Let’s suppose for a moment that we live in a world where:
1. Human-grade artificial intelligences not only exist but actually are granted rights
2. That we follow Schweitzgebel and Garza’s argument for AI rights and follow their design principle.
Following Matthew Liao’s arguments in his paper “Selecting Children: The Ethics of Reproductive Genetic Engineering”, where he supports the Human Rights approach regarding the ethics of reproductive genetic engineering over the Libertarian approach or the Perfectionist approach,
If we apply the logic behind the ethics of reproductive genetic engineering to Human grade AI manufacturing, is there a moral duty to design them in a certain way that fulfills the fundamental conditions and capacities that Liao talks about?
How is reproduction parallel to robot manufacturing? Reproduction becomes analogous to manufacturing when reproductive genetic engineering comes into the equation. Even though there are physical differences in the way that AIs and humans come to life, (manufacturing process vs biological reproduction) the ability to either modify or select desired genetic traits makes biological reproduction more like a manufacturing process. With AIs, we are in full control of their design and manufacturing process, we conceive them from scratch. Their design, the programming behind it, the materials and form, their function…Whereas in reproduction we are not naturally able to “manufacture” or design our offspring, until recently. Before reproductive genetic engineering, reproduction was a question of pure chance (and, of course genetic traits). However with modern science and with the availability of these technologies that can help “engineer” or choose our offsprings traits, the line that used to divide human reproduction from robot/AI manufacturing becomes more blurred: we are able to manipulate and design humans just like we are able to design and manufacture robots from scratch.
If we are capable of designing and manufacturing a Human-grade AI that is psychologically continuous and that follows Schweitzgebel and Garza’s Design Principle, are we then not morally obliged to design them in a way that would ensure that they have all the fundamental capacities that we have?
Liao emphasizes that the Human Rights approach on the ethics of reproductive genetic engineering is the proper way to address reproductive genetic engineering. (Remember that we are granting the same rights as humans to Human-grade AIs)This approach claims 4 things:
- We should not deliberately create offspring who will not have all the fundamental capacities.
- Offspring that lacks some of the fundamental capacities is already created. It can be permissible to bring this offspring to term.
- Other things being equal, it is not permissible to cause an existing offspring to lack some of the fundamental capacities.
- If it possible to correct certain defects in the fundamental capacities of an offspring, and if it is not too demanding to do so, it can be impermissible not to do so.
Liao criticized the Liberalist approach to reproductive genetic engineering (AKA Liberalist eugenics) that states that “it is morally permissible to create any kind of offspring we like.” We are allowed to put any combination of genes that correspond with a particular conception of the good.” In his talk Liao said that the Libertarian approach “seems to permissive” because it would “allow the existence of any offspring whose life is not fulfilled under the fundamental capacities.”
However the Liberalist approach seems to be an adequate way to approach AI ethics: it seems to be morally permissible to create any type of (human-grade) AI that we like, as long as it corresponds with a particular conception of the good. It is certainly not considered morally wrong to create an AI that lacks some of the fundamental capacities, because they are created to fulfill a purpose that corresponds to a particular conception of the good. Usually, with manufacturing we tend to alienate ourselves from the product, we consider the manufactured product a “thing” and thus it has no rights. But when we have the possibility to create Human-grade AI (and in a world that they have rights) and you limit or prevent it from having the “fundamental capacities” that every right bearer has, aren’t you violating their rights?
This makes me wonder, why is it that the Liberalist approach to reproductive genetic engineering is not as “appropriate” as the Human rights approach, but a liberalist approach to human-grade AI manufacturing might be considered appropriate? Is there actually something “biological” attached to our rights?