Agreed, it should just be a permission like any other. If I'm on a website where I expect to play media I can remember the preference.
They probably don't want to break websites (as autoplay has no permissions model), but breaking the most annoying feature on the web right now is a good thing.
In Chrome I can type Ctrl+L You <TAB> <Search Query> <Enter> to search YouTube. Chrome figured this out on its own by learning that when I type "You" I usually mean "YouTube". Ever since then that behavior has been very predictable for me. Why would Chrome learning what sites I usually play video on be any different?
Reminds me of fuzzy matching systems like Alfred that will try to learn which key combos you use to launch each app instead of trying to offer you predictability.
But "fi<enter>" never launches Firefox anymore because of that one time you accidentally launched Finder instead. And since you keep accidentally launching Finder that way, it never learns that you want Firefox.
So you try "f<enter>" but you accidentally launch Flux. And because you keep forgetting it thinks "f" = Flux, it always launches Flux.
So you have to use "fir<enter>" to launch Firefox. But "fire<enter>" somehow launches Firewatch, a game you forgot you had installed on your laptop.
It's the most obnoxious system ever where each additional keystroke reorders the entire fuzzy result list in the name of "learning." It really is the opposite of predictable.
And then there's Chrome, where you can start typing something in the URL bar, and it comes up with some results immediately, but in the milliseconds before you select a result it changes to something "better" so you go to the page instead. Infuriating.
I have pretty much the opposite anecdote: to me the Chrome URL bar is a magic component that finds any of the tickets I've lately worked on (from three different ticketing systems) with typing a few words/characters, knows that "n" is HN etc, "g" takes me to gmail.com etc. And all without any effort to manually teach it.
Google Spreadsheets is where it often fails to work perfectly - hard to get it to take me to the spreadsheets homepage if I don't remember exactly which sheet I want to open.
Honestly, this has been the most annoying part of agency work for me. Every three months I work in a different github repo, so I constantly need to forcefully make it relearn what gi+down-arrow means
Sounds like the "classic" problem where someone wants to build a reinforcement-learning system (because "self-improving AI" sounds so cool) but don't actually have a suitable reward function that would describe their problem.
Nevertheless, they don't let themselves be caught up by this minor obstacle and use whatever random reward function they can implement with the data they have.
The resulting system won't actually learn to solve the original problem - but it will learn something, so, hey, it's self-improving!
See also: Probably every single recommender system in use. (At least that's my subjective impression)
Quicksilver on macOS solved this by having a contextual "add weight" option for the results returned by Quicksilver. For me it's been one of the most intuitive solutions to this problem, where you can add incremental weight or just pile on +1000 to the result's weight to shoot it to the top.
For me (and relevant to TFA) this is what is often missing from such learning engines; an option to manually intervene when necessary. It's hubris to assume that the heuristics will always get it right, and I would have to imagine there's valuable data in seeing what manual corrections a User makes, so I'm not sure why it's not included more. A .conf file, the "add/remove" weight as mentioned before, just some way for the user to correct the heuristics when they get it wrong. And they will get it wrong, and frequently do.
Unfortunately, macOS does the exact same thing. I'm typing "ph" to "pho", and suddently it matches "photo booth" instead of "photoshop". Absolutely unpredictable...
Fucking photo booth. Perplexed me that the system always things I want that over Photoshop or Photos.
This is the part where Apple apologists get the whole “Siri/Spotlight is bad because they do everything locally and they respect your privacy” thing wrong. Nothing about this problem needs to be done “in the cloud” - their local algos need work.
For a while “c” was calendar, “ca” was calculator, “cal” was calendar again. I use launch bar and that has been my only irritation. It seems to continue learning and not get stuck on certain key combos.
They probably don't want to break websites (as autoplay has no permissions model), but breaking the most annoying feature on the web right now is a good thing.