Tests are called "evals" (evaluations) in the AI product development world. Basically you let humans review LLM output or feed it to another LLM with instructions how to evaluate it.
Interesting, never really thought of it outside of this comment chain but I'm guessing approaches like this hurt the typical automated testing devs would do but seeing how this is MSFT (who already stopped having dedicated testing roles for a good while now, rip SDET roles) I can only imagine the quality culture is even worse for "AI" teams.
Governments do test products now and then, yes. But they aren't in the habit of routinely testing new products when they hit the market. I don't know what prompted this particular test, but it wouldn't surprise me if the ANFR just one day decided to test a bunch of models of iPhone (or a bunch of phone models in general) kind of arbitrarily.
There was a similar case recently where Sweden's equivalent decided to test a bunch of models of EV chargers and found some lacking.
In general, outside of these random tests, "they check the paperwork, not the thing itself" is correct.
Sure, but the "accredited" in accredited laboratory is itself a paperwork audit-based process. This is how you get your own paperwork as a manufacturer to give to an auditor to say "Look at our paperwork". It doesn't mean the regulator is somehow testing things itself.
You are wrong, the idea is much older and the idea for the current version (initially called 1-2-3 Ticket) was suggested by the green party in 2013 already.
Jitsi does not re-encode video streams on the server, it's a SFU (selective forwarding unit) and just forwards (some) video streams to the other participators.
When Simulcast is enabled, multiple resolutions of your video are streamed to the Jitsi server and it only sends those streams to clients which they asked for (participants can choose the quality of video they receive, in the UI)
https://www.lennysnewsletter.com/p/beyond-vibe-checks-a-pms-...