[ExI] reason on data centers

[spike at rainier66.com]

From: Keith Henson <hkeithhenson at gmail.com> 
Subject: Re: [ExI] reason on data centers

 

On Tue, Sep 22, 2026 at 8:04 PM <spike at rainier66.com <mailto:spike at rainier66.com> > wrote:

 

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

Keith, it is comedy with a point.  Of course it is mixed with silliness, but that is Heaton’s style.

>…I did not find it funny at all. Reminded me of the very worst commercials on TV back in the 1950s…

 

OK no worries, I never saw commercials in the 1950s.  I hadn’t been born then.  But even if I had been, my folks didn’t have a TV until 1965, so I missed half of the 60s.  

 

>…Probably true. One interesting gotcha is that the Chinese may crash the entire AI buildout in the US by releasing free versions that are good for many uses but take a small fraction of the compute power. All of a sudden we will have outmoded data centers that gobble oceans of power and lakes of water and have no income to support them….

 

Ja, I too have thought of that as well as other failure modes for the current bubble, which is why I sold out of my Nvidia recently.  Keith something like that happened before: when Deep Blue defeated Kasparov in 1999, IBM immediately disassembled it.  Reason: it took a lot of power.  Only 14 years later, grandmaster level chess could run on a cell phone and could finish a game on the battery power in the phone, no charging cord.  IBM didn’t want to risk damaging its mainframe biz once we realized we no longer needed mainframes.

 

>…Something like is happening to Microsoft when whole countries switch to Linux. That crash could outdo the dotcom crash by 50 or 100 fold….

 

We are pouring astonishing amounts of capital into the idea.

 

>…I wonder if they could be repurposed as shelters for the homeless? If this happens, there will be a lot more homeless people who lost their pensions in the crash…. Keith

 

Anyone whose pension is heavily dependent on AI has an incompetent fund manager.  You can have that component in there, of course.  But it is a longshot bet.  We know this whole data center buildout is risky as all hell.  Just to be sure, everyone here, John, Ben, BillK, the resta yas, you do realize the current wild AI buildout is risky, ja?  I don’t mean the risk it will be successful, go malicious and kill everybody.  I mean the bigger risk that it will fail and actual money will be lost.  How shall we cope?

 

At some point, we should all sober up and estimate the risk that AI will go Eliezer’s If Anyone route.  Think about it, what do you consider the risk of AI destroying humanity in the next decade?  I will offer my estimate: about 20%.  Estimate, subtract from 1.  I am getting about 80% chance it will not kill everybody.  But at some point one must break down the 80% into plausible scenarios, estimate the probability of each path, then figure out how to invest so as to make a buttload if one is correct.

 

spike

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260922/dc4e44bb/attachment.htm>