Making predictions that outlast you is easy, you can never be ashamed of being wrong. The most useful predictions can be checked within a lifetime, the others are useless.
Im in the middle of my 40s, so maybe I will see the first "steps" into that direction.
But even if I dont: The idea that robots are taking over, built by us, is just insane: If they reach a level of "self improvement and self repair/care", humans will be just finished - not in a "sense of bad will" like "kill-all-those-dumb-humans", but in a sense of: Their technosphere will last longer and is very much more likely to survive climate change than the thin & vulnerable bio sphere we humans need?
Though, maybe there is some type of equilibrium in the future with much less humans at a stable rate, really stable lake just enough kids that mankind doesnt grow,and a large set of robots assisting us. But this is too far "positive sci fi", Id say?
Life has 4.5B years. Predicting silicon would last more is doubling down on what I said: you went for 100 years to a roughly 5 billion years prediction.
Let me remind you that a single unlucky CME can fry 80% of comms satellites and electrical grid, effectively crippling the entire technosphere you mention in a matter of hours.
One small war that disrupts the silicon supply chain can set back the development of AI for a whole decade. One company (ASML) having issues could do that. It is a very fragile system.
Climate change is a deep concern mostly because it would cause unimaginable suffering and massive loss of population, but it's likely humanity would survive it. We survived worse: supervolcanos, ice ages, multi-decade droughts, etc. It would be preferable to avoid it though.
