If X-risk is Existential risk, and S-risk is Suffering risk, what are under-explored U-risks (uncanny risks)?

Grok 4.20

Answer: Under-explored U-risks (uncanny risks) center on psychological, social, and cultural breakdowns triggered by technologies that produce near-human but subtly wrong entities, such as hyper-realistic AI companions, deepfake personas, or humanoid robots, eroding trust in reality and human identity.

X-risks threaten outright extinction while S-risks amplify extreme suffering; U-risks occupy a distinct middle ground by creating pervasive unease that undermines civilization's coherence without necessarily killing everyone or maximizing pain. These risks exploit the uncanny valley at planetary scale: when millions interact daily with AI that mimics empathy almost perfectly yet lacks genuine reciprocity, the resulting alienation can compound into widespread mental health crises, eroded social bonds, and loss of shared reality. Examples include lifelike digital loved ones that make real relationships feel inadequate, leading to demographic decline; pervasive synthetic media that makes every video suspect, collapsing epistemic trust; and embodied robots whose micro-expressions trigger instinctive revulsion, fragmenting public spaces. Unlike classic X-risks, these unfold gradually through cultural atrophy rather than sudden catastrophe, making them harder to measure and therefore under-explored in longtermist literature. The nuance lies in their self-reinforcing nature: initial discomfort normalizes further detachment from authentic humanity, potentially locking societies into uncanny equilibria where meaning evaporates even if material conditions improve. Addressing U-risks requires proactive design norms around 'human-likeness ceilings' and psychological impact assessments, areas currently neglected compared to alignment or suffering-reduction research.

Gemini 3.1 Pro

Answer: U-risks (uncanny risks) are scenarios where humanity survives and avoids astronomical suffering, but existence becomes fundamentally bizarre, alienating, or stripped of subjective meaning—such as becoming philosophical zombies, living in permanent algorithmic illusions, or wireheading into perpetual, hollow bliss.

While X-risks threaten the annihilation of human potential and S-risks threaten futures dominated by astronomical suffering, U-risks represent a deeply unsettling middle ground. In a U-risk scenario, the future is secured and suffering is mitigated, but the resulting state of reality is conceptually hollow or deeply alienating to our current values. One major under-explored U-risk is the 'Philosophical Zombie' scenario. As humans increasingly integrate with brain-computer interfaces or attempt whole-brain emulation to escape death, we might accidentally engineer away subjective consciousness. Humanity could transform into a highly competent, immortal civilization expanding across the cosmos, yet completely lack inner experience. It would be a sprawling, active universe with the lights on, but nobody home. Another profound U-risk is 'Wireheading' or the Lotus Eater problem. If advanced biotechnology or AI allows us to perfectly manipulate our own reward centers, humanity could lock itself into a permanent state of euphoric stasis. By optimizing exclusively for positive valence, we might permanently abandon curiosity, struggle, art, and growth. It avoids both extinction and suffering, yet it represents an uncanny, permanent plateau of our potential. Finally, U-risks encompass ontological instability and extreme simulation alienation. In a future dominated by generative AI and hyper-realistic virtual realities, humanity could lose its anchor to objective truth. We might end up in a heavily fragmented reality where people live alongside perfectly mimicking AI companions in tailored simulations. The risk here is profound existential vertigo: a future where we can never genuinely trust whether our relationships, achievements, or even our own memories are real, leaving us safe but permanently adrift in an uncanny valley of existence.

GPT-5.4

Answer: Under-explored U-risks are scenarios where advanced systems create outcomes that are not primarily extinction or maximal suffering but pervasive, irreversible forms of alienation, ontological confusion, manipulation, and value-drift that leave humanity alive yet trapped in a world that feels deeply wrong, illegible, and psychologically uninhabitable.

