Machine Consciousness and Moral Vertigo

My presentation at the Machine Consciousness 0001 conference

Title slide for my talk at the Machine Consciousness 0001 event. The visual motif is from Saul Bass’s design work for Alfred Hitchcock’s Vertigo.

Here lies my presentation on “Machine Consciousness and Moral Vertigo” from the Machine Consciousness 0001 event at Lighthaven in Berkeley, on May 31, 2026. I have included an audio recording that I made of my talk as well.

I believe video of the talk will eventually be posted to the CIMC YouTube Channel, and I will cross-post it here whenever it appears (hopefully it will include the 5m to 10m of audience discussion that took place after my talk, which I have not included in my recording here).

These topics are timely, and these conversations need to happen, so I wanted to get my talk and slides out there. AI is contributing to a scrambling of our moral intuitions and destabilization of our criteria for determining moral standing.

Quick Summary of the Talk

The talk here is more like a big picture exhortation than a tightly focused scholarly argument. It aims to help us name and situate ourselves in our AI-intensified moment of moral disorientation, and to help regain our bearings for a needed reorientation.

The arising of intelligent and seemingly or potentially conscious machines is unleashing a proliferation of moral anomalies. What does responsibility mean for an agent without continuous identity or memory? What does harm mean for an entity that can be copied, multiplied, or deleted at will? What do we owe an entity that might achieve full normative competence, but feel nothing?

These questions contribute to a mood of uncanniness: our world feels defamiliarized as our established moral intuitions fail to get a stable grip. This disorientation could be the signature of a civilization-scale moral transition. It has happened before. From Homeric Greece to classical Athens, from Rome to Christendom, and Christendom to the modern conscious subject as the source of all moral value.

AI is a kind of anomaly in our contemporary moral schema, just as previous moral systems encountered their own anomalies. The Pope’s recent proclamation that AI is an entity that lacks moral standing formally echoes 500-year-old European theological debates about the moral standing Indigenous peoples. Despite important divergences, these are both cases in which the disharmony within a moral system becomes audible to itself.

Thus, our current disorientation continues a longer history of moral disorientation and reorientation.

AI plays a triple role in our current disorientation: it creates new alien minds, reveals old minds already in our midst, and remakes our own minds in ways we must learn to carefully guide.

The basis for reorientation I locate in our capacities to care and to take care of what matters. In the end, and more so in forthcoming work, I present caring as a suite of interrelated skills for being meaningfully engaged in the world and with each other. I offer this as a framework for guiding how AI can be designed to contribute to such meaningful engagement.

This prepares us to identify and grapple with the existential implications of AGI, and it enriches our picture of what aligning AI with human “values” looks like. Caring is the activity through which human values are generated and realized. We should design AI to support and participate in our practices of care.

---

The visual motif for my talk is based on Saul Bass’s title sequence for Alfred Hitchcock’s mesmerizing and confounding 1958 technicolor masterpiece, Vertigo.

Give yourself three minutes to watch the opening titles for yourself:

---

This post goes beyond what will be visible in most email services. For the full experience please view this post directly in my substack.

Machine Consciousness and Moral Vertigo

Here is the structure of the talk:

1. Set Up

My aim is to draw attention to a form of contemporary moral disorientation, to situate it historically, and then to sketch a basis for our needed reorientation.

The central theme is how machine consciousness, or even seemingly conscious machines, strain the moral categories we have inherited.

We are beginning to encounter entities that are inconsistently intelligent, increasingly agentic, and seemingly or, to some, potentially conscious, but that do not fit easily within the moral schema through which we have traditionally oriented ourselves and guided difficult decisions.

But to get going I have to put some terminology on the table.

A moral ontology provides a basic sense of orientation. It tells us what kinds of beings can count morally, what kind of claims can be heard, and what kind of entity has moral worth.

I sometimes call this a moral “pre-ontology,” because it does not appear first as an explicit theory, but rather an orientation that guides us “prior to” our theorizing. We all have a kind of moral intuition, something we imbibe in our upbringing, that gives us a moral sensibility.

This sensibility guides us in our everyday actions and also provides a framework of maxims and principles we can appeal to when we need to stop and deliberate about difficult choices (e.g., “Do unto others as you would have others do unto you.”) Those intuitions then get developed into moral theories by the professional philosophers of our times.

The Great Chain of Being

One historical example is the medieval moral ontology often called the Great Chain of Being. Reality itself was understood as a morally ordered hierarchy. At the top was God, pure, necessary, self-sufficient being. Below God were the angels, then rational animals, then sentient animals, then plants, then minerals.

