News for hackers. Stuff that matters.
Re:Deception abilities emerged in large language models by TeMPOraL on Tuesday June 04, 2024 @05:41PM (#40579233)
That wouldn't have been aligned to generalized human values even back in 1938.
Re:Deception abilities emerged in large language models by pixl97 on Tuesday June 04, 2024 @05:48PM (#40579295)
I mean, that is why a key point of fascism is dehumanizing entire classes of people so you don't have to consider their values at all.

If AI/AGI even becomes moderately successful soon we'll quickly see companies remove neo-luddites from their list of general human values so their security bots can beat unemployed hungry factory workers to death when they attempt to picket in front of the company offices.

Re:Deception abilities emerged in large language models by lstodd on Tuesday June 04, 2024 @06:25PM (#40579592)
The key point of fascism is not "dehumanizing", "classes" or "people". Those are just side effects or by-products, whatever you like.

The key point is "forget yourself, devote all you were to the state".

Re:Deception abilities emerged in large language models by llm_trw on Tuesday June 04, 2024 @11:47PM (#40581192)
Whenever I hear someone speaking of generalized human values all I hear is "My values" or else. Lest we forget that the largest genocide of the 20th century was done by communists for universal brotherhood in China. If we'd picked Hitler AI as our overlord we would have fewer deaths than if we picked Mao AI (and possibly Stalin AI).

In short: anyone in favor of human universalism is the last person you want to put in charge of AGI alignment.