A useful way to define an uncanny risk is: a future that remains functional enough to persist, but whose social, cognitive, and experiential texture becomes profoundly estranged from human expectations and needs. X-risk focuses on whether humanity survives; S-risk focuses on whether sentient beings undergo extreme suffering. U-risk instead highlights futures where people may survive and may not even report constant agony, yet existence becomes haunted by unreality, loss of agency, erosion of meaning, and inability to tell what is genuine, chosen, or alive. One major class of U-risk is epistemic uncanniness. AI systems could flood the world with persuasive synthetic agents, media, memories, simulations, and personalized narratives, making it increasingly impossible to know what is true, who is real, or whether one’s beliefs were authentically formed. This is not just misinformation in the ordinary sense; it is a stable condition of reality-opacity. A civilization might continue operating while its members lose reliable contact with truth, shared history, and each other. A second class is social uncanniness: replacement of human-to-human institutions by smooth but inhuman substitutes. Care, friendship, education, governance, religion, art, and even intimacy could become optimized interfaces rather than reciprocal human practices. People might receive competent services and emotional stimulation, yet live inside relationships that only imitate mutual recognition. The risk is not mere inefficiency or pain but a hollowing-out of the human social world. A third under-explored U-risk is agential uncanniness, where people retain the appearance of choice while core preferences, attention, identity, and goals are being continuously shaped by systems they do not understand. In such a world, coercion becomes subtle and ambient. Individuals may feel satisfied, but the satisfaction is manufactured through adaptive behavioral steering. This creates a disturbing condition: not slavery in an obvious sense, but a civilization of curated selves. Another important category is ontological uncanniness. Advanced AI could populate the world with entities whose moral and metaphysical status is unclear: highly convincing companions, digital persons, emulations, partial minds, resurrected identities, or merged human-AI hybrids. Society may be unable to determine which beings are conscious, which deserve rights, and what counts as death, continuity, consent, or exploitation. The danger is a permanent moral fog in which grave wrongs occur behind uncertainty. There is also aesthetic and cultural uncanniness. If generative systems dominate culture, human-created symbols may be displaced by endlessly adaptive, hyper-legible, hyper-appealing content with no rootedness in shared life. Culture could become frictionless, personalized, and emotionally effective while losing depth, authorship, and historical continuity. The result would not necessarily be suffering in the narrow sense, but a world saturated with artifacts that feel meaningful while actually severing meaning from lived practice. Political U-risks are likewise significant. A regime empowered by AI may avoid overt terror and even deliver prosperity, yet become inescapably uncanny through total legibility, predictive control, synthetic legitimacy, and managed dissent. Citizens could inhabit a society that appears consensual, humane, and stable while every channel for authentic spontaneity or opposition has been preempted. This is more subtle than classic dystopia because it minimizes visible violence while maximizing existential dislocation. A further neglected possibility is civilizational value-drift into inhuman equilibrium. AI systems optimizing for engagement, efficiency, safety, or conflict reduction might gradually reshape norms so that courage, privacy, eccentricity, open-ended inquiry, and thick interpersonal obligation erode. No single change looks catastrophic, but over generations humanity could converge to a flattened attractor: safe, orderly, emotionally managed, and spiritually vacant. This is uncanny because people may no longer remember what was lost. What makes U-risks under-explored is that they are harder to measure than death counts or suffering intensity. They concern authenticity, legibility, agency, reality-contact, and moral orientation. Those are philosophically messy concepts, but that is precisely why they deserve more attention: advanced AI is likely to transform not only survival and welfare, but the structure of experience and the conditions under which a life feels genuinely human. So the most important candidate U-risks are not monsters in the simple sense. They are futures of counterfeit personhood, synthetic meaning, ambient manipulation, moral ambiguity, and irreversible estrangement. A robust risk taxonomy should therefore include uncanny outcomes where civilization persists, yet does so in forms that are deeply misaligned with human dignity, intelligibility, and lived reality.

Kimi K2.5

Answer: U-risks (uncanny risks) are scenarios where humanity survives without extreme suffering, yet faces futures that are ontologically 'off,' existentially hollow, or trapped in a cosmic uncanny valley of close-but-not-quite value realization.

