Imagine it's 1900, and two people are watching a car chug down the road.
The first person says, "These new machines are revolutionary! They'll usher in a new age of mobility for all humankind, connecting people faster than ever!"
The second person says, "They're death machines! They'll crash into walls and people, destroy property and kill millions of people!"
Both people are right. With the benefit of history, it's obvious. The modern world relies on cars, and simultaneously, a million people a year die in car accidents. So it is with AI. We have doomers and accelerationists and pro- and anti-AI opinions and... well, a lot of vocal people on one extreme or the other. (And incendiary positions get engagement.)
It's critical to consider both of these positions simultaneously. They're even mostly compatible, though the course of action each person recommends might be drastically divergent. Unintended consequences make this even harder to navigate, even as those may also be obvious in hindsight (look at Prohibition, or cane toads, or Wikipedia has these and many more in its unintended consequences article.) Economists are already discussing what happens if people's personal agents move everyone's savings to high-interest savings accounts automatically, and I'm sure you have a dozen examples in your own life for far less earth-shattering advancements.
Now think about it for a technology that is showing signs of being able to accelerate itself.
Imagine it's 1950, and people are still talking about cars like it's 1900.