I'm not asking for faith in our direction - the thing I love about the Firefox community is how open, honest, and technical it is.

But I do ask that you don't have the opposite of faith. Like, try not to be determined that we're going to do the wrong thing here.

…

have you told the people actually pushing the releases ? Currently the opt-in is hidden behind the about:configs page, and you actually have to opt out. That's not how an opt-in works.

May I suggest everyone in your organisation needs to attend a primer course on consent.

@me no. That's a direct quote from the post.

blog.mozilla.org/en/mozilla/le…

@SamatSattarov so your definition of opt-in includes enabling a bunch of browser.ml about:config settings after updates, including all the ones I’ve already disabled, just in case I change my mind and want my browser to be full of absolute horseshit?

that’s fucking worthless and I’d tell you to feel ashamed that this dark pattern crap is what you think constitutes consent, but let’s be real: you’re a PR mouthpiece for an AI corporation and are incapable of shame.

@jor

“AI” is crap, produces wrong or misleading answers all the time, and this cannot be fixed, by design. That’s in addition to all the ethical concerns (stealing people’s work, awful environmental impact, etc.) that should have excluded it a long time ago.

Unfortunately, 'AI' is a meaningless marketing term. While most of the things under that umbrella are nonsense I have been impressed with the on-device translation models that the Firefox folks have produced:

• They are about 50-100 MiB.
• They run well even on a mobile phone CPU.
• They are trained on documented, public, open data sets, which were specifically created for training translation models.
• They have done a lot of work to make sure that it is possible for an individual to reproduce the training on a single desktop.

There aren't many things in the AI hype nonsense that I say positive things about, but this is the biggest exception.

@jor @david_chisnall I think this is where some of the communication has gone badly.

We're also looking at using on-device models to provide alt text in cases where it's missing, which should improve a11y for those who need it most.

I don't think that's the kind of thing people think of when they hear AI, but it is.

@jor
My number one request to Mozilla would be: stop using the term AI.

It is a nonsense marketing term being pushed by Snake Oil Sam and his merry band to try to pretend that machine learning models are magic.

Talk about:

• What your systems actually do (and why I should care).
• What you have done to make them secure (on device, not able to run commands) and therefore why I should trust them.
• What you have done to ensure that they are ethical (trained on properly licensed data, from sources that consent to this use).

Do that, and you will regain my trust. Tell everyone that you are an 'AI-first company' making an 'AI-first browser' and I will assume you're a bunch of grifters trying to cash in on the current fraud bubble.

@david_chisnall I am not sure it really is the main issue here, as this implies again that it would be a bad communication problem, while it is not.

Although it would indeed be better to stop using misleading marketing terms like “AI” and replace them by better and more descriptive terms, the main issue to me here is: the inclusion directly in the browser of a technology that is by design unreliable, unethical and dangerous on many accounts.

Whether we call them “AI chatbots and summaries” or “dangerously misleading and unreliable LLM chatbots and summaries”, in both cases none of this should ever have been included in the first place.