4 ms·
"Don't anthropomorphize token predictors" is a reasonable take assuming you have demonstrated that humans are not in fact just SOTA token predictors. But AFAIK
by fastball 1y ago
"Don't anthropomorphize token predictors" is a reasonable take assuming you have demonstrated that humans are not in fact just SOTA token predictors. But AFAIK that hasn't been demonstrated.
Until we have a much more sophisticated understanding of human intelligence and consciousness, any claim of "these aren't like us" is either premature or spurious.
- krackers 1y agoEvery time this discussion comes up, I'm reminded of this tongue-in-cheek paper. https://ai.vixra.org/pdf/2506.0065v1.pdf https://ai.vixra.org/pdf/2506.0065v1.pdf
- lostmsu 1y agoI expected to find the link to https://arxiv.org/abs/1703.10987 https://arxiv.org/abs/1703.10987 (which is much better imo)