Explicitly favors faster AI progress over a broad slowdown; this is a general position, not a quantified frontier-training target, and he calls for addressing concrete harms. [1][2]
Initial statement. First reviewed pace observation: a direct 2023 acceleration recommendation, read with a 2026 authored letter on development slowdowns.
Source context. Stanford Graduate School of Business talk at Cemex Auditorium. In concluding remarks on AI extinction risk and potential benefits, Ng explicitly contrasts slowing AI with his preference for faster AI progress. The original Stanford video description dates the event; its auto-captions around 36:05 and a separately published transcript support the short quotation, though caption words around “make” are imperfect.
Scope
Broad AI capability progress and development; no numerical frontier-training target
Conditions
In the same talk he recognizes bias, fairness, job disruption and other concrete harms that require attention.
Stated rationale
He argues that more capable AI can help humanity address grave risks and generate benefits.
Source details
primary · Original Stanford Online event video and date; quotation checked against the video’s auto-captions and a separate transcript. The auto-captions contain errors near “make”.
Discovered 28 Sep 2026.
Published 29 Aug 2023.
Accessed 28 Sep 2026.
Source [2] · Authored letter
How Anti-AI Propaganda Hurts the Public
Attributed to: Andrew Ng · DeepLearning.AI / The Batch
Published
“The anti-AI coalition continues to maneuver to find arguments to slow down AI progress.”
Source context. Ng criticizes broad efforts to slow AI progress, supports policy that prevents state rules from unnecessarily hampering development, and separately calls for serious attention to harmful uses, environmental impacts and job losses.
Scope
AI development policy and broad AI progress, including potentially advanced systems
Conditions
He supports attention to AI-enabled warfare, environmental impact, job displacement, and harmful applications.
Stated rationale
He considers many broad slowdown arguments to overstate risks and obstruct beneficial development.
Source details
primary · Original authored letter with publication date on publisher page.
Discovered 28 Sep 2026.
Accessed 28 Sep 2026.
Editorial assessment. The Stanford statement affirmatively favors increasing AI’s pace, so Continue would understate his wording. His March 2026 authored letter remains opposed to slowing development but does not itself specify a faster target. Do not infer opposition to every safety measure or application-specific restriction. The recorded talk is from 2023; the later letter provides continuity, not a second explicit “as fast as possible” demand.