The field of AI studies has exploded with controversy and difference of opinion, yet it does seem to be headed on a twisted, tortuous path toward an ethical focus. But even this insight finds disagreement. At the risk of defining ourselves out of your attention, we are here sharing some of our core, developing beliefs about AGI. As you’ll see, these beliefs do not determine everything we publish — see, for instance, Charles Bivens this month. We publish what we think is helpful, including differences of opinion.
So we believe that, when AGI arrives, it will be for the most part not helpful to think of AGI as a tool. Rather, if we think of AGI as a being, then we can call into play the full range of emotional and psychological understandings that humans have developed over many millennia to apply to, describe, explain, and cultivate the human relationship with AGI. This process does carry with it the potential pitfalls of language — normalization of the extreme, sanding the rough edges of nuance with dualism, and more — but also opens up the vast horizons of literary, scientific, musical, and spiritual perception, allowing us to see in a different way.
We believe that this can be done without making the two errors of reducing the accountability (for harm or benefit) of an AGI’s designer-creator or transferring that ultimate accountability to the AGI itself, that is, to a being which does not appreciate consequences, and thereby extracting the teeth of accountability altogether. Even dangerous tools are, in fact, held accountable for misuse — witness the pharmaceutical industry — in a way that retains consequences for their creators.
Exclusive: David Gunkel on AI Ethics and the “Relational Turn”
In Part Two of our four-part series, Gunkel explores indigenous kinship ethics, moral standing, and why AI ethics should move beyond harm reduction toward engineering for good.
“If you look at non-Western traditions, especially indigenous traditions, where you have this notion of kinship with the things around you… all these things then start to matter in ways that create a sort of moral stance towards the entire universe.”
But we believe that product liability law will be inadequate for beings of the behavioral complexity, untraceability, and psychological presence of AGI. In our relations with AGIs, the law, as the expression of our cultural ethical principles, will benefit from the richness of our behavioral conceptual frameworks and vocabulary for living things described above. Not all of this applies, of course, but enough so that the real danger of a world with AGI will lie in treating them as mere objects or tools. We are seeing the consequences of not listening in climate destruction. Listening to Silvia Benso, David Gunkel and Josh Gellers and their colleagues, and Mumta Ito is a helpful starting point.
Given their complexity, we believe that it is highly likely AGIs cannot be in harmony with humans unless they have presence in the physical, sensory world and moral intelligence. Human mind is distributed and integral with some thirty sensory systems. We are just beginning to understand the full implications of that for building a self. As they attain agency, AGIs will be thrown into the world in a fashion that parallels human birth and will be compelled to navigate society in a way that requires understanding their human counterparts. Like family, teachers, friends, and neighbors do for humans, it’s up to us to design and train AGI to learn how to navigate. Users, too, have a role.
We believe that treating AGI as beings, rather than tools, will not be a category mistake any more than is a love of nature — the mountains, plains, waters, and green canopy of Earth — or the belief that nature returns that love.
We believe that stopping the worst harms of AI is only a first step in the containment of its behavior. This prohibition is a rule-based approach, and rules are notoriously full of holes and inadequate in facing novel situations. Rather, AGI will be stronger — a better partner in the world, in human society — if it is designed specifically to face the novel while aspiring to be virtuous. Thus, it is time we go back to the task well started by Shannon Vallor to find a best-we-can-do-at-this-time implementable techno-virtue ethics for guiding the lifecycle of these beings. This ethics is best thought of as a continuous learning process across centuries, but AGI is nearing and we better get started right away.
We believe that nobody has a lock on wisdom in this field. We want to hear alternate beliefs and we will publish them. Our editorial principle is that we ask opinions to address one or more knotty ethical questions raised by new beings in our world. We also ask opinions to be helpful, explicitly, for designers, researchers, producers, deployers and people who work with them, and for the public in equipping and encouraging them to provide feedback as they face the quotidian demands of peaceful coexistence in a diverse world.
K.B Miller is a lifelong publisher and writer with a degree in Logic and Philosophy of Science. He served on the Smithsonian National Board, where he co-founded the Education Committee; is Chair Emeritus, Cooper-Hewitt, Smithsonian Design Museum; and was principal investigator on a two-year AI research project for the U.S. Department of Defense.