Okay I think I understand what you're trying to say. So let's go back to our hypothetical observer outside the universe, looking in at the source code. (Not that this is possible, but I find it clarifies my thinking and what I'm trying to say.) The human is looking at an oak tree. The observer is looking at the human's brain, and sees that certain neurons are sending signals to other neurons. The human is closing their eyes and visualizing an oak tree. There's a similar but not identical neural pattern going on–I find the subjective experience of visualizing an oak tree using my imagination isn't quite the same as the experience of looking at one, but the neural firing is probably similar.

If you don't mind I would like to play off your analogy. I agree that the arrangement of neurons will not be identical, but I would pose the question how does the observer know that the human is closing his eyes? When he is looking at the tree perhaps there is wind and a feeling of coolness; but when he is closing his eyes it can also be windy and cool. If there is a lack of wind in one model how does the observer know that the neurons are the result of a mental construct and not the result of looking at a tree through a window while sitting inside?
The way memories/ mental images work is that they are networked. When we recall a past memory we are irrevocably altering it by attaching it to our current consciousness. So for example, let’s say when I am 20 I remember an exploit of my early teens while at a sleep over drinking vanilla soda. The next time I go to recall that memory, I will also unintentionally, and unavoidably, activate memories of that sleep over and vanilla soda. Every time I reactive that memory the soda and sleep over get activated too, strengthening their place in the memory. In another 10 years the two memories are indistinguishable. Back to our observer, when he is looking at me thinking about an oak tree, it irrevocably activates a network of sensory experiences that will not identically replicate the reality of the oak tree in front of me, but will present an equally believable reality of some oak tree. Not the same oak tree, but I would suggest that the workings of the human brain are more complicated than what you have imagined. Where the observer would not be able to tell the real from the construct.

For me there's something basically wrong with claiming that something doesn't exist if no human being knows about it.

I have no disagreement here. It is wrong to claim that something doesn’t exist if humans do not know about it. What I have been arguing about is not ontology (what does or does not exist), but epistemology (how humans come to know). It is not that I am saying no territory exists outside of what is human, but that humans have no other way of knowing territory besides through human means, linguistic means.

Do you think research in hard science will stop advancing, or that is should stop advancing? If so, why?

It is not that I think scientific research will or should stop, but that it should be moderated. What is the purpose of science-technology? I understand the purpose of these things to help humans to be able to better know and predict their environment for the sake of creating a safer niche. Is this what the scientific institution currently does? I would argue no. Currently, I see the driving impetus of science and technology to be profit. That is not a critique in anyway of A.I or the projects of this community. To the contrary, I think the motives of this group are exceptions and exceptional. But I am talking about the larger picture of the scientific institution. The proliferation of new technologies and sciences for the sake of profit has rendered the world less knowable to people, harder to predict, and no in some sense more dangerous (when every technological victory brings with it more sever problems). I am not against hard science. I am against the overemphasis of this one technique to be superimposed onto every facet of human reality.