I can't think of a large successful -ism that doesn't have that property.

Similar to particle-wave duality?

One reason a lot of sci-fi and AI safety researchers did a relatively good job at predicting the future we see now is a lot of it is the same problems we see emerge in biological systems. Free rider problems as a means to save energy expenditures, different versions of game theory and stag hunt. How to bake in rules that apply to one society (or part of it) but not another. We like to think of these things as stable in human scale systems, but they are not at all. Things can go from hunky dory to your neighbors stabbing each other in the streets in mass revolution very quickly.

Worse these AI systems are not in a universal island just affecting themselves, what they do affects us, what we do trains them and as the rate of progress continues to accelerate social structures are going to further destabilize (and they are already rapidly changing and strained). It is very likely we are going to see a world order rearrangement soon, much like the rapid changes in the early 1900s brought.

I'm not sure anyone meaningfully understands it: "Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens" https://arxiv.org/html/2505.13775v3

> More interestingly, our experiments also show that models trained on corrupted traces, whose intermediate reasoning steps bear no relation to the problem they accompany, achieve performance largely comparable to those trained on correct traces.

My understanding is that the extra token vectors generated as reasoning are still useful, but that their surface form (tokens themselves) do not necessarily reflect the underlying reasoning. i.e. reading the reasoning traces could be complete gibberish, but the hidden-dim vectors themselves still refine the latent probabilities and help in generating the correct answer.

Not an expert in LLMs, but this seems supported by the abstract of the paper cited in the above article:

  it remains unclear to what extent these performance gains can be attributed to human-like task decomposition or simply the greater computation that additional tokens allow. [...] our results show that additional tokens can provide computational benefits independent of token choice. The fact that intermediate tokens can act as filler tokens raises concerns about large language models engaging in unauditable, hidden computations that are increasingly detached from the observed chain-of-thought tokens.

https://arxiv.org/html/2404.15758v1

And you best control the birth rate by royally messing up the economy. Do you know how bad things have to be, for a mammal to voluntarily decide not to reproduce?

>he quietest way would be to control the birth rate and sterilise the population gradually,

Honestly we are doing this pretty well without AGI. Nearly world wide the birthrate has fallen below replacement rate. In places like Japan and Korea these are already critical problems in the medium term.