Understanding “longtermism”: Why this suddenly influential philosophy is toxic

It's not about an ethical obligation to create digital people, more a way to quantify what's at stake if humanity were to be destroyed. The point that he makes is that most people use 8B deaths as the "cost" when doing cost benefit analysis around a hypothetical apocalypse. Bostrom thinks the cost is much higher since it should also include all humans that will ever exist. When you try to quantify how large that could possibly be, you run into the possibility of far-future digital humans and he includes those in his calculation.

Digital, posthuman, galactic empire futures may or not happen. There may be serious practical bumps in the road. Digital existence seemed nice in some positive fiction works (like in Greg Egan's Permutation City) until I've read Lena ( https://qntm.org/lena ), and that assuming that it will be possible or desirable at all. The same goes with unseen possibilities that our current stage of development may not see as an idea, even our today's world with everyone connected with internet sharing their lives was something with little or no fiction covering it not so long into the past.

In the other hand, having no future not only erases those weird speculations that may not happens, but also the past, a tree falling in a desert forest doesn't make sound, after mankind is gone things will be as we never existed, all we did never happened.

I do find it interesting that most people don't object to the idea of thinking about what is best for your hypothetical grandchildren.

I agree, my objection to the serious (say, Bostrom's) version of the argument is exactly along that line. When we add new entities to our set of those worthy of moral consideration, we don't imply they should all be given equal value. In less abstract terms, it's perfectly consistent to hold the following two moral positions: "animals should not be mistreated gratuitously", and "it is morally permissible to eat meat"; this would amount to having an utility function where animals have some nonzero weight, but not enough weight to offset reasonable human desires.

Likewise, that averting existential risks should trump all other considerations would not only require that future humans have some nonzero moral value, but that their moral value ought to be high enough that everything else pales in comparison.

Something like "my concern with future humans decreases exponentially with time", like you propose, still leaves us with a "normal" position: "sure, we should care somewhat about the future of mankind, but not, like, above all else". Both the premise and the conclusion of this "boring" position seem much more reasonable to me.