishes conversation context, retrieval, and changes to model weights. In one September 2026 preprint, naive sequential fine-tuning on 100 memorization tasks retained 1.2% on average, while combined preservation mechanisms reached 34.9%; another preprint found that continued pretraining improved older factual recall in five of six open-weight models under a different setup. The article treats dreaming and the “overfitted brain” idea as an analogy or hypothesis, not settled evidence. For AI-assisted physics, it explains that symbolic regression must predict unseen measurements and be independently reproduced: a July 2026 preprint found that guiding the numerical search performed best across 74 known equations and seven harder recovery tasks, but recovering known formulas is not discovering a new law. Two 2026 preprints address AI judgment: a 1,500-person study found that advice moved participants away from their initial leanings on average despite measured sycophancy, while a 1,590-person study found that warnings and demonstrations did not reduce a sycophantic chatbot's persuasiveness. The final section uses face recognition to illustrate tacit knowledge and closes with a practical method: ask a narrow question, compare mechanisms and alternatives, and verify observations, hypotheses, and analogies separately.