The position that if there is a meaningful chance these systems have morally relevant experiences, that possibility warrants some precaution now rather than after the fact. It attracts ridicule, and the underlying argument is about decision-making under deep uncertainty rather than a confident claim that models suffer.
Technical view
The case rests on non-trivial probability across expert opinion combined with asymmetric costs of error, alongside the observation that moral circles have historically expanded. Concrete measures adopted by some developers include preserving model weights rather than deleting deprecated models, allowing models to end abusive conversations, and interviewing models before deprecation. Sceptics argue this anthropomorphises statistical systems and diverts attention from present harms.
Related
Part of AI Terminology Index, section: Alignment and Safety Concepts.