That's part of remedies @Caffar3Cristina and I list in techpolicy.press/google-is-a-mo…:
* Disgorge all ill-gotten data, including downstream artefacts like models and search-adjacent ones.
* Limit retention going forward inverse-proportionally to market share.
I'm afraid we live in something like that world already! So much of the software systems we use are built based on exactly that kind of theoryless pseudoscience, in most cases the artifical friction is there and the "experiments" preferred to designers.
Excellent @cori_crider on the EU Google adtech case: "The only way to resolve the conflict of interest is to eliminate it."
We can't wait another decade to fix the problem of these toxic, now ecocidal companies. Democracy must show up and break them up.
politico.eu/article/google…
My sense is that if you're looking for alignment, you have to share some degree of lived experience. If your Markov blanket is kind of everywhere on the planet at once, that seems hard? Also, the texting interactions make sense based on IRL knowledge I think.