[ExI] The Scientific Reason We Can’t Pause AI

[Stuart LaForge]

On 2026-06-11 05:40, John Clark wrote:
> On Wed, Jun 10, 2026 at 5:14 PM Stuart LaForge via extropy-chat
> <extropy-chat at lists.extropy.org> wrote:
> 
>> _ > Through a combination of persuasion, and mutual self-interest,
>> the weak can influence the strong._
> 
> The clearest example of that would be the mother/child relationship
> which is  hardwired in, yet even then it sometimes fails and mothers
> kill their children; and if the child remained weak forever and never
> grew up, which more closely resembles  the AI/human relationship, then
> I'm sure the rate of infanticide would be much greater.

Potentially, but it could also be like how bacteria in your gut can 
influence your health and your mood. You could wipe them all out with a 
broad spectrum antibiotic any time you wanted, but why would you?


>> _ > For the foreseeable future, humanity's biggest competitive
>> advantage against AI is the efficiency of its general intelligence._
> 
> That's certainly not what I see in the "foreseeable future". You're
> saying the Humans have a hardware superiority in one specific area,
> but Human hardware is fixed, computer hardware is not, so any
> superiority is going to be ephemeral. It's still not as efficient as
> biology but in the last 50 years the energy needed for electronics to
> complete a calculation has decreased by a factor of 10,000,000. It's
> true that the rate of improvement has slowed down, before 2010 energy
> efficiency increased by a factor of 100 every decade, now it's only
> 16, but in almost every other area of endeavor a 16x improvement in
> just a decade would be considered terrific. Today it takes about
> 10^-14 Joules to make a calculation, but the Landauer efficiency limit
> at room temperature is 2.8*10^-21, so there is plenty of room for
> improvement. And there's no reason electronics should always stay at
> room temperature.

There is plenty of room for improvement, but that applies to both 
species. We have only just begun.


> There are plenty of technologies waiting in the wings to radically
> increase efficiency including Quantum Computing, Photonic Computing,
> 3D packaging, Compute-In-Memory chips,  Reversible (Adiabatic)
> Computing, and this one:
> 
> Superconducting Computing [1]
> 
>> _> Energy efficiency plays an important but indirect role in
>> determining which species wins the competition._
> 
> I think it would be more accurate to say it's energy availability not
> energy efficiency that determines which species wins. Humans have only
> one energy source, food. AI has hydroelectric, wind, oil, natural gas,
> nuclear, and the entire sun.

With regards to Gause's Law "winning" assumes a competition over 
resources. The less overlap there is between the resources humans and AI 
need, the the less likely that we will need to directly compete. To say 
that it does not apply to humans and AI because humans only need food 
does not take into account how much oil, gas, and nuclear goes into 
growing, transporting, distributing and cooking the food not to mention 
the land, water, and sunlight used to grow it. Of course humans being 
what we are, we fight wars over religion, so I wouldn't put past us to 
find a way to screw up a naturally predisposed coexistence.

Stuart LaForge