I think there is a step or two missing in between. Abusing dogs isn't just as bad as abusing something that suffers. It's, e.g., doing something that has no value to society (unlike animals being mistreated in a feed lot), or it's something that sews unacceptable levels of aggression. It feels much more complicated than you make it seem.
Hm? There is no objective test for whether I am a philosophical zombie or not.
Consciousness is subjective, so it's simply incoherent to ask whether something is conscious or a zombie.
I imagine thane once someone manage to produce an AI with enough agency to seem sentient it will far exceed human capabilities.
We have deep learning systems that perceive and strategise. Who’s to say that AlphaGo doesn’t have some experience of playing out millions of games?
Or talk to an ant.
There's no reason to expect to be able to understand something's ideas just because it's smarter than you.
Right now, we are in no position to argue it's wrong, because we have no accepted framework for understanding qualia.
That said, we can still make some simple assertions with confidence such as:
1: A one-line program which prints "I am afraid" does not experience fear qualia.
2: A human being which utters the words "I am afraid" may be experiencing fear qualia (and thus perhaps experiencing pain/suffering).
If you assert that the world is full of philosophical zombies rather than humans like yourself, you may not agree with assertion #2. But we build our ethical and legal frameworks around the idea that we aren't philosophical zombies, rather, that we have feelings and want to avoid suffering.
Once you start talking about more complicated programs, though, we don't know what the right assertions are. We don't have an understanding of what things can suffer and what things cannot, for starters. We generally agree that, say, dogs can suffer. That's why we have laws against animal abuse. But what about programs?
It is absolutely a real problem, because some day we will have to build laws and culture around how we treat AI. Whether or not an AI can suffer (ie. by experiencing qualia that can be considered undesirable) will likely affect how we treat that AI. If we have no way to answer "can this program suffer?", then we are setting ourselves up for regret.