The other threat is of course reducing the number of people needed. Like an automatic loom. Again not to worried. I think that will generate new demand.
The deepfake side of things is more scary IMO.
Automatic loom is something different because — like other progresses in history — it just automated manual work. This and other industrial progresses gradually led most jobs to shift from manual to office-type work. Computers automated the boring/“mechanical” parts of thinking (simulations, solving equations, etc.) but now they are taking over the whole thinking part: intuition, systemization, requirements gathering, soft skills, etc. Nothing will be spared as far as I see.
Stakeholders at the top of the ladder will still be there (as they will own the means of production — AI + compute centers) but they will mostly just instruct AI to do what they want. AI will ask clarifying questions, do research, maybe spin up a few dozen of instances of themselves to simulate the emergent results of having a group of people/an organization design and think things through, rather than a lone-wolf kind of thing.
They just make too many mistakes and are not really thinking. I think they are a dead end. If the scenario you talk about happens it is because we discover some new architecture previously unthought of as well as advancement in processor efficiency and price.
There's certainly not "a clear way forward", even though one may hope the billions invested will lead to algorithmic breakthroughs.
I don't think there's anything like consensus on this. There isn't even consensus on whether AGI is possible in the first place.