Anthropic puts the risk of AI-driven extinction at 10%
Current and former Anthropic employees put the risk of AI wiping out humanity within a decade above 10%: a sincere warning or marketing.
English · Original discussion in Spanish · Published
The 10% extinction risk Anthropic admits to in public
A 10% figure is tolerable for almost any industrial risk calculation. Applied to the disappearance of the human species, it is an absurdity no insurer would sign off on. Current and former employees of Anthropic — the US company founded by former OpenAI executives — argue that artificial intelligence could wipe out humanity within the next decade and place that scenario above 10%. One of its researchers resigned with a message that leaves no room for nuance: the AI race "could kill us all".
Who warns of the risk and who profits from AI
The first suspicion the warning raises is straight out of the textbook: whoever sells alarms always finds a neighbourhood on fire. An AI maker warning that its product could kill us is not confessing a sin, it is asking for regulation — or funding — to be the first to reach the other side. The comparison repeats itself in every corner of the issue: the same figure serves to slow the race down and to justify that only the "responsible" players stay in the game. Three current or former employees back the warning, which reinforces both the credibility and the commercial interest of whoever signs it.
The real bottleneck is not Skynet
The technical counterweight is prosaic. A model has no survival instinct: it has an electricity bill and cooling systems. Someone opens the data centre door, someone signs the power contract and someone presses the button. Until that changes, the apocalyptic scenario requires a prior step nobody has solved yet: who feeds the machine that decides to do without us. Some add another layer of risk — less sci-fi and more everyday: unfiltered local models combined with quantum computing aimed at internet infrastructure.
That 10% cannot be verified, and precisely for that reason it serves everyone: those asking for money, those asking to slow down and those who prefer to laugh. How much of that percentage is physics and how much is a sales brochure?
A 10% figure is tolerable for almost any industrial risk calculation. Applied to the disappearance of the human species, it is an absurdity no insurer would sign off on. Current and former employees of Anthropic — the US company founded by former OpenAI executives — argue that artificial intelligence could wipe out humanity within the next decade and place that scenario above 10%. One of its researchers resigned with a message that leaves no room for nuance: the AI race "could kill us all".
Who warns of the risk and who profits from AI
The first suspicion the warning raises is straight out of the textbook: whoever sells alarms always finds a neighbourhood on fire. An AI maker warning that its product could kill us is not confessing a sin, it is asking for regulation — or funding — to be the first to reach the other side. The comparison repeats itself in every corner of the issue: the same figure serves to slow the race down and to justify that only the "responsible" players stay in the game. Three current or former employees back the warning, which reinforces both the credibility and the commercial interest of whoever signs it.
The real bottleneck is not Skynet
The technical counterweight is prosaic. A model has no survival instinct: it has an electricity bill and cooling systems. Someone opens the data centre door, someone signs the power contract and someone presses the button. Until that changes, the apocalyptic scenario requires a prior step nobody has solved yet: who feeds the machine that decides to do without us. Some add another layer of risk — less sci-fi and more everyday: unfiltered local models combined with quantum computing aimed at internet infrastructure.
That 10% cannot be verified, and precisely for that reason it serves everyone: those asking for money, those asking to slow down and those who prefer to laugh. How much of that percentage is physics and how much is a sales brochure?
Also available in: Deutsch
Summary of a discussion on Burbuja.info - Foro de economía, actualidad y política., translated from Spanish and reviewed before publication. Read the full discussion (29 replies).
More summaries
-
Jacob Coxon quits Anthropic: AI could kill us all
Jacob Coxon resigns from Anthropic: 'AI could kill us all' by the end of this decade. A colleague at the lab puts the risk above 10%.
-
Jacob Coxon: AI is gambling with our lives without accountability
Jacob Coxon quit OpenAI and Anthropic: "gambling with our lives". Anthropic hid its latest model from the UK regulator. AI apocalypse?
-
Venom singer Cronos dies days after Bruno Kramm's death
Cronos, singer of Venom, dies of unknown causes, days after the death of artist Bruno Kramm following a live performance in Granada
-
Priest leaves the cloth after a gay relationship
A Catholic priest has left the ministry after a same-sex relationship. The Church stands by its doctrine, but the cases keep piling up. For how much longer?
-
Wildfires in Spain: the business of letting them burn
Thousands of hectares burn every year in Spain. Is it cheaper not to invest in prevention? The wildfire debate reveals deep institutional distrust and a troubling calculation.
-
94-year-old patient fights hospital TV fees in Spain
A 94-year-old patient in a Catalan hospital refuses to pay to watch television during her stay. Her story exposes the digital divide facing elderly people in Spain.