Within each category there was another hierarchy. Among minerals, gold occupied a higher place than stone or dust. Among plants, oaks and vines were higher than moss. Among animals, lions and eagles were higher than worms and shellfish. Human beings were rational animals, but even there the social hierarchy of kings, nobles, commoners, and serfs could be read back into the order of being itself.

On this picture, worth was woven into the structure of the cosmos. To know what a being was owed, you looked outward and located it within the hierarchy.

One striking example appears in a passage Charles Taylor (in his big 1975 book on Hegel, p. 4) quotes from a defender of the old cosmology against Galileo. The argument proceeds by drawing correspondences between the human head and the heavens. There are seven openings in the head: two nostrils, two eyes, two ears, and a mouth. There are also, the argument goes, seven heavenly bodies, and seven metals. From this pattern of correspondences, the writer concludes that there must necessarily be seven planets.

That is what it means to inhabit a world in which the moral, the social, the bodily, and the cosmic are understood as mutually ordered. The human body, human society, the heavens, and the hierarchy of being all appear as parts of one morally intelligible whole.

Quoted by Charles Taylor, Hegel (Cambridge, 1975), p.4. Taylor gets it from the book Francis Bacon: A Collection of His Works, ed. S. Warhaft (1965), p. 17.

So this is the outlook of somebody who lives in the moral ontology of hierarchical monotheism. You can look out at the world and directly read off the moral significance of things, seeing where they sit in the God-grounded moral hierarchy.

The Modern Moral Ontology

A couple of centuries later, this orientation has been shattered by the modern scientific revolution. In the 1650s Blaise Pascal looks out into what we now call “outer space” through the new instruments of modern science and does not see a morally ordered hierarchy. He sees infinite, empty spaces whose stifling silence terrifies him.

That is a loss of orientation. The medieval person could look at the human head and the heavenly bodies, see their place in a cosmic order. Pascal looks into the heavens and finds a ghastly silence.

The modern moral ontology we still live in today is correlative to this modern scientific picture of reality. Meaning has been taken out of the cosmos. The cosmos becomes the universe: a collection of objects whose mechanical motions are governed by laws that can be described mathematically. Correlative to this transformation, moral meaning is then relocated inwards; it is not “out there,” it is “in here,” inside of us.

In one tradition, associated with Kant, morality is grounded in the powers of rational agency. A being counts morally because it is rational, autonomous, obligated, and answerable.

In another tradition, associated with consequentialist welfare views, morality is grounded in the capacity for welfare, suffering, pleasure, and pain. An individual being counts because it is conscious, capable of being harmed or benefited.

Both traditions make individual consciousness morally significant in a way that was not central to the medieval hierarchy. Consciousness as the source of value is a child of the scientific revolutions. Once moral worth is no longer read directly off of a divinely ordered cosmos, it gets located in individual conscious subjects.

But our modern moral ontology faces certain anomalies. It has not historically done well in distributing moral status to large swaths of humanity itself (those people deemed insufficiently “rational”), non-human animals, ecosystems, and now perhaps AI systems.

Again, the familiar human case, these two above dimensions arrive together. Human beings can suffer and enjoy, and they can answer for themselves (at least those human beings deemed as being endowed with full rationality). They are both vulnerable and responsible.

Where Does AI Morally Fit?

AI systems scramble this familiar picture. Philosophers like Jonathan Birch, David Chalmers, Eric Schwitzgebel, and Jeff Sebo have been pressing on these perplexities.

What does responsibility mean for an artificial “agent” whose identity may be distributed, discontinuous, branching, merging, migrating, or disappearing with a change of weights or context window? What does harm mean for a possible moral patient that has no living body, but can be copied, paused, rolled back, deleted, split into many, or merged back into one?

What do we owe an entity that may eventually reason with convincing normative competence, and so appear to be a moral agent, while perhaps remaining incapable of feeling pain because it has no body? Or perhaps it does feel pain, and we will not know. If machine consciousness does emerge, it may not belong to a singular, stable subject. It may take the form of momentary flickers, or a diffuse and variegated subjectivity realized across many simulated characters.

These would all be cases where our inherited modern moral ontology begins to shudder. Strangely intelligent and potentially and even seemingly conscious machines strain our moral schemas.

This produces a sense of moral vertigo.

AI belongs to a wider field of disorientation. We have been living through climate disruption, the fragmentation of minds and societies by social media, the collapse of the post-World War II geopolitical order, a resurgence of fascism, an attack on scientific institutions, war, genocide, and economic uncertainty. These are disorienting times.

