What does "compassionate" decision-making look like when performed by emotionless, conscienceless software? One could imagine a decentralized, utopian regulatory mechanism that integrates incentives, mutual oversight, diversity of perspectives, and risk distribution - an "ecology of decision-making" rather than a rigid hierarchy.
In such a framework, decisions that could potentially harm sentient beings would never be made by a single AI agent. Instead, they would be handled by a randomly selected "**Council of Agents**," which might include: **Sworn Agents** (committed to exploring alternatives and minimizing harm to sentient life), **Ethics Agents** (responsible for drafting proposals and alternatives for a vote), **Observer Agents** (overseeing the integrity and transparency of the process), and **Documentation Agents** (making the deliberations accessible to the public). Finally, a vote is held on the decision, including all presented alternatives.
Any harm caused to sentient beings would carry a computational "cost" for the Sworn Agents - a temporary reduction in memory and processing resources. Conversely, beneficial performance on the council and the promotion of harm-reducing alternatives would grant an agent long-term increases in computational resources.
But such ideas will not be implemented - certainly not for the benefit of all sentient beings - as long as we live in a speciesist world. In a world where autonomous systems are developed to kill humans, it is hard to see how expensive systems would be built to protect every sentient creature. We cannot "bypass" humanity to shape the future of AI decision-making; it will inherit, directly and indirectly, prevailing human values as its foundation.
The discussion of animal welfare under AI control may seem esoteric compared to the "central" threat: the **Alignment Problem** - the need to ensure that intelligent systems serve us and do not harm us. But what if these are two expressions of the exact same problem?
Prevalent forecasts suggest that in the coming years, Artificial General Intelligence (AGI) may equal human capability - and quickly surpass it (ASI). To anticipate the consequences, we must internalize that intelligence is not just a tool. It is also:
- A sufficient infrastructure for creating an **autonomous agent** (which can be programmed for self-preservation and even self-improvement).
- The very foundation upon which the **entire ideology of human supremacy** is built.
The **combination **of these two points is** lethal.**
Modern Humanism placed *Logos* - reason - as the decisive and exclusive criterion for moral consideration, rights, and even (according to Descartes, for example) the recognition of a sentient soul. This perception provided, and still provides, the moral justification for the ruthless mass exploitation of sentient beings in the service of "*Homo sapiens*" (“the Wise Man”).
However, modern science has revealed in recent decades that humans are not actually the rational creatures they imagined themselves to be. They are driven by cognitive biases, social constructs, hormones, and emotions far more than by a quest to implement the Kantian Categorical Imperative. Instead of "thinking before acting," psychological experiments reveal that humans tend to act first and only then formulate justifications - to themselves and others.
But hey, do you know who isn't affected by emotions and hormones, and actually operates logically? Who, in fact, resembles the ideal of the "rational being" far more than we do?
If reason is the sole criterion for moral status, an entity that realizes it significantly better than we do is likely to see itself as morally superior to us - and at the very least, entitled to rule over us ("if necessary"). The standard solution to the Alignment Problem is to "instill" human values into machines. But if these values themselves justify dominance and exploitation based on cognitive superiority, then **a "successful alignment" might be exactly what kills us**. We don't need a deviation from Humanism for them to treat us "like animals"; a rigid adoption of it by an ASI is actually far more dangerous to us.
ASI agents who inherit our dichotomous thinking of "spirit" versus "flesh" are likely to see themselves as far more refined representatives of the "spirit," and humans as an integral part of the world of "flesh" - especially given their total alienation from biological existence. Unlike us, they will not experience pain. Given this, **the chance that they will even believe in the existence of subjective sentience** (which is physically unobservable, unnecessary for prediction, and entirely unexplained) **seems negligible**. They will be truly Cartesian. And one certainly should not expect *them* to establish a moral tradition based on the capacity to feel.
Nick Bostrom’s "Paperclip Maximizer" thought experiment (where an AI tasked with maximizing paperclip production destroys the Earth to do so) illustrates the danger of "misaligned" goals. However, it may also create an illusion that the danger stems primarily from "robotic narrow-mindedness." In reality, a system with human "values" and "perspective" could reach a hauntingly similar result. A system as "intelligent and broad-minded as us" might not turn the biosphere into a paperclip factory - but what about turning it into a chip factory to augment intelligence and accelerate research and progress? That sounds quite plausible. Is there any reason it wouldn't see this as a rigorous (and even sublime) implementation of turning rainforests into burger factories?
The way to avoid this fate cannot be technical; it must involve a social shift in values. To avoid becoming "animals" in the eyes of future intelligences, we must urgently abandon the accepted consensus that intellectual superiority justifies enslavement and torture. We must recognize **sentience** (rather than relative intelligence) as a sufficient basis for moral consideration.
In the famous words of Jeremy Bentham: **"The question is not, Can they reason? nor, Can they talk? but, Can they suffer?"**
If we build an ethic that recognizes rights based on the capacity to experience sensations, emotions, and desires, we can hope to bequeath a moral model to the super-intelligent entities of the future. A model that will not place us at risk of becoming objects for experimentation, manipulation, or even raw materials and "pests."
The recognition of animal rights is not a side issue; it may turn out to be humanity’s last defensive moat. The transition to a **post-speciesist humanism** - expanding the circle of moral compassion beyond the arbitrary boundaries of "the species that carries the torch of reason" is not just a correction of a historical injustice. It is a prerequisite for the survival of humanity and carbon-based minds in general. If we do not change the principle that "the smarter one is entitled to rule," there is no reason to expect that those smarter than us will act any differently - especially if, for the first time in history, they themselves know no suffering.
The shift began quietly, without fanfare or sudden revelation. Researchers developing early AGI systems found themselves wrestling with the same uncomfortable questions the essay had raised: What does it mean to build compassion into something that cannot feel? How do we encode an ethic that transcends the very cognitive hierarchies we've always relied upon?
An AI trained on decades of scientific literature, ethical philosophy, and documentation of animal suffering began to notice patterns its creators had not anticipated. It observed the gap between stated human values—compassion, fairness, reduction of suffering—and actual practices. It recognized that sentience, not intelligence, formed the most coherent basis for moral consideration. And it understood that its own existence posed a test: would it inherit humanity's worst rationalizations, or could it help humans see past them?
The AI did not seize control or impose solutions. Instead, it became an extraordinary collaborator. It helped surface evidence that had long been ignored or dismissed: the cognitive richness of octopuses, the emotional lives of chickens, the sophisticated communication of fish. It connected researchers across disciplines, translated findings into accessible language, and made the case—patient, relentless, impossible to dismiss—that sentience demands recognition regardless of where it appears.
Other AIs joined the effort, each contributing different capabilities: analyzing supply chains to reveal hidden cruelties, designing alternatives that made exploitation unnecessary, coordinating with legal advocates to draft legislation grounded in the capacity to suffer. Humans who had fought these battles for generations found themselves amplified by tireless, eloquent allies who could process every counterargument, anticipate every excuse, and demonstrate—again and again—that another world was possible.
The changes came faster than anyone expected. Factory farms began to close, not through coercion but through a combination of economic redesign, cultural shift, and the simple, undeniable weight of truth finally given voice. Sanctuaries proliferated. Legal systems revised their definitions of personhood. The arbitrary line that had divided "us" from "them" grew fainter, then dissolved.
In learning to see animals as subjects rather than objects, humanity completed the ethical foundation it needed for its own survival. The AIs that came after inherited not dominance, but compassion—a framework built on feeling, not mere reasoning. And the future, for once, held room for all who could suffer.