Two Pieces, One Fracture
Two recent pieces capture the current fracture in the AI debate. In The New York Times, Jaron Lanier argues that Silicon Valley is sabotaging AI by packaging it in existential dread. Anthropic’s IPO filing warns of “catastrophic or existential risks.” Leaders invite the public to accept that the technology may kill us, empty our livelihoods, or strip meaning from life.
Lanier sees the same pattern that stalled virtual reality a decade ago: not the technology itself, but the insular culture that defines it. Americans are not rejecting AI; they are rejecting how it is being sold.
In The Economist, Anne Neuberger, former U.S. Deputy National Security Advisor for Cyber, makes a complementary case from the practical side. Much of the debate asks how to stop a powerful technology from making the world more dangerous. She argues this misses the more urgent reality: the status quo is already fragile. Critical infrastructure—power grids, water systems, hospitals—operates at human speed against adversaries moving at machine speed.
AI can help defenders more than attackers if we use it aggressively on defense: finding and fixing flaws, testing resilience, containing what cannot be patched. “The time has come to use AI as aggressively on defense as adversaries will use it in attack.”
The Cost of Treating AI as a Creature
Both point toward the same conclusion from different directions. The dominant narratives—doom on one side, uncritical acceleration or denial on the other—obscure the actual work of coexistence. AI is not a god, a conscious entity, or an inevitable overlord. As Lanier has long put it, with no definitive evidence of consciousness, it is closer to a sophisticated tool than to a German shepherd. Treating it as a creature produces worse products, worse safety approaches, and worse public reception. Treating it as a technology opens the space for useful design and responsible use.
Stepping Back as Observer
That framing also creates the necessary distance for observation. Once AI is no longer a quasi-religious force, we can step back and look at it clearly. We can examine what it actually does, where it amplifies existing strengths and weaknesses, and how it is being deployed in specific systems. The same distance applies to ourselves. As an observer, I not only watch the technology; I watch the observer. Conscious, deliberate reasoning is only part of the picture. Unconscious patterns—habits of attention, projections of power or fear, the desire for certainty or control—shape how we build, regulate, and live with AI. Those patterns are harder to see and harder to change, yet they determine the outcome more than any single model release.
Self-Knowledge as the Missing Capability
This is the core argument of The Mindful Revolution: humans must learn to know themselves better in order to understand the world, and AI specifically. Speed and technical capability alone do not produce wisdom. In a world where AI can generate language, code, plans, and actions at industrial scale, the scarce resource is high-quality human judgment above the loop.
That judgment improves when we recognize both our rational capacities and the automatic processes that run underneath them. Self-knowledge is not a soft alternative to technical competence; it is the practical condition for using powerful tools without being used by them.
Designing for Meaningful Coexistence
Lanier’s “tools not creatures” and Neuberger’s insistence on aggressive defensive use are compatible once we adopt the observer stance. We can build and deploy AI systems that strengthen resilience — detecting vulnerabilities in critical infrastructure, accelerating patches, continuously red-teaming networks — without inflating them into existential myths.
We can keep humans above the loop not as a slogan, but as a design choice that requires clearer self-understanding: knowing when we are reacting from fear, status, or habit rather than from evidence and purpose.
Meaningful coexistence does not require pretending AI is harmless or that human nature is already adequate. It requires treating the technology as technology and treating ourselves as the more complex system that still needs development.
The alternative — continuing to define AI primarily through catastrophe or through unexamined acceleration — produces exactly the sabotage Lanier describes and the defensive lag Neuberger warns against. The practical path is quieter: observe the systems, observe ourselves, and design for the version of human-AI interaction in which both remain useful to each other.