[ExI] alpha customer

Tue Nov 25 02:30:32 UTC 2025

From: Mike Dougherty <msd001 at gmail.com> 
Subject: Re: [ExI] alpha customer

On Mon, Nov 24, 2025, 8:34 PM spike jones via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org> > wrote:

>…I'm in no position to catch engineering mistakes…

Mine would have become clear anyway.  A typical approach is to solve a problem from multiple directions.  All the answers should come back looking pretty similar.  When you do that calc with 40 anythings dumpbing a kW each inside a metal box, things get hot quickly.

>…My intuition was that it would be a solution looking for a problem…

As I understand custom AIs, they take huge processing and power to train the AI, after which it is reasonable.  I might be wrong on this, and if so, do hipsterize me please.  

If there is a big power, processing and bandwidth demand to create a custom AI, we could imagine applications for the calculation station which might be able to compete with big guys like Gates, Ellison, China and Musk.

For instance…

The buzz is this year’s controversial hot Christmas toy, the ChatGPT enabled talking teddy bear.  Of course the sensationalism has been grossly exaggerated in the press, but after all the blather, there is a grain of truth: one might be a little reluctant to let one’s preschooler chat away on GPT.

However… one can imagine a specialty bot for that purpose, derived entirely from the children’s section of the library.  Its chatbot wouldn’t go where you don’t want your young larvae to follow.  It would be a kind and gentle bear, powered by the specialty bot… ChatG.  

As soon as a prole came up with the idea, this prole could set the calculation station to creating ChatG, without the risk that Gates, Musk or Ellison would steal the idea or the resulting ChatG.  It would of course be replaced by the risk that I would steal it.  But hey, three of those guys, one of me.  A quarter the risk.  And besides. I won’t steal the idea, sheesh.

I might try to derive a competitor.  But that’s different.  That’s just… capitalism!  

We would need to buy power at triple the cost going in, then still have a less efficient operation.

>…From thart starting point, your niche is in the realm of nefarious and illegal activities which pays the overhead from highly risky endeavors….

Ja, no.  Definite no-fly zone, but I get the concern: if this idea worked, bad guys will do it.

>…  You gonna truck in mobile hacking rigs to break nearby networks?  I think this is interesting guerilla cyberwarfare... but you're proof of concept is is white-hat until purchased and used elsewhere in the world,  yeah?

Ja, white hat always, but after I did the calcs, it is a no-hat.  There is a lesson in all this, Mke.  Read on please.

>…Do you think this entire approach to EV is viable?  …

I now think not.  But this is an area in which I claim no particular expertise.  Back in around the 1995-2002 timeframe, it was fashionable to opine that AI and the singularity would not necessitate enormous compute power: the magic was in the software.  With the advent of ChatGPT and the other bots, I think that view has diminished.  

I wouldn’t say the probability of getting AGI first is proportional to the processor power available, but I would now argue those two are generally correlated.  If so… then this exercise has its value.  Developing this idear convinced me that the first to AGI will be one of the biggies.  If ranked by probability, I would estimate possibly in this very approximate order: Musk, Gates, NVidea, Ellison, China Inc, Apple, Thiel, US military, Bezos, Samsung, Zuckerberg, spike.

OK, granted there may be a few other names between me and Zuck.  We shall see.

spike

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20251124/e1be1334/attachment.htm>