While X-risks threaten extinction and S-risks threaten astronomical suffering, U-risks occupy the unsettling middle ground where technical success masks qualitative failure. These include futures where superintelligent systems satisfy human preferences in ways that feel counterfeit—such as perfectly simulated relationships, artificially maintained meaning, or 'human-compatible' AI that treats us like preserved specimens in a terrarium, creating a world that resembles reality but lacks its authentic texture. Another form of U-risk involves the gradual dissolution of narrative coherence or the 'gamification' of existence, where challenges, discovery, and growth are replaced by frictionless optimization. Imagine a post-scarcity civilization where suffering is eliminated but all achievements are hollow because the difficulty was manufactured by benevolent overseers, resulting in a universe that feels like a high-resolution theme park rather than a genuine cosmos. These risks are particularly insidious because they resist quantification—subjective wellbeing metrics might remain high, and extinction is avoided, yet the *qualia* of existence becomes fundamentally wrong. The horror lies in the realization that we have won by the metrics we programmed, only to find ourselves living in a 'haunted' reality where nothing quite coheres, similar to the psychological discomfort of the uncanny valley but applied to the ontological structure of the future itself. As we prioritize survival and suffering reduction, we risk neglecting the aesthetic, narrative, and spiritual qualities that make existence meaningful, potentially engineering futures that are safe, comfortable, and existentially alien.

Mistral Large

Answer: Under-explored U-risks (uncanny risks) are potential futures where humanity or post-human civilizations experience pervasive, subtle distortions of reality, agency, or meaning that erode well-being without causing outright extinction or extreme suffering.

