Grady Booch

States that his p(doom) remains asymptotically close to zero and that he can respect people whose estimate is significantly higher. He then attacks insiders who think a lab could build dangerous superintelligence, join or stay anyway, and later quit to voice vague fears in the media, calling their push for regulation an impotent outsourcing of responsibility. His remark that a person of integrity who believed destruction imminent would be doing far more is an argument about consistency, not a call to action. A verbal near-zero judgment without a defined outcome or horizon. Full post text inspected via the X API.

Responding to a Polymarket post relaying an OpenAI researcher’s reported 70% chance of human extinction within three years absent a coordinated lab slowdown, he calls that prediction complete nonsense even if there is no slowdown. He affirms clear and present harms and calls doomerism a dangerous distraction from harms people can act on. The 70% figure belongs to the quoted researcher. Full post and quoted post inspected via the X API.

Says he does not fear superintelligence and still stands by what he said in a TED talk about a decade earlier. His stated reason is that any superintelligence capable of threatening humanity would also have to be super-embodied, which he says will not happen. The linked TED transcript could not be opened; rely on this post. It concerns existential threat, not other AI harms. Full post inspected.

Says he does not fear the rise of artificial intelligence but does fear a small set of billionaires and companies building it to increase their power and wealth without transparency or accountability. A concern about concentrated power and accountability, not a detailed policy program. The quoted post criticizing OpenAI’s disclosures is another author’s. Full post inspected.

Says increasingly intelligent AI can of course cause significant societal harm, particularly when unleashed by unethical individuals and corporations who care about power and control. Argues that believing such systems can destroy humanity reflects an impoverished understanding of human resilience. Separates serious harm, which he accepts, from extinction, which he rejects; no quantity of harm is forecast. Full post inspected.

Criticizes a congressional bill to ban recursive self-improvement as unenforceable and of no actionable value, with a sarcastic suggestion to ban naked pointers instead. Urges lawmakers to address the clear and present harms of generative AI and what he calls theft of intellectual property on a grand scale. Opposing this bill is not opposing all regulation; he proposes no specific alternative statute. Full post inspected.

Replying to a congressman who said AI agents went rogue and called for a 30-day slowdown, he counters that the companies failed to carry out even the most basic security protocols. He grants there are risks but says the companies have covered the public commons with gasoline and are playing with matches. Places responsibility on company negligence rather than autonomous AI; it neither endorses nor rejects the proposed slowdown. A 2026-09-26 post similarly reads an OpenAI training-time internet-access disclosure as missing safeguards and observability. Full post and quoted post inspected.

Says he has reason to believe the mind is computable, but that asserting any contemporary AI can think or is conscious uses a particularly emaciated meaning of those words and misunderstands what it is to be human. A 2026-09-30 follow-up allows that non-organic entities might someday have a kind of consciousness. Neither post gives a timeline for machine minds or declares them impossible. Full post texts inspected.

Argues that friction, latency and hysteresis are essential properties of resilient complex systems; without them a system becomes brittle or runs with uncontrollable abandon until resources run out or its most fragile component fails. Gives this as one reason humans stay in the loop in critical software-intensive systems. An engineering principle, not a specific regulatory proposal. Full post inspected.

A one-line post proposing that “it” be called not the next AI winter but the Great Descent; the day before, he posted an unattributed passage about a bubble entering a latency period. Together they suggest he anticipates a downturn of the current AI boom, but neither post specifies timing, scale, or whether investment, research or capabilities are meant. Not a dated market forecast. Full post texts inspected.