Regarding "[...] I'm not personally in a position to action results of such a poll":

Are you in the position to bring up the results of such a poll in a meeting with your team & boss?

And if so, did you?

@jaffathecake @firefoxwebdevs

@duke_of_germany @firefoxwebdevs they're aware of the sentiment. I'm sure you're aware that Mastodon has a high representation of folks who don't like AI, so presenting evidence that Mastodon users don't like AI is kinda… well… not really useful.

@duke_of_germany @firefoxwebdevs whereas the poll I actually ran, which shows how target users of the feature view translation, was practically useful.

@firefoxwebdevs Calling them "enhancements" sure is a choice.

@viralobscurity both Librewolf and Vivaldi have built-in translation, which people overwhelmingly consider to be AI (https://mastodon.social/@firefoxwebdevs/115849251057488746). Vivaldi even sends the text to a server for translation.

Your choice is your own, but we are up-front about the AI features we have.

@vex I’ve been on the side of the maintainer who misjudged something and has gotten heated responses, and from that experience I think that I got too heated.

It’s really hard not to get defensive when there is a lot of heat already.

And FF devs are regularly taking the heat, even though most just try to do good work.

When you know someone receives a lot of aggression already, you shouldn’t add more.

I realized too late that we were effectively ganging up on them.
@firefoxwebdevs @dveditz

@condret @mike our translation, tab naming, and link preview features all use local models.

The AI sidebar doesn't, but I'm assuming that's obvious as you select a third-party service to use.

@ArneBab people indicated pretty strongly that they wanted a way to block AI, but re-enable particular features https://mastodon.social/@firefoxwebdevs/115849251057488746.

It used to be referred to as a 'kill switch', which is much stronger wording, and a lot of folks wanted that wording (see the replies https://mastodon.social/@firefoxwebdevs/115740500373677782). But I think "block" is a reasonable middle ground. I understand that you don't.

@firefoxwebdevs @sarah the fact that this is the best you can do with a technology designed from the ground up to ignore/bypass consent, is exactly what's pissing everyone off.

The very fact that this technology is being implemented directly into the browser telegraphs the intention to not keep that promise.

@ArneBab @vex @dveditz I found our discussion here reasonable, fwiw. I appreciated your viewpoints here, even if I didn't agree with all of them.

@vex @sarah is there a particular AI feature/flow that you feel bypasses consent?

Models are downloaded after the user indicates they want the feature.

@firefoxwebdevs Sounds great. Fuck your browser.

@firefoxwebdevs
Better late than even later… thank you!

@wojtek
Because optIn is better than optOut.

@fabio @firefoxwebdevs

@firefoxwebdevs YES! Thank you.

@ArneBab @firefoxwebdevs @dveditz as a former developer myself (10 years), I understand where you're coming from. Were this a normal unpopular feature, I'd agree.

But "AI" is designed to break consent from the ground up. It cannot function without theft, & there's no ambiguity around the harm of such an integration, both now & in the future.

The logical conclusion of inviting a digital bandit to the entrance of the Internet is the control of information by those who own the bandit. It has no place in a supposedly privacy-focused browser. & It's right & correct & moral to yell & gang up on anyone trying to invite the bandit to permanently gatekeep a popular entrance to the internet. Especially when the person inviting isn't naive about the consequences & is gaslighting about the what why & how.

@ambiguous_yelp @firefoxwebdevs I was convinced the damage is irreparable. Wasn't everything just deleted?

also, several people in the replies here point out that firefox questions the users choice after they already toggled the kill switch. im normally fine with software giving me a warning if it could potentially mitigate unintended consequences, particularly when it comes to security and/or privacy. in this case, a warning message would be more appropriate if you turned AI features ON, not OFF. i think thats why the demo video especially irks me

@vex AI built on free culture content and staying true to its licenses (including attribution) is not theft.

Wikipedia is explicitly licensed to allow derivative content -- if it’s under cc by-sa, too. The same goes for almost everything I create outside my job.

Project Gutenberg provides many books that are in the public domain. Training a model on these is unproblematic.

Mozilla voice gathered voice data provided by volunteers with full consent.

@firefoxwebdevs @dveditz

@firefoxwebdevs@mastodon.social Well, it should have been there first and disabled by default, but at least it’s here now. Cool.