Emergence in A.s.a. H. physical AI experiments

An A.s.a. H physical AI learns/creates a hierarchically organized concept structure (like the simple early example in my 4 September 2024 blog). Concepts higher up in the structure "emerge" from those lower down. Frequently recurring spatiotemporal patterns may be learned/discovered that involve only concepts that are defined low down in this hierarchical structure.* Other recurring spatiotemporal patterns may be learned that involve concepts that are defined higher up in the concept structure**, these can be considered to be "emergent."   

* Things like "bigger forces are accompanied by bigger accelerations".

** Things like "collisions are accompanied by reduced health". 

Programmable Logic Controllers

Most robots* are controlled by PLCs. I have tended to avoid the use of PLCs** since I haven't wanted to use ladder logic and PLC hardware and software tends to be expensive.

* industrial

** with the exception of Controllinos 

Turbulent thermal insulation in inertial confinement fusion

 Vincent et al* report the observation and importance of turbulent thermal insulation** in high beta plasmas.

* Experimental observations of microturbulence-suppressed parallel heat conduction in weakly collisional high beta plasma, arXiv:2610.08530v1, 6 October 2026

** See R. Jones, Preliminary experimental observations of turbulent thermal insulation, Plasma Physics, vol. 22, pg 753, 1980.

Spacex Starship

What's needed is for a Starship* to be recovered and reflown. We would then want to know how much refurbishment went into flying again and what the weight of the thermal protection system was. These would determine if the system was economical or not. Longer term, how many times can the ship be reflown? At what cost?

*i.e., the second stage 

Issue with LLM generated code*

When I was a professor I always found it harder to debug code written by my students as compared with debugging code I had created myself. This is now even more of an issue when I try to debug code created by an LLM**. 

* See my blog of 1 January 2025.

** No, coding is not "solved." That's the capitalists lying again. There are still lots of bugs to find and fix.

Tech

I am glad to see the effort and resources currently being spent on A.I., robotics, and space travel. I don't expect that these efforts will produce anywhere near what's being promised of course. Each of these efforts is currently too narrowly focused, they should all be more broadly based.

LLM slowdown

The "AI slowdown" is not caused by a coordinated reaction to safety concerns*. That's being used as an excuse. Rather, it is due to the terribly weak scaling law for LLMs. This was foreseen years ago; Ilya Sutskever's "end of the age of scaling" and return to another "age of research."**

* really they're doom trolling

** See also my blog of 1 January 2024.