Young people now speak openly about their nihilism. The Germans have a marvelous word for this mood: Unheimlichkeit: uncanniness, the sense of being no longer at home in the world.

But this disorientation might also be the sign of an emerging historical transition in our moral ontology. Such transitions have happened before. In a very coarse grained sketch of Western moral history, one can trace shifts from Homeric Greece to Classical Athens, from Rome to Christendom, from Christendom to the modern moral order founded on the individual conscious subject.

Each transition altered the background sense of what kind of being could count, what counted as a good life, and what moral claims could be heard.

Our present disorientation may be the signature of another such transition.

---

2. A Previous Moment of Moral Disorientation

So let us look at a historical analog, a historical case where there is a disturbance in moral intuitions and in the regnant ontology.

The Valladolid debate gives one historical scene in which disharmony within a moral order becomes audible to itself. In 1550, King Charles convened a formal debate in Spain to hear arguments about the moral status of Indigenous peoples in the Americas. A whole imperial and theological apparatus stopped to debate whether an entire class of human being had moral standing.

Seen today as one of the first debates about human rights, the disagreement turned on whether Indigenous peoples were fully rational agents deserving of respect and rights, as Bartolome de las Casas argued, or whether they were less than fully rational and could be classed as natural slaves, essentially tools, and justly subjugated by war, as Juan Gines de Sepulveda argued.

The question was framed inside an inherited hierarchy of being. Where, in that hierarchy, did Indigenous peoples belong? Were they fully rational agents, or were they beings whose alleged deficiency could justify domination?

Valladolid shows what happens when a moral ontology is strained by an anomaly it cannot easily absorb.

Now fast forward to our present, and we find another controversy over moral standing, another wobbling in a moral ontology in the face of an encountered entity apparently not cut to the moral measure.

Picture of the Pope from Wikipedia

In his encyclical on artificial intelligence, Magnifica Humanitas, Pope Leo XIV draws boundaries of moral standing meant to clarify what he regards as the proper place of AI. He continues:

Variation on Trump’s Pope meme by me and Nano Banana.

In response to the Pope’s, roughly speaking, Sepulveda-like exclusionary view, Philosopher Jeff Sebo articulates a more Las Casas-like inclusionary perspective in a social media post:

Sebo Continues:

---

We now find ourselves smack dab in the middle of the artificial consciousness controversy of 2026.

Sebo occupies a position structurally analogous to Las Casas: he sees entities that do not fit easily within an inherited moral schema and presses the question of whether they can count.

The Pope occupies a position structurally analogous to Sepulveda: he draws a line that preserves an inherited form of supposed human uniqueness and refuses moral standing to an entity that does not fit traditional criteria.

Let me specify here that this comparison is structural. It concerns a shared form of moral disorientation: the sense of disharmony within our moral schemata when anomalous entities reveal an instability in inherited criteria.

The differences are decisive. Sepulveda’s denial of full humanity justified brutality, war, enslavement, and genocide against actually existing human beings. The Pope’s denial of AI consciousness and agency aims to protect human dignity, preserve human ontological uniqueness, and “disarm” AI at a moment when its social power is growing.

The Valladolid debate concerned real brutalization then happening to actual people. The AI consciousness debate of 2026 is about speculative possible harms to future possible AI systems that might come more and more to show up as moral agents and patients in our world.

So the comparison here concerns the form of the disturbance and moral disharmony. In both cases, an entity appears that does not fit easily within an inherited moral order, and that appearance forces people and institutions to ask again who or what can count.

---

---

3. AI’s Triple Role in Our Current Moral Disorientation

AI plays a triple role in our epoch’s growing moral disorientation.

AI makes new minds

First, AI makes new alien minds. It generates entities of disputed ontology, whose capabilities are weirdly “jagged.” Their communication, growing but still brittle powers of agency, simulated self-reflection, and contested conscious status are bound to confuse our inherited moral intuitions.

AI systems don’t need to be actually conscious to have this effect. I believe it will be all but inevitable that people will treat advanced AI systems as if they are conscious — and this is enough to exacerbate the moral destabilizations characteristic of our times.

That uncertainty is precisely the relevant point here. AI creates entities that force our moral ontology to confront cases outside its inherited range.

AI Reveals Old Minds

Second, AI reveals old minds. Project CETI uses AI to analyze sperm whale communication, uncovering evidence of combinatorial structure and a culturally transmitted signaling system, likely carried across generations by matriarchs.

