Although LLMs are not deterministic in normal operation (because of the temperature values used) they are still strongly biased in their outputs. Bias in their outputs is literally how they function.

So you get a different type of conformism where people who interact with LLMs a lot are going to be influenced towards the median, as encoded in those token probability look up tables.

The effect is basically the same as what Heidegger and Illich were worried about, and I think we're seeing it happen already everywhere.

A solution(?) is to try to be your true weird authentic self when you act in the world, rather than bending your words and behaviours to the machine and LLM probability shape.

LLM force humans to think like machines far more than just writing code.

Writing code is a creative activity with large number of degrees of freedom and opportunities to express yourself. Button pushing and approving code changes is a deeply limiting activity that alienates you from your work.

LLM's are probabilistic but with a very limited range. They are tuned to be predictable enough to serve as a consumer product, so most ideas you get from them will revolve around the same concepts repeatedly.

Unfortunately, mere nondeterminism still admits unsatisfactorily constrained output distributions. Training these models imbues them with certain inherent attractors, some good, some bad.

We see some of these attractors across most or all frontier models due to similarities in datasets and training methods. We are seeing a digital manifestation of Jung's collective unconscious. And we have seen just how reinforcing archetypes are in human behavior over time. Sometimes it's good, sometimes it's pretty bad.

For one, it has turned us into socially programmable machines who can be manipulated quite easily by associating desired thoughts and behavior with certain archetypes: the patriarchal/matriarchal nature of US politics is a great example of this. The same evolutionary predispositions which causes children to listen and adhere to their parents is used to cause adults to listen and adhere to their politicians.

Given the nature of the core datasets these models are trained on, we should assume they are also vulnerable to such archetype-coded social manipulation, and this is troubling in a post-singularity world: When raising domesticated rats, I learned that I had to be careful what I fed them. Rats in the wild are careful about eating new substances. They will nibble and wait a while, see how it affects them or other rats, before committing to larger bites. Domesticated rats, however, will often blindly eat anything you feed them because they have learned to trust you as a safe source of food. We are going to have this exact relationship with our digital agents.

Ok this fascinating. I happen to have some Heidegger files, and went a-greppin. I'd say he's really talking about "inaminate objectness", as opposed to our human animate subjectness. Reasons:

1. In Science and Reflection, he says that both the 'completely calculated' nature of classical mechanics and the 'statistical character' of quantum mechanics are two representation of an underlying 'objectness of material nature'.

2. In Discourse on Thinking, he's far more concerned with the instrumental, utilitarian nature of machines than literal determinism, a concept he (somewhat confusingly/uniquely) connects to computation/'computational thinking', as opposed to the inherently-human capability for 'meditative thinking'. I personally would say this bit is refuted by LLMs in the end, but because they have intuitive rationality, a capability that seemed impossible to artificially design to the scientists of the time -- not because they involve dice.

3. In What is Thinking?, he gives the closest connection of this dichotomy to one of his better-known main topics, namely [handi]craft as a special subset of work: 'what maintains and sustains even this handicraft is not the mere manipulation of tools, but the relatedness to wood.' In this sense machines would be merely manipulating tools, stochastic or otherwise.

4. Overall I get the sense from skimming that lecture that his concern is a little mischaracterized, and in a kind of beautiful way that backs up my "Dune is still about AI" thesis described in a top-level comment somewhere below: his concern is about any society that could build complex machines (a 'technological' one), not the behavioral aspects of the machines themselves. So that concern applies equally to Star Trek (AGI), 1984 (no AGI), and our current post-singularity society (it's complicated/potentially ""just"" guessing).

All of this fits very well into his central thesis, which is continuing the Kantian & Hegelian tradition of examining how humans might be moral, free, and spiritual despite being meat computers of a largely-unknowable nature. In Heideggerland, this is expressed best in dramatic, nigh-unreadable passages like:

  Temporality already maintains itself ecstatically within the horizons of its ecstases; and in temporalizing itself, it comes back to those entities which are encountered in the "there".
  With Dasein's factical existence, entities within-the-world are already encountered too.
  The fact that such entities are discovered along with Dasein's own "there" of existence, is not left to Dasein's discretion.
  Only what it discovers and discloses on occasion, in what direction it does so, how and how far it does so—only these are matters for Dasein's freedom, even if always within the limitations of its thrownness.