This doesn't seem quite right, or at least underspecified. We can talk about this stuff concretely these days, at least in the context of digital systems. E.g. i can draw up a diagram of a system that takes in some camera and audio data (and tactile, proprioceptive, etc.), tokenizes it then runs that + past state data through some autoregressive VLM to drive an inference process. The state being passed around can be written out analytically for a given trained model - the external and internal environmental representations, the linear algebra that transforms them into latent action representations, the process by which that is transformed into control signals. It seems difficult to claim that the computational process that implements this has any more or less of a gestalt then one multiplying two matrices together. So it's not just the existence of certain representations or computational loops that seems to lead to possessing a gestalt.

If one wanted to try that, it would likely take the standard LLM finetuning route - scrape a bunch of data pairs at some expense, write some prompts, and have at it.

It's also a more natural question to ask, since building projections on top of frozen foundation model embeddings is both common in an absolute sense, and much more common, relatively, than building projections off of tiny frozen networks like a ResNet-50.