My largest objection so far to Robin Hanson's The Age of Em is that ems are probably going to be more trouble than they're worth: Why emulate the whole brain, when so much of what the brain does, and needs, is completely superfluous from the point of view of the labor market? Most brain functions add no value at all to a worker's job; they're pure waste. What's worse, after you emulate a brain you will then have to persuade that brain to work for you, which entails both expense and some degree of ethical difficulty.
These would be groundless objections, of course, if no alternatives existed. But I think they do exist: Isn't there a good chance that machine learning will very soon give us everything that basic ems - like basic people - might be good at?
If so, then we won't need (or want) full-spectrum AIs, or ems, to serve as our lawyers or doctors. We'll want machines that perform those tasks alone, that never need to rest or relax, and that work for the cost of processor time alone, without all of the complications.