The question is not whether machines will think. It is whether the people directing them will.
There is a version of the technological future that is genuinely good, and it is not utopian. It looks like this: computation absorbs the repetitive cognitive labour that currently consumes most of a working life, and the released attention goes into judgement, care, craft, teaching and repair — the activities that machines are worst at and that human beings find most meaningful. Models help us see coupled systems we cannot hold in one mind: ecological monitoring at planetary resolution, materials discovery for the energy transition, early detection in medicine, translation that removes a barrier between traditions.
None of that requires a technological miracle. It requires only that the released capacity be spent on something worth doing, which is not a technical question at all.
The failure mode is equally clear, and it is the default rather than the exception: released capacity is immediately reinvested in more throughput. More content, more consumption, more persuasion, more speed, and a workforce required to keep pace with an amplifier. On this path nothing about the human situation improves; it merely intensifies.