His point seemed to be that many technologists like us have a strong tendency to believe they know what is ethical despite their narrow experience in the world.
I think the important thing is to admit that we can't know everything, and to err on the side of caution.
When asking whether it's ethical to have a relationship with a synthetic person, the answer depends on the answers to many other questions.
Is the synthetic person simply a robot following a basic program, or is it capable of learning, adapting, and making decisions that its programmers didn't anticipate?
If a synthetic person can learn and adapt, then at what point should it be considered sentient?
If a sentient synthetic person appears to consent to a relationship, how can we be certain that it understands what it's consenting to, or that it even understands the concept of consent?
I'm not sure we can answer all the questions that need to be answered, so I'm not sure we can say with any degree of certainty that it is ethical for a human to have a relationship with a synthetic person. Since we can't be reasonably certain that it is ethical, we must assume that it isn't ethical until we achieve certainty beyond a reasonable doubt, because to assume otherwise might cause harm.
I think this is similar in some ways to how we treat relationships between an adult and a person below the age of consent. It doesn't matter how strongly a child affirms that they want to have a sexual relationship with an adult; it simply is not possible for a child to give consent because it's not possible to establish with certainty that the child actually understands the nature of the relationship or the consequences it could have. So the only ethical course of action is never to allow sexual relationships between adults and children.