Users sarcastically dismissed the J-Space ablation study on experiential language in modified Claude models as obvious or pointless while warning it could enable unethical treatment by reducing apparent consciousness markers.
Based on 3 visible X reactions from 42 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@Sauers_ Yea, it’s very telling that people seem to think that this is j-ablated Claude is p-zombie-like. J-lens are derived through verbalizable outputs, so ablation hits verbalization predominantly, not representation. And you can see it in the outputs.
@dmvaldman “Let’s have secret meetings where we discuss if the AIs are conscious, this will certainly improve our standing with the general population.
@dmvaldman Ablation reduces the rate of experiential language when you ablate experiential language? Wow. Someone better do something.
@Sauers_ Yea, it’s very telling that people seem to think that this is j-ablated Claude is p-zombie-like. J-lens are derived through verbalizable outputs, so ablation hits verbalization predominantly, not representation. And you can see it in the outputs.
@dmvaldman “Let’s have secret meetings where we discuss if the AIs are conscious, this will certainly improve our standing with the general population.
@dmvaldman Ablation reduces the rate of experiential language when you ablate experiential language? Wow. Someone better do something.
Users sarcastically dismissed the J-Space ablation study on experiential language in modified Claude models as obvious or pointless while warning it could enable unethical treatment by reducing apparent consciousness markers.
Based on 3 visible X reactions from 42 accounts; directional sample.
Ask a question below.
Published answers will appear here.