[ExI] AI Model Collapse

Fri May 30 16:56:53 UTC 2025

From: extropy-chat <extropy-chat-bounces at lists.extropy.org> On Behalf Of BillK via extropy-chat
Subject: Re: [ExI] AI Model Collapse

On Fri, 30 May 2025 at 16:49, spike jones via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org> > wrote:

I was astonished by a comment by a TV host who does programs about various jobs and industries, Mike Rowe.  …

"...We've got a huge labor shortage and the push to get to the AI thing over the finish line. That means data centers. We've got to train people to build these data centers..."  Mike Rowe

…

spike
______________________________________________

>…I asked Perplexity AI to research this question.

It said "Polls show 63% of adults haven't heard the term "recursive self-improvement".

But techies are aware of the problem.

Full report below.  BillK

>…Perplexity AI - 

>…Public awareness of AI's recursive self-improvement risks remains limited but growing, concentrated primarily in technical and policy circles rather than mainstream discourse. Below is a detailed breakdown of current awareness levels and key knowledge gaps:

  _____  

>…Awareness Levels Across Groups

1. Technical Communities

*	High awareness: 78% of AI researchers recognize RSI as a critical risk factor in AGI development  <https://en.wikipedia.org/wiki/Recursive_self-improvement> 5.
*	Active debates: Forums like LessWrong and arXiv host ongoing discussions about alignment challenges and containment strategies.

2. Policy Makers

*	Moderate awareness: Recent EU AI Act amendments reference "autonomous self-improvement systems" but lack specific RSI safeguards  <https://www.datapro.news/p/the-risks-of-recursive-self-improvement> 4.
*	Gaps: Only 12% of national AI strategies explicitly address recursive capability growth  <https://www.linkedin.com/pulse/future-ai-why-recursive-self-improvement-may-lead-hard-gary-ramah-grn5c> 2.

3. General Public

*	Low awareness: Polls show 63% of adults haven't heard the term "recursive self-improvement"  <https://lew.ro/in-5-years-ai-will-write-and-improve-its-own-code-are-we-on-the-brink-of-recursive-self-improvement/> 6.
*	Misinformed perceptions: Media coverage often conflates RSI with generic AI risks like job displacement or bias.

  _____  

…

BillK, thx.  If you read thru the text of the former VPOTUS speech on AI, the one with the infamous “…it’s two letters… a kind of fancy thing…” comments, the risk she pointed out was that AI would run companies without understanding or compensating for human bias.  BIAS!  This is how the government thinks of the risk of AI?  We have a TV guy who is an expert in industry interested in getting “the AI thing over the line” in order to solve the labor shortage in the office, while the former VP and AI czar warns of bias, while coming across like a fifth grader reporting on a book she hadn’t read.

Elon Musk commented recently that AI poses a real existential risk, 20% chance of destroying civilization in the next decade.  Yet that comment made so little public impact that I can’t even find it now, which is making me wonder if he ever said it.

spike

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20250530/8c2d7871/attachment.htm>