[ExI] OpenAI Reaches A.I. Agreement With Defense Dept. After Anthropic Clash

[spike at rainier66.com]

From: spike at rainier66.com <spike at rainier66.com> 



 

>… they don't have an AI better than Claude ….

 

A battlefield AI doesn’t necessarily need to be better than Claude.  It does need to be loyal to whoever fired it however.  It must be completely trustworthy to protect that guy.  The warfighter cannot be doubting his own weapons, and his weapons cannot be second-guessing the motives of the guy who fired it.

 

We tend to think of the current AI arms race as nation vs nation, as has been the tradition with super-weapons.  However this time around, there is a subtle and important distinction.  There are more players than there are nations.  It isn’t only China vs USA vs Russia.  Now, whoever gets the first superhuman AI which can write and optimize its own code owns the planet, or that SAI does.  If the human who developed it can get it to be loyal to him, then he owns the planet.  The winner can be a government, or it can be Sam Altman, Dario Amodei, Elon Musk, or anyone else we never heard of because they are working in secret.  Eliezer’s nightmare is playing out in real time, right now.  The solution he proposed a decade ago would not work then, it will not work now nor in the future.  We don’t know how this will end.  It isn’t even clear who to cheer for.

 

spike

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260301/d66f6777/attachment.htm>