Neuroscience pioneers propose AI ethics framework

Neuroscience giants Jean-Pierre Changeux (famous for the nicotinic receptor model) and Morten L. Kringelbach have published a provocative paper on arXiv proposing a complete rethinking of AI ethics in life sciences. Their work contrasts the energy-intensive computation of current AI systems with the brain's efficient 'global neuronal workspace' architecture.

The researchers argue that AI ethics shouldn't be treated as a special case but should be grounded in how human brains naturally handle ethics through shared neural architectures. They propose that AI systems built on these principles would require governance structures focused on 'upbringing' rather than traditional regulatory restraint. The paper introduces the concept of 'reward' not as maximization but as a dynamic cycle of wanting, liking, and satiety - fundamentally different from how current AI systems operate. They outline specific institutional requirements for this future vision while acknowledging major open questions about implementation and governance.