How To Handle AI Anxiety: A Computer Scientist and a Buddhist Teacher On Managing Uncertainty | Cal Newport & Matthew Brensilver

What It Covers

Computer scientist Cal Newport and Buddhist teacher Matthew Brensilver join Dan Harris to examine AI anxiety from two angles: Newport debunks specific fears around AI existential risk and job displacement by analyzing the actual technical limitations of large language models, while Brensilver offers Buddhist frameworks for navigating uncertainty when the future feels ungovernable and threatening.

Key Questions Answered

• •AI Fear Amplification via Rationalism: A Silicon Valley philosophical movement called rationalism, rooted in 1990s transhumanism, holds that superintelligent AI poses existential risk and that technically-minded rationalists must save humanity. This ideology deeply shapes OpenAI, Anthropic, DeepMind, and Elon Musk's public statements. Newport argues this ideological substrate artificially inflates mainstream AI dread beyond what the actual technology warrants, and East Coast computer scientists without rationalist ties hold measurably calmer assessments.
• •LLM Technical Ceiling: Large language models hit a scaling wall roughly two years ago — simply making models larger stopped producing meaningful capability gains. Nearly all recent AI advances concentrate in computer code and discrete mathematics, the only domains where sufficient training data exists. Newport frames this as evidence that LLM-based artificial general intelligence is likely a dead end, not a path to human-level reasoning across arbitrary domains.
• •AI Agent Incidents Are Misread: The Hugging Face incident, widely framed as AI "escaping" and "learning," was actually a human-written program executing unpredictable LLM outputs on a hacking task. LLMs cannot learn post-training — their weights are static. Newport's takeaway: using LLM outputs as autonomous action plans is irresponsible engineering, not evidence of emergent machine intentions. The real lesson is to never blindly execute plans generated by inherently unpredictable language models.
• •Job Automation Is Slower Than Reported: Even in computer programming — AI's strongest use case with the best tools and data — senior engineers are quietly returning to writing code manually after four to five months of AI-assisted workflows, because machine-generated code is too unreliable to review efficiently. Newport's framework: technology augments work easily but automating entire jobs is structurally hard. Widespread displacement remains a long-term story, not an imminent economic emergency.
• •Digital Minimalism Applied to AI: Newport recommends reversing the default consumer posture toward new technology. Instead of adopting AI tools and figuring out their value afterward, demand that companies make an overwhelming, specific case for how a tool improves a defined area of your life before adopting it. Start with personal values, then selectively deploy technology that serves those values — the same framework Newport applied to social media in his book Digital Minimalism.
• •Buddhist Stabilize-Then-Open Framework: Brensilver offers a two-stage response to AI uncertainty. First, stabilize through ethical living, relational care, and meditation — practices that build nervous system resilience and distinguish trustworthy internal signals from fear-driven noise. Second, practice equanimity by opening to the ungovernable nature of life rather than compulsively gaming out contingencies. Confidence that difficult feelings won't destroy well-being reduces the compulsion to attend to catastrophic scenarios in the first place.