Facebook/meta was very early in the “AI” space. They built lots of tools and models being widely used right now. Just look at their popular github repos, faiss, SAM, vision transformers, visual embeddings, core ML libraries, all very essential components of many ML systems.
Saying facebook “just” dove into AI is wrong, facebook has been there from the early days along with Google (and probably some more) in building the very foundations these stupid LLMs “thrive” on
The readings will have an extremely low signal to noise ratio, at best you coils differentiate a truck from car by length of the pulse, but with speed already being ±30% variability any small difference in cars inherent properties (size in length or amount of metal) would quickly be reduced to nothing. Then for every make and model you have thousands of cars already, I don’t see how you could possibly differentiate between those even in theory.
Gtk 3->4 made a lot of internal changes, and at least some were related to making wayland work. Wayland “worked” in gtk3, however it was very much an afterthought, and half the toolkit was useless under wayland. Other changes are usually required for changes related to rendering, gtk4 had vulcan rendering which may require some breaking changes. Another thing is just general breaking changes that are good, sometimes you realise some decision was bad, and a new major release is just a way to make these.
From the end users perspective nothing much changes, it maybe looks a bit different, but not much besides that. But a vulcan renderer and being fully wayland compatible are major improvements that also improve the user experience, even if you don’t notice directly.
A couple of months ago I downloaded a 1970/80s movie that was so obscure only private trackers had it. I am virtually the only seeder, my radio is currently 1.8 which is the highest I have. (Might not be high for others, but I just download 10yo tv shows that have soo many seeders they dont do much)