Do not develop superintelligent AI until the problem of keeping it under human control is solved; this does not demand an end to all AI research. [1][2][3]
Initial statement. Do not develop superintelligent AI until the problem of keeping it under human control is solved; this does not demand an end to all AI research.
Source [1] · Original broadcaster interview report
Geoffrey Hinton on ABC Radio National Breakfast
Attributed to: Geoffrey Hinton · ABC News
Published
“Until we've solved that problem, it would be foolish to develop them.”
AI superintelligence smarter than humans, quoted sentence immediately after the control problem · View original source ↗
Source context. ABC reports its own Radio National Breakfast interview. The antecedent of them is superintelligent AI. Hinton also calls Amodei slowdown warning sensible. Interview recording date is not separately provided; article publication is September14. Original ABC Listen episode independently opened and dated September14 6am; recording day may differ.
Scope
Development of superintelligent AI smarter than humans.
Conditions
Do not develop the systems until human control is solved; no calendar duration or claim that all existing AI must stop.
Stated rationale
ABC reports its own Radio National Breakfast interview. The antecedent of them is superintelligent AI. Hinton also calls Amodei slowdown warning sensible. Interview recording date is not separately provided; article publication is September14. Original ABC Listen episode independently opened and dated September14 6am; recording day may differ.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 14 Sep 2026.
Accessed 3 Oct 2026.
Source [2] · Supporting original statement
CNN Amanpour: Geoffrey Hinton interview
Attributed to: Geoffrey Hinton · CNN
Published
“I thought it was completely unrealistic.”
HINTON answer to SREENIVASAN asking why he did not sign the six-month pause letter · View original source ↗
Source context. CNN original rush transcript. China/benefits and inevitable development motivate feasibility objection. Does not supply CONTINUE. Preserve historical distinction with 2026 conditional superintelligence nondevelopment.
Scope
Specific March2023 proposed six-month pause beyond GPT-4.
Conditions
Rejects feasibility; instead recommends roughly equal resources for capability and control research. Not affirmative current-pace endorsement.
Stated rationale
CNN original rush transcript. China/benefits and inevitable development motivate feasibility objection. Does not supply CONTINUE. Preserve historical distinction with 2026 conditional superintelligence nondevelopment.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 9 May 2023.
Accessed 3 Oct 2026.
Source [3] · Supporting original statement
What if automating AI R&D triggers an intelligence explosion?
Attributed to: Geoffrey Hinton · arXiv / authors
Published
“First, policymakers should develop ways to pace and constrain scale-ups of automated AI R&D.”
PDF page7, Steering and constraining an intelligence explosion, first numbered policy component · View original source ↗
Source context. Name appears on first-page author roster; paper explicitly says views are authors own, not employers. arXivv1 datesSeptember28. This is coauthored support for tools and managed capability growth, not a standalone demand for an immediate all-lab halt. Benefits and safety R&D still valued.
Scope
Scale-ups of automated AI R&D and the risk of software-driven intelligence explosion.
Conditions
Governments should consider safety/input conditions, capability growth limits per period, verification of agreements, incident responses and targeted workload pauses; weigh abuse and costs of delayed progress.
Stated rationale
Name appears on first-page author roster; paper explicitly says views are authors own, not employers. arXivv1 datesSeptember28. This is coauthored support for tools and managed capability growth, not a standalone demand for an immediate all-lab halt. Benefits and safety R&D still valued.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 28 Sep 2026.
Accessed 3 Oct 2026.
Editorial assessment. Research proposal only; independent adversarial review and explicit human approval required. Contrary/scope evidence: He supports broader slowdown and pre-release testing as well as conditional nondevelopment of superintelligence. Risk estimates and regulation are not themselves PAUSE. Do not frame this as all-AI abandonment. September28 coauthored paper explicitly recommends pacing/constraining automated R&D scaleups and targeted workload pauses; it does not explicitly retract conditional superintelligence prohibition. Remaining limitation: Whether latest coauthored September28 paced-growth policy should shift representative category to PACE rather than explicit conditional PAUSE for superintelligence; reviewer must assess scopes, not assume retraction.