Here's some discussion points to build on OP.
Eventually the rising cost of human labor will meet the lowering cost of machine labor. Once this point has been reached, we'll be forced to make some tough decisions.
What will we do with all these displaced unskilled manual laborers?
And when machines begin to replace skilled laborers such as mechanics, technicians, and mechanical engineers, what will we do with them?
As our AI becomes better at talking to people, eventually it should be able to (cheaply and effectively) hold small, pointed conversations with people in order to resolve queries presented in natural language. What do we do with all the displaced desk job workers?
At some later point I'd expect machines to be trusted enough in industrialized countries to begin collecting medical readings and samples to be fed to a central intelligence such as Watson. What do we do with the nurses and general practitioners that no longer provide the first line of care for medical problems?
On the theme of trusting machines, what happens when we entrust these(by this point in the thought experiment) incredibly trusted machines to raise and educate our children?
Let's steer this thought experiment in an impolite direction. Let's say everyone displaced from a job magically dies, either on the end of a bayonet, of old age after being sterilized, or from starvation. After the groups discussed above are phased out, who do we have left in the world?
The bankers, the government/military, the owners of corporate entities, and those with skills desired by the former(lawyers, entertainers, engineers, scientists) are now the remaining population that either cannot be phased out without creating a technological singularity, or would simply refuse to give up control.
In such a world, what would be a downside of machines taking control of remaining operations and neutralizing anyone who stood in their way?
This will be my singular post for this thread.