I Am a Silicon Valley Veteran. We Are Sabotaging Ourselves

Jaron Lanier argues that public hostility toward A.I. reflects Silicon Valley’s way of presenting the technology more than its capabilities. He compares today’s claims that A.I. could destroy humanity or make people obsolete with leaders’ past visions of virtual reality as a substitute for ordinary life: grandiose, alienating rhetoric helped repel the public. He sees a contradiction in companies warning of extinction while racing to build ever more powerful systems, and says the culture’s fascination with A.I. as a godlike, possibly conscious being distorts both product design and public debate. A.I. models, in his view, are valuable tools: vast records of human knowledge that can be recombined to support discovery and creation. Apocalyptic claims also have a long history as a way to attract attention and research funding.

Lanier focuses particular concern on “agents,” software that can act autonomously around A.I. models. He argues that agents are often unnecessary, citing scientific achievements such as A.I.-assisted protein research, and that autonomy introduces risks including security breaches and uncontrolled changes. Rather than asking how to stop A.I. agents from wanting to harm people, he proposes treating them as tools and requiring special review, oversight and accountability before deploying agents, especially in swarms. He describes his alternative, “Mind Gliders”: custom interfaces that let people explore ideas with A.I. while retaining control. This approach, he says, can be faster and cheaper while preserving human judgment and creativity. His central prescription is to build A.I. to assist people, not replace them.