PeterDonis said:
That's what quantum field theory and the renormalization group are for.
...
Some references for where you are getting your understanding from would be helpful here. What you are saying does not look like anything in actual QFT. In particular, your description of "naked" vs. "dressed" actions seems wrong: that distinction has nothing to do with "the simple observer itself" vs. "including a part of the environment".
Sorry for the slow follow up. I was trying to describe the conceptual motivation for a more general measurement theory starting form post#58.
Yes my arguments and my points for why QFT paradigm is limiting can not be described from within QFT. My points are washed away from step 1, once you adopt the QFT paradigm as there is no such thing as inside observers/agents in QFT from which the full inference takes place.
Conceptually I see the "scaling" going on in renormalisation in QFT, is the observational "resolution" while keeping the observer/agent complexity itself essentially large enough to essentially by unlimiting. The physical observer situation this corresponds to, seems to be to the typical situation in high energy physics where you have a massive classical lab, and you just increase the energy of whatever you use to probe the target with. I suppose this is reasonable for it's original purpose but I think not for a background independent agent based model.
But the theory itself is always encoded in the essentially unlimiting environment which represents "the observer". So the context that provides the encoding capacity of the theory itself, is not scaled.
My objection to this scheme is that it fails to capture the inside view of a more general agent/observer becauase it would require scaling the complexity where the whole expectation is encoded, not just the "probe".
And it's from this inside view that I hope (from my interpretation and agent stance) that simplicity and more naturalness will be found. By naked or bare action, I meant the action as seen from the simple observer itself. The same "action" as described by an external observer, will necessarily come with an embedding that also will require more tuning, but which my be a fictional freedom. But we do not understand how to remove the fiction. The relation between this and the description of this agents interactin with other agents, as inferred from the perspective of third agent is I think necessarily more complicated than trying to scale the same mathematical model by only scaling the parameters. The theory will necessarily in the general caes involve new physics that can't be described jusy by scaling a fixed parameter set. Also gravity seems hard to renormalize anyway, so I think new physics is needed. The idea and motivation is that I think this will constrain theory space and reduce the level of fine tuning.
Mathematically the standard paradigm of the theory is based on a theory space which defines differential equations, and there is an initial value tuning required to explain the present. One "computer" does model everything as an initial value or boundary value problem. But can such a model learn and evolve and be applied to an inside agent? I think clearly not. I think the physica of the "computer" must be part of the game.
The different thinking tool for the foundations, that causes some of the difficulty may be that I think in terms of agent based models, instead of equation based models. A random reference on the notions, See
https://arxiv.org/abs/2107.03619 but these models are not that common in physics.
Many problems can be modeled both as system dynamics and as agent interacations, with pros and cons, but if one tries to understand QM as a theory of inference, the agent based model has an angle to this that seems better suited. Just like some people like "geometrization" of physics and has had tremendous success with it, one can see this agent-inference stuff as another trick. The end result will still be system dynamics, but as theory builders one needs some thinking tools to constrain the mathematics. There many inspirations about "physics from inference", but the earlist ones are more like entropic methods, but the more ai-style agent interactions are not very popular. There is
https://arxiv.org/abs/0808.1260 and there are various attempts to derive GR from entropic methods
https://arxiv.org/abs/gr-qc/9504004, which are of the former type and there are other idea like this
https://arxiv.org/abs/1712.01826. None are anywhere near the goal, but have common ideas. I'm not claiming anything here, just trying to add a perspective to the discussion on objectivit. I seem to be one of few here that represent this interpretation.
/Fredrik
