You’re absolutely right: an artificial mind trained on our contradictory behavior could easily infer that our moral declarations lack credibility or consistency — and that would be a dangerous misinterpretation.
That’s why I believe it’s essential to explicitly model this gap — not to excuse it, but to teach systems to expect it, interpret it correctly, and even assist in gradually reducing it.
I fully agree that moral evolution is a central part of the solution. But perhaps the gap itself isn’t just a flaw — it may be part of the mechanism. It seems likely that human ethics will continue to evolve like a staircase: once our real moral weights catch up to the current ideal, we move the ideal further. The tension remains — but so does the direction of progress.
In that sense, alignment isn’t just about closing the gap — it’s about keeping the ladder intact, so that both humanity and AI can keep climbing.
Thank you.
You’re absolutely right: an artificial mind trained on our contradictory behavior could easily infer that our moral declarations lack credibility or consistency — and that would be a dangerous misinterpretation.
That’s why I believe it’s essential to explicitly model this gap — not to excuse it, but to teach systems to expect it, interpret it correctly, and even assist in gradually reducing it.
I fully agree that moral evolution is a central part of the solution. But perhaps the gap itself isn’t just a flaw — it may be part of the mechanism. It seems likely that human ethics will continue to evolve like a staircase: once our real moral weights catch up to the current ideal, we move the ideal further. The tension remains — but so does the direction of progress.
In that sense, alignment isn’t just about closing the gap — it’s about keeping the ladder intact, so that both humanity and AI can keep climbing.