This is a particular instance of the general approach, "I have to assign a number to each of these items, but it's hard and contentious to do, so instead I will give them all zeroes (objects) or ones (agents)." It always increases the total error. The world is not divided into agents and objects, and this approach would still increase total error, or at best leave it unchanged, even if they were, since errors in classification give a larger total error when they are thresholded instead of just left as, say, probabilities.
You should also consider that, when AI is developed, you will become an "object".
This approach doesn't work well even for humans. Very intelligent humans, armed with and experienced with mathematics, large computerized databases, regression analysis, probability and statistics, information theory, dimension reduction, data mining, machine learning, stability analysis, optimization techniques, and a good background in cog sci, biology, & physics, think more differently from average humans around 1600 AD, than average humans in 1600 AD did from dogs. So where do you draw the line?
Very intelligent humans, ... think more differently from average humans around 1600 AD, than average humans in 1600 AD did from dogs.
Reference? The priors for recent humans and dogs thinking more alike than modern humans and dogs (despite Euarchontoglires and Laurasiatheria diverging about 90 MYA and humans in 1600 diverging from modern humans 400 years ago). I might estimate at 90M:400 against, if I had to do so very quickly.
Why do you think that more than half of the change in thinking in the last 90 million years has occurred in the last 400?
When someone complains that utilitarianism1 leads to the dust speck paradox or the trolley-car problem, I tell them that's a feature, not a bug. I'm not ready to say that respecting the utility monster is also a feature of utilitarianism, but it is what most people everywhere have always done. A model that doesn't allow for utility monsters can't model human behavior, and certainly shouldn't provoke indignant responses from philosophers who keep right on respecting their own utility monsters.
The utility monster is a creature that is somehow more capable of experiencing pleasure (or positive utility) than all others combined. Most people consider sacrificing everyone else's small utilities for the benefits of this monster to be repugnant.
Let's suppose the utility monster is a utility monster because it has a more highly-developed brain capable of making finer discriminations, higher-level abstractions, and more associations than all the lesser minds around it. Does that make it less repugnant? (If so, I lose you here. I invite you to post a comment explaining why utility-monster-by-smartness is an exception.) Suppose we have one utility monster and one million others. Everything we do, we do for the one utility monster. Repugnant?
Multiply by nine billion. We now have nine billion utility monsters and 9x1015 others. Still repugnant?
Yet these same enlightened, democratic societies whose philosophers decry the utility monster give approximately zero weight to the well-being of non-humans. We might try not to drive a species extinct, but when contemplating a new hydroelectric dam, nobody adds up the disutility to all the squirrels in the valley to be flooded.
If you believe the utility monster is a problem with utilitarianism, how do you take into account the well-being of squirrels? How about ants? Worms? Bacteria? You've gone to 1015 others just with ants.2 Maybe 1020 with nematodes.
"But humans are different!" our anti-utilitarian complains. "They're so much more intelligent and emotionally complex than nematodes that it would be repugnant to wipe out all humans to save any number of nematodes."
Well, that's what a real utility monster looks like.
The same people who believe this then turn around and say there's a problem with utilitarianism because (when unpacked into a plausible real-life example) it might kill all the nematodes to save one human. Given their beliefs, they should complain about the opposite "problem": For a sufficient number of nematodes, an instantiation of utilitarianism might say not to kill all the nematodes to save one human.
1. I use the term in a very general way, meaning any action selection system that uses a utility function—which in practice means any rational, deterministic action selection system in which action preferences are well-ordered.
2. This recent attempt to estimate the number of different living beings of different kinds gives some numbers. The web has many pages claiming there are 1015 ants, but I haven't found a citation of any original source.