Physics of effective agents

(very speculative, armchair philosophy/physics)

I started writing this essay with a thesis that the strong, genius-type, embodied (0 resistance between mind and body - effectively a good relationship of mind to the body and vice versa) men are the scarcest type of men that the world needs the most and I somehow ended up speculating about the physics of effective agents. Oh well, that's what Universe literally made me write.

I am interested in how change is made in this Universe by living beings, which leads me to "life-force" driven agents. Agent is effectively a proactive controller of future states; it is an important distinction, because a thermostat is reactive.

Instead of just humans, as I became an adult, I started seeing agents at all scales: single cell organisms, plants, insects, animals, humans, institutions, [insert something you consider an agent here], states, civilizations and sometime in the near future AIs. If one were to count it all, that's a lot of agents.

Not all agents are made equal though - some of them are more effective than others. There are plenty of intelligent people, top 1% worldwide is 80 million people, the entire population of Germany yet not so many people are world-historical figures. Who knows, maybe one day Shinichi Mochizuki will be viewed as the most effective agent by the beings at the End of Time, but at human timescales, it is likely not the case.

Take for instance Genghis Khan, Napoleon or Elon Musk: they caused a lot of atoms to be put in a different arrangement, either approaching or reaching their desired configuration of spacetime.

Simply making atoms be arranged differently around is not enough though - without these guys, some events might have never happened or would have been be severely delayed - and that might have been good, but let's not get into morality and teleology here as that's beyond the scope of this essay.

Question remains though - what constitutes an effective agent?

Let's start from a simple example - two stable equilibria A and B, ball at point A, goal being point B.

What is the maximally effective agent?
An agent that gets the ball from A to B:

  • in the shortest time possible
  • utilizing as little energy as possible

Ok, that's classical mechanics 101, but where is the agent? To first approximation, agent is an entity with goals, capable of action: to reach them, it must possess intelligence which is the "ability to pursue goals at an appropriate level of scale" which requires a model of the world.

Ok, but it could still do better - if it had enough information that it takes less time and/or energy (depending on its internal optimization model) to drill through, it should do it.

Enough theory, let's summarize what an effective agent is:

  • it has goals
  • it possesses enough energy to achieve the desired goal
  • it is capable of action
  • it has a model of the world that allows it make to make "shortcuts" to reach that goal, ie. being aware of the path that minimizes energy. Some paths might seem impossible as the agent don't know about a shortcut!
  • robust: ideally it isn't harmed the action of reaching the goal and the failure to reach a goal is not catastrophic to it
  • does it within its lifetime

Effectively, the only difference between the an agent and an effective agent is its superior model of the world.

Translating this to humans, because that's the reason I started writing this essay: we humans operate on memes and some of us have high signal-to-noise mental models which allow for these "shorcuts" facilitating reaching many more goals and other people don't. The latter are called midwits.

Strong memeplexes help reaching goals much easier - these feel like a shortcut to a midwit. However, which of these memes prove to be strong in Darwinian sense can only be determined in retrospect and changing environment does not make i easier. Nevertheless, some of these memes are invariant.

I would like to formalize it a bit, put some physical units to it, but not today, I have to think way more and one day per idea is not enough.
Nonetheless, someday, someone will develop Hamiltonian mechanics for life agents.

PS. I realize I should have titled it anatomy of effective agents, but I don't feel like changing it now.

Things to ponder about:

  • Nietzsche would have said that an effective agent is the one that operates on instincts as opposed to Socratic theoretical man
  • Physics of effective agent in virtual world - is it any different? Well, if the physics engine for the simulated reality looks similar to our Universe than clearly not.
  • effective agent as the ones propagating viral and stable memes like Jesus Christ

Leave a Reply