I frequently say what you say, but there is a more charitable reading: today's AIs are not "minds", as they are not stateful. Maybe continual learning ones (like the mini AGI exhibited today on the HN front page) will be perceived as "soulful".

I think the working space in a given chat or session still counts as mind enough, but your argument is different, in that it argues about specific architectures rather than simply stating "software therefore cannot be conscious".