On the danger of seeing double

The previous post on "seeing double" is incomplete. Actually it went astray when I followed my stream of thought and got carried away. What I was trying to say is that, with the "next level" thinking becomes a norm, a social habit, it gets much harder, not easier, to keep an empricial view of life in general.

The problem can be stated two ways. For one, consider someone who is getting used to "seeing double" when using LLMs, and has established a formal way of using it, not in the natural conversational way, but more like handling a machine with sets of parameters and indicators. Let's just say he/she is building a certain template for certain purposes (i.e. writing a script for a video, doing an industrial research, etc.) when using LLMs. But we should be aware of a simple fact, that is, we really don't have that much tasks that need to be automated, unless we are living not for ourselves but for many many other people. A Youtuber with 100k+ followers belongs more to his fans than to himself, and is more appropriately called the "influenced" than the "influencer." It is as if his/her life is split into 100k+ pieces to be exchanged with each and every follower's attention, so that he/she can no longer do something simply out of personal desire, but has to consider how his/her audience would respond. This is the precise moment when he/she decides to automate certain processes during his/her production, to de-personalize his/her output and to be related by the audience. You simply cannot deal with 100k+ followers one by one, although sometimes you are faced with such demand, and what LLMs are offering is a cyber double who can interact with your followers in your stead. But now that you are not dealing with any single follower personally, you are stuck with your digital doppelganger and have to deal with it instead. You start to be seeing double. You start to focus not on the original reactions from your followers, but the interaction between your agent and your followers. The agent cannot be responsible for everything it says, and neither would you want to be held responsible, so the only way is to finetune the agent regularly, not only by updating your bibliography, but also by trial and error with all the "parameters and indicators," and the job itself becomes more and more mechanistic, dehumanized.

This is the opposite direction of "personal" writing, the latter of which is what I refer to as "writing too personally." Because nobody is interested in your total life as you do, simply writing everything that concerns you is just pushing your potential audience away. For example, I wouldn't write about indie rock music, or contemporary gin, or the history of the Chinese Communist Party, or Andrei Tarkovsky's films here, even though I am very interested in all of them, because I want to make this blog just a slice of my life which has a steady focus on certain topics, so as to stretch myself further in that direction. But by doing so I face the danger of seeing double, because I now am not living fully as the person that I am, I am restricting my being on these topics, and I am focusing not only on these topics but also the reactions I might get, the audience per se.

The second way of seeing it also stems from the fact that we really don't have that much tasks that require automation. In a previous post I distinguished between a skill and a habit. A habit is some brainless repetition that does not and need not grow. A skill is something that grows on you constantly. The automation only works for habits, not skills. You can and should only automate those activities that you don't expect to get better at. For example, a PE teacher can turn some sports into brainless repetitions because he/she does not care about his/her own performance but only use these sports as educational tools. But an athlete of the sport can never do that. The athlete always consciously adjusts his/her motions to get better and better. The skill fo the PE teacher is not sports but education, so what he/she is consciously adjusting is just the educational moves he/she does to the students. The skill (education) is the end, the habit (sports) only the means. So the problem is, unless we have some very specific skills we really want to improve, and unless such skills include a lot of brainless repetitions that can be outsourced, either to AI or more commonly to the market, the larger part of our time is always devoted to the improvements of our skills, not the adjustments of our habits. But the current AI scene is luring us into thinking that we can get the most of it if we simply do the opposite, to outsource our daily tasks as much as possible to LLMs. This comes the real danger of seeing double: that we really forget how the two are interconnected, how, by outsourcing the habit, we actually lose the skill altogether, and, finally, how, by losing the habit and the skill, we lose the core of our being, the identity.