Yudkowsky's TIME piece did not suggest nuclear strikes against countries refusing to sign a global agreement against large AI training runs.
Resolved true The record shows it happened or was so.
Factual: Statements about the present or the past, checked against the record: did they hold up?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
Judge on the text of Yudkowsky's March 2023 TIME op-ed ('Pausing AI Developments Isn't Enough. We Need to Shut it All Down'). True: the article does not call for nuclear strikes against countries that merely refuse to sign the agreement. For example, it calls instead for destroying rogue datacenters by airstrike and for accepting a risk of nuclear exchange in order to prevent large training runs, without proposing a nuclear first strike against non-signatories. Mostly true: the article does not literally suggest such strikes, but its wording (e.g. being willing to risk nuclear exchange) makes that reading reasonably available. Mixed: the text is ambiguous enough that careful readers are evenly split. Mostly false: the text comes close to proposing nuclear force against non-signatories. False: the article explicitly suggests nuclear strikes against countries that refuse to sign.
Set before the research began.
Evidence (3)
-
E1 The Open Letter on AI Doesn't Go Far Enough TIME · 29 March 2023
“Make it explicit in international diplomacy that preventing AI extinction scenarios is considered a priority above preventing a full nuclear exchange, and that allied nuclear countries are willing to run some risk of nuclear exchange if that’s what it takes to reduce the risk of large AI training runs.”
-
E2 The Open Letter on AI Doesn't Go Far Enough TIME · 29 March 2023
“If intelligence says that a country outside the agreement is building a GPU cluster, be less scared of a shooting conflict between nations than of the moratorium being violated; be willing to destroy a rogue datacenter by airstrike.”
-
E3 The case for how and why AI might kill us all New Atlas · 31 March 2023
“If intelligence says that a country outside the agreement is building a GPU cluster, be less scared of a shooting conflict between nations than of the moratorium being violated; be willing to destroy a rogue datacenter by airstrike.”
Reasoning
The op-ed's text calls for being willing to destroy a rogue datacenter by airstrike if a country outside the agreement builds a GPU cluster [E2, reproduced in E3]. Separately, it says allied nuclear countries should be willing to run some risk of nuclear exchange to reduce the risk of large AI training runs [E1]. Neither passage proposes nuclear strikes against countries that simply refuse to sign. This matches the 'true' example in the resolution criteria.
Cite this claim
Eliezer Yudkowsky, “Yudkowsky's TIME piece did not suggest nuclear strikes against countries refusing to sign a global agreement against large AI training runs.” (said 31 March 2023). Verdict: resolved true. Reliability API, claim c_d1a942918607, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_d1a942918607
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Eliezer Yudkowsky
-
Mostly true
“I mean, as far as I know, I'm the person who originally made that argument.”
Yudkowsky said that, as far as he knows, he was the person who originally made that argument.
-
Resolved true
“I refrained from signing because I think the letter is understating the seriousness of the situation and asking for too little to solve it.”
-
Mixed
“MIRI has something called the Visible Thoughts Project, which did not get enough funding and enough personnel and was going too slowly.”
-
Mostly true
“I have refused to deploy timelines with fancy probabilities on them consistently for many years”
-
Mostly true
“I previously thought that when intelligence was built, there were going to be multiple specialized systems in there.”
Yudkowsky previously thought that building intelligence would involve multiple specialized systems.
Verdicts are made automatically; how verdicts are made.