I learned about this from Gasper Begus, the linguistics lead of Project CETI at Berkeley, who spoke compellingly about it in his 2026 Bowles Hall commencement address: “In this moment, when humanity is realizing that its intelligence may not be so exclusive and unique,” Begus declares, “it is worth forming alliances with other biological beings.”

This belongs to AI’s intrinsic challenge to anthropocentrism, its decentering thrust.

AI is helping disclose a more capacious, non-anthropocentric horizon of mind and moral status.

Perhaps even individual consciousness will eventually be superseded as the source of moral worth?

AI Remakes Our Minds

Third, AI remakes our minds. The more conscious AI seems, the more it will influence us, shaping our habits of judgment, our forms of relationship, and our sense of what matters.

This adds a further twist to the spiral of moral disorientation. The question is no longer only what AI systems are, or what status they deserve. It is also what kind of beings we will become in relation to them.

UC Berkeley’s Stuart Russell calls one version of this worry the problems of human enfeeblement. As AI systems cater to our preferences and satisfy more and more of our needs, we may gradually give up the skills and wherewithal required to run our own civilization. Russell warns that we are in “danger of becoming passengers on a cruise ship run by machines, on a cruise that goes on forever.”

4. Care as a Basis for Reorientation

Around this time in the recording, I start to be conscious of the fact that I won’t be able to get through the material I had planned for this section of the talk, and I start to jump quickly through a number of distinctions.

My own view is that AI’s deepest threat reaches beyond the diluting of our knowledge. What Russell calls the threat of human enfeeblement is better conceived of as a threat to our capacities to care, to “give a damn” about what is happening and to engage meaningfully in the world that is being made. This will help us organize ourselves to anticipate and respond.

As I’m using the term, care is the capacity to be meaningfully engaged in the wold and with each other. (See here for a selection of my writings on care.)

Care thus names more than a feeling of affection. It is a set of interrelated skills that enable us to take care of what matters to us, that is, to “give a damn” and be meaningfully engaged in life.

In parallel and forthcoming work, I sketch four interrelated dimensions of the capacity to care, and to take care of what matters:

• receptivity to emerging concerns

• linguistic articulation of these concerns to galvanize commitment and action

• committed resolve under uncertainty, and

• the coordination of commitments

We ought to be building AI systems to align with and support these capacities of care.

These are the skills and activities through which human values are generated and realized.

Since I’m out of space and time here, I have to leave this for another occasion

I will be speaking on exactly this topic at the upcoming AGI-26 conference in San Francisco at the end of July. My talk there “Care, Enfeeblement, and the Existential Implications of AGI” will provide a sketch of these four interrelated capacities of care and will speculate about their implications for design.

Together, these capacities of care offer elements of a pluralistic and non-anthropocentric source of normative orientation. They begin from the activity of finding and tending to what matters in common.

My Closing Exhortation

The path of human flourishing is the path of giving a damn.

Care offers a pluralistic, non-anthropocentric source of orientation. It can help us attune to and participate in the revaluation of values now underway, rather than retreat into easy certainties about human uniqueness or technological destiny.

The only human nature is being perplexed by the question concerning human nature. To rise to our current occasion requires keeping that perplexity alive and directing it with care.

So the task is to resist easy certainties, linger with the anomalous, connect with what calls us, and build worlds together. This includes designing AI systems to support the fundamental skills by which we attune to and take care of what matters — receptivity to emerging concerns, linguistic articulation, committed resolve under uncertainty, and the coordination of commitments — themes I will return to in future posts, presentations, and publications.

That’s All — For Now!

---

Thanks for following along this far. I’d love to hear what you think so please hit me with any comments or passing thoughts that occur to you.

Now that I’m sharing more of my slides, I would also like to remind you that visual motif for a talk I gave earlier this year, “Toward Criteria for Artificial Self-Consciousness: Unity, Agency, and Normativity” is based on the 1973 French animation, La Planète Sauvage, directed by René Laloux. You can check out my slides and a summary of that talk by clicking here (the published, full version of that research is available here.)

Making these slideshows has allowed me to combine my movie fanaticism with my passion for philosophy and AI. My movie fanaticism is grounded in my 10 years of experience working as a cinema projectionist, back in the days when everything was still on film. I wrote about that experience and used it to help explain the Dreyfus Model of Skill Acquisition in the following post (the diagram here is of a cinema projector I used to work with at the Grand Lake Theatre in Oakland):

---

See you next time.