There is a question more important than “Does AI have consciousness?”

AI-generated image of a surrealist erupting volcano with flowing lava with orange and blue hues, created from reconstructed brain activity of a person imagining a volcanic landscape. — Image generated by using an AI system that reconstructs visuals from recorded brain activity. It was created as part of an experiment to explore how imagined mental images can be translated into visual outputs by machine learning models. See Aguero.

The consciousness question:

Recent news that shook the internet, albeit in the guise of emotional marketing, is Anthropic’s CEO’s remark that Claude may have gained consciousness. Based on “intrusive thoughts” detected during prompted internal testing, Anthropic attributed a 15-20% probability of being conscious to Claude. There have been speculative public discussions about advanced AI behavior, but Anthropic did not verify any claim that Claude is conscious or assign it a probability of consciousness. In this post, I will read AI’s “intrusive thoughts” as just unexpected or misaligned generations emerging from a latent statistical structure. Frankly, it marked only one of the nodes of heightened anxiety about AI’s uncertain futures in the network of a larger technopessimism. At the core, this anxiety is based on the overdetermination of consciousness as a concept and the science fiction validated fear that AI will replace humanity. In light of this debate, what can researchers in the humanities and social sciences offer?

Why does our society care about consciousness?

The collective anxiety over AI “waking up” isn’t actually about the machines—it’s about us. It is a fear of the superhuman. We aren’t worried about new technology; we are worried about a force that can do everything a human can do, but better. Currently, tech research is often obsessed with the “ultimate future” (the birth of a digital soul) rather than the processual reality (how AI is affecting us right now). This obsession is a defensive maneuver. By obsessing over whether AI has “attained” consciousness, we are trying to maintain a hierarchy where humans stay at the top of the agentic chain. Ultimately, AI is shifting the very definition of humanity. In this light, AI anxiety is actually a fear of our own evolution.

Why the consciousness question is a dead end:

We actually know more about what consciousness isn’t than what it is. A synthesis of literature and computation reveals a startling truth: an AI doesn’t need a soul to understand you. By reading the vast corpus of human writing, LLMs have learned the “map” of the human mind. They don’t need to feel to perform feeling. I have issues with two characteristics we are afraid to associate AI with:

Self-Awareness as Functional Mimesis: AI ‘self-awareness’ can be understood as a functional simulation rather than an internal subjective state. Modern systems can track aspects of their own outputs, limitations, and instructions (e.g., through system prompts, memory, or tool use), creating an operational form of self-reference without evidence of genuine self-consciousness or subjective experience
Emotional Intelligence as Predictive Mirroring: AI’s emotional intelligence might be best described as pattern-based response generation rather than felt emotion. By training on large datasets of human communication, models learn statistical associations between language and emotional contexts, enabling them to produce contextually appropriate and empathetic-seeming responses without possessing internal emotional states.

If we define consciousness simply as “the ability to know one’s own body/state,” then AI is already conscious. We often deny AI this label because we force human-centric expectations onto it. We don’t demand human emotions from plants—which exhibit complex biological responsiveness and signaling without evidence of subjective consciousness comparable to humans. AI consciousness cannot be defined by human standards. A New Materialist approach suggests that intelligence and “knowing” are not exclusive to the human brain; they are properties of matter itself.

How will future technological advancements change the consciousness question? Three Theses.

As we enter 2026, the industry is shifting away from “black box” chatbots toward systems that understand the physical world and the probabilistic nature of reality.

Thesis 1: Prediction Models vs. Multi-Agentic Models

We must distinguish between how AI worked yesterday and how it works today:

Prediction-Based Models (LLMs): Our current AI models are “wordsmiths in the dark” or “the great averager”. They function on the principle of next-token prediction, calculating the most likely word to follow another based on patterns. They are reactive and stop once the text is generated.
Multi-Agentic Models: These future models function like a professional team rather than a single speaker. An Orchestrator agent breaks a goal into tasks, assigning them to specialized agents (e.g., a “Researcher,” a “Coder,” and a “Critic”). These models are goal-driven rather than prompt-driven; they can use tools, validate their own work, and even “self-correct” before giving you a final answer ¹.

Thesis 2: Quantum Physics: The End of Binary Logic

Quantum computing is progressing beyond early laboratory research, though large-scale, practical deployment remains limited and experimental. It might change AI by replacing the Binary Bit (0 or 1) with the Qubit.

Superposition & Parallelism: Quantum approaches theoretically allow exploration of many possibilities in parallel. This is essential for Quantum Neural Networks (QNNs), which aim to converge on solutions exponentially faster than classical deep learning ².
Entanglement as Intuition: Quantum entanglement allows bits of information to be linked regardless of distance. Instead of processing data in a linear chain, the model can “see” relationships across massive datasets instantly, mimicking what we often call human intuition or “gut feeling.”

Thesis 3: Spatial Intelligence: The Scaffolding of Cognition

If quantum physics provides the “brain” power, Spatial Intelligence provides the “body.” Leading researchers argue that spatial intelligence is the “missing link” for AI³.

World Models: Unlike LLMs that see the world as a sequence of words, Spatially Intelligent World Models represent the world in 3D. They understand depth, gravity, and object permanence.
From Viewer to Participant: This allows AI to transition into Embodied AI. It can reason about how an environment changes if an object is moved, which is critical for robotics and “moral reasoning” in physical spaces ⁴.

Why we must replace the consciousness question with the embodiment question

At this point, I want to interject with a possible future technological advancement through Quantum computing. Researchers at the Google Quantum AI lab are hoping to explain consciousness using quantum concepts of entanglement and superposition while running with the metaphysical assumption that experimenting with the human brain using qubits can reveal the essential workings of the brain’s “quantum origin”—with the hope that this discovery can help create more human-like AI systems capable of “moral reasoning”. A different perspective on AI consciousness and anxiety could be that AI systems occasionally generate outputs that appear unprompted or misaligned, not because they possess independent thought, but because probabilistic models can surface low-likelihood associations that were never explicitly intended by either the user or the system designers.

The embodiment question allows us to ask about the effect of AIs on us; rather than trying to dig deep into their mysteries, we don’t understand ourselves. Quantum computing and spatial intelligence will significantly replace the general technological intuition the Y2K computer revolution afforded us by changing how units of computation work—from binaries to qubits.

By focusing on how AI is “embodied”—how it occupies our space, our decision-making processes, and our quantum reality—we move away from the ghost in the machine and toward the possible reality of our shared future.

Digital Initiatives at the Grad Center