(For the younger hackers: https://en.wikipedia.org/wiki/File:Mozilla_Phoenix_logo_vect...)
[1] https://logos-world.net/wp-content/uploads/2021/08/Firefox-L...
Show the “full image” with a pond of Google and Microsoft crocodiles. Because that’s what’s really going to happen. Mozilla’s little fox is going to get eaten alive.
A browser is literally a user agent. What well-funded org should be entrusted with making an open source agent for the user instead?
I use LLMs quite frequently, but there are some places I do not want them. "Use AI to chat with your PDF!" The only thing I'd want to have it remotely touch in my browser is translations.
The problem is not that a browser should not act intelligently on behalf of the user. The problem is that what is usually called "AI" is known to sometimes act erratically and invasively, and also consume a lot of (local) resources. That is, the "AI" is not trustworthy enough. And the key feature of Firefox for much of its audience is that it's more trustworthy than the browser named after a particular shiny transition metal.
I don't want the AI to be front and center in my browser. I do want the AI, if present, be local, and be distributed among tools: a better reading mode, fuzzy search on the page that searches by meaning, recognizing text on images (and also make it searchable and selectable), creature comforts like that.
If I need to chat with an LLM, I want it to not be bound to my browser.
And yes, I want to never need to start Chromium because a rare specific site refuses to work correctly in Firefox. If AI can help with that, splendid! But I suspect something else may be needed more.
It should be opt-in by default.
Why: Because AI is constantly and very frequently changing and evolving with lots of security concerns given how much scope/context/permissions it's typically granted. By having it enabled by default means that you have zero expectations that whatever settings/preferences/configs you changed in order to "opt-out" may no longer be respected/preserved/effective.
This is a major problem before we ever get to "what are the specific problems" regarding AI.
This has nothing to do with what an AI “agent” is.
Firefox should be the browser that respects you privacy (the only one...). Integrating AI undermining the efforts of making it the privacy oriented browser.
For now the AI is forced and ridiculously complicated to disable (with new options in about:config poping in each new version). The promise to have an "disable all IA features" is still a promise.
I'm personnally sad that now we have to consider banning Firefox for company use, because it's hard to verify that we've disabled every AI "feature" that might funnel our data to remote servers.
Correct.
> but it's crazy to be against them for AI.
Disagreed.
> A browser is literally a user agent.
In the same way that a car is literally just some wheels. It's overly-reductionist to the point of being adversarial.
> What well-funded org should be entrusted with making an open source agent for the user instead?
What does that have to do with the topic at hand? Maybe if you didn't try to strip the context (Mozilla, it's reputation, it's actions, it's incentives, and how this AI initiative conflicts with the userbase' expectations and references therein) all this would seem a lot less "crazy" even if you still disagree.
Mozilla's users aren't being unreasonable or irrational for voicing criticisms here.
Sure, there's plenty of blind-hate for AI. But even many of us that aren't don't like the way Mozilla is going about this for a number of very valid reasons/concerns well beyond "I don't like AI"
I think something like xkcd comics or something similar has always been received well by the community. Given that it's a high-effort piece of content as a digital painting, I think it should be ok - or at least not treated like it's in the same bucket as memes.
turtleyacht•1mo ago
Something we haven't observed yet are hyperlinks automatically created from a web of documents. This is usually a manual process: which word or words to select, and which specific URL to go to.