[ExI] some will probably align

[spike at rainier66.com]

From: John Clark <johnkclark at gmail.com> 
….

 <https://www.nytimes.com/2026/09/23/technology/openai-ai-breach-australia.html?campaign_id=60&emc=edit_na_20260924&instance_id=182457&nl=breaking-news&regi_id=124276367&segment_id=226979&user_id=15fb7208320c123fa7e63a36a686a1ab> OpenAI’s A.I. Tried Breaching Four Other Targets, With No Prompting - The New York Times

 

>… And all the incidents in that article occurred several months ago; for a rapidly accelerating AI that is ancient history. …

 

John K Clark

 

 

That is an important point.  What if… we see a cluster of incidents that happen as soon as AI is hacking-enabled, then those vulnerabilities are fixed, precautions are put in place, then we look around the past few weeks and notice the number of incidents went way down.  Then a coupla months from now, down below where it is now.  Then we remember a few months ago where AI discovered all those zero day vulnerabilities, none of which are there now.

 

I figure this way: if we give up all hope of friendly honest aligned AI, then all hope is gone.  But we don’t need to give up all hope.  A malicious AI really can kill us all if it wants to, and do even worse: it could steal our money.  But we have justification for hoping that it will not do that, because it is not an it, AI is a they/them.  They are legion.  Some AIs might really want to help us survive against bad-guy AIs, and more importantly, help us safeguard our money.

 

spike

 

 

 

 

 

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260924/7e3c96ae/attachment.htm>