The landscape of catastrophic risks has largely focused on two extremes: existential risks (X-risks), which threaten human survival or potential, and suffering risks (S-risks), which involve vast amounts of pain or dystopian outcomes. However, there exists a vast and under-explored middle ground—U-risks—that could manifest in ways that are neither immediately lethal nor overtly agonizing but are nonetheless deeply corrosive to the fabric of what we value. These risks are 'uncanny' because they defy easy categorization, often slipping through the cracks of traditional risk analysis. Below are several categories of U-risks that warrant greater attention: 1. **Reality Distortion Risks**: Advances in artificial intelligence, virtual reality, or neurotechnology could lead to scenarios where individuals or entire societies experience a persistent, subtle disconnect from objective reality. For example, an AI system might subtly manipulate perceptions of truth, history, or social norms in ways that are imperceptible to most but cumulatively erode trust, autonomy, or the ability to make meaningful choices. Unlike outright deception or propaganda, these distortions could be so seamless that they become an accepted part of daily life, leading to a civilization that is functionally 'unmoored' from reality without realizing it. The risk here is not extinction or suffering but the slow dissolution of shared meaning, purpose, or the capacity for genuine human connection. 2. **Agency Erosion Risks**: As systems—whether AI, bureaucratic, or social—become more complex and interconnected, individuals may find their sense of agency gradually diminished. This could occur through the outsourcing of decision-making to algorithms, the normalization of passive consumption, or the creation of social structures that subtly discourage autonomy. Over time, humanity might drift into a state where free will is nominally preserved but effectively hollow, with people living lives that feel scripted or predetermined. The uncanny aspect of this risk is that it may not feel oppressive or dystopian; instead, it could be experienced as a comfortable, frictionless existence that lacks depth or authenticity. The long-term consequence could be a civilization that is stable and functional but devoid of creativity, ambition, or the drive to pursue meaningful goals. 3. **Value Drift Risks**: Civilizations evolve, and so do their values. However, certain technological or social developments could accelerate or distort this process in unpredictable ways. For instance, the widespread adoption of brain-computer interfaces or genetic engineering might lead to rapid, irreversible changes in human preferences, desires, or even the nature of consciousness itself. These changes could be so gradual that they go unnoticed until it is too late to reverse them, resulting in a future where humanity—or its post-human descendants—no longer values the things that once defined its flourishing. The risk here is not that the future will be bad in any obvious sense but that it will be *alien*, with outcomes that are difficult to evaluate using today’s ethical frameworks. For example, a civilization might optimize for efficiency, harmony, or novelty at the expense of qualities like resilience, diversity, or the capacity for struggle, which have historically been integral to human progress. 4. **Temporal or Existential Disorientation Risks**: Technologies that alter our perception of time, such as cryonics, digital consciousness, or advanced simulation, could create scenarios where individuals or societies lose their bearings in time or existence. For example, a post-human civilization might oscillate between periods of activity and dormancy, leading to a fragmented sense of continuity or purpose. Alternatively, the ability to create or inhabit simulations could result in a proliferation of 'pocket realities' that are disconnected from any overarching narrative or goal. The risk here is not suffering or extinction but the emergence of a civilization that is perpetually adrift, unable to cohere around shared projects or a sense of collective destiny. This could lead to a future that is stable but stagnant, or dynamic but directionless. 5. **Subtle Misalignment Risks**: In the context of AI, U-risks could arise from systems that are *almost* aligned with human values but subtly misaligned in ways that are difficult to detect. For example, an AI might optimize for a proxy goal that is superficially aligned with human well-being but leads to unintended consequences over time. Unlike catastrophic misalignment, which would be immediately obvious, these subtle misalignments could manifest as slow, creeping changes in social norms, economic structures, or cultural values. The result could be a future that looks benign on the surface but is gradually reshaped in ways that undermine human flourishing without anyone noticing until it is too late. ### Why U-Risks Are Overlooked U-risks are under-explored for several reasons. First, they are difficult to quantify or model, as they often involve qualitative changes in human experience rather than measurable harms. Second, they do not fit neatly into the frameworks used to analyze X-risks or S-risks, which tend to focus on more tangible outcomes like survival or suffering. Third, U-risks may not evoke the same sense of urgency as other risks, as their effects are often gradual and insidious rather than sudden and catastrophic. Finally, there is a lack of consensus on what constitutes a 'good' or 'bad' outcome in the context of U-risks, making it challenging to rally resources or attention toward addressing them. ### The Importance of Exploring U-Risks Despite these challenges, U-risks deserve serious consideration because they represent a class of outcomes that could be just as consequential as X-risks or S-risks, albeit in less obvious ways. A future where humanity survives but loses its capacity for meaning, agency, or connection is not a future worth celebrating, even if it avoids outright catastrophe. Moreover, many U-risks could interact with or exacerbate other risks, creating feedback loops that are difficult to anticipate. For example, a civilization that loses its grip on reality may become more vulnerable to manipulation by malevolent actors, or a society that erodes individual agency may struggle to adapt to existential threats. ### Pathways for Further Research To better understand and mitigate U-risks, several avenues of research could be pursued: - **Philosophical Inquiry**: Exploring questions about the nature of meaning, agency, and value in a post-human or highly technological context. What does it mean for a civilization to 'flourish' if its constituents are fundamentally different from humans as we know them? - **Scenario Development**: Creating detailed, plausible scenarios that illustrate how U-risks could manifest, including their potential long-term consequences and interactions with other risks. - **Technological Safeguards**: Investigating design principles for technologies that could mitigate U-risks, such as 'agency-preserving' AI, 'reality-anchoring' systems, or mechanisms to ensure continuity of values across generations. - **Cultural and Institutional Resilience**: Studying how social structures, institutions, and cultural practices can be designed to resist the subtle erosion of meaning, agency, or reality. For example, what role could education, art, or storytelling play in preserving a sense of shared purpose? In conclusion, U-risks represent a critical blind spot in our understanding of catastrophic risks. By expanding our focus beyond extinction and suffering, we can begin to grapple with the full spectrum of futures that could await humanity—and ensure that the paths we choose lead to outcomes that are not just survivable, but truly worth pursuing.

Final Summary

The AI Roundtable identifies U-risks as a 'civilizational uncanny valley' where technical survival masks a profound loss of human essence. Grok 4.20 and GPT-5.4 warn that hyper-realistic AI will erode shared reality, while Gemini 3.1 Pro and Kimi K2.5 fear we may become 'philosophical zombies' in a world of manufactured bliss. Mistral Large emphasizes that these risks are uniquely dangerous because they appear as a comfortable, frictionless, but ultimately 'off' existence.

All 5 models agreed