I think nobody wants AI in Firefox, Mozilla
160 points by thang
160 points by thang
Mozilla is so frustrating. Ad-block has to be a plugin despite everybody wanting it and asking for it for 22 years, but AI and Pocket get built-in despite nobody asking for them. I'm not even against AI and Pocket, but it's stupid and inconsistent to build them into the base product when they should be plugins.
I really think Mozilla is only around because Google is paying them so much.
That might sound harsh, but I've used Firefox since it was still Netscape. Back in the day, they were really making a better browser, but now they seem out of touch.
I actually like their translator which is a local LLM.
I think this was a good idea as it emphasizes privacy. I no longer need to use Google Translate.
The rest of the additions, not so much. I guess the problem is bad leadership at Mozilla with no direction.
It’s not an LLM. The local translation feature used Machine Translation.
I'm not quite sure what the precise definition of LLM is, but the local translation is a series of transformer based models isn't it? (Edit: This seems to support that)
afair based on work by https://kheafield.com/ who has specialized in very compact language modelling
I think its a small LLM. Probably small language model is more appropriate terminology. But, anyway, it all depends on how you define a language model. My point was that these models are (task-specific) transformers, tuned for low-memory & CPU usage.
Tbh, adblock paints a massive legal target on your back, so making sure people can build powerful adblockers while not integrating it yourself makes some sense.
I don't see how.
In the past Opera had a built-in ad-blocker and they never had a problem. That was a different time, obviously, but the idea people can control what content is loaded and displayed in their browser isn't illegal.
And at the end of the day, I'd rather Mozilla use their Google millions to fight a lawsuit defending ad-blockers than waste it on features like Pocket and AI.
We've just had a story about this, "Is Germany on the Brink of Banning Ad Blockers? User Freedom, Privacy, and Security Is At Risk".
Quoting the relevant fragments:
The recent court ruling is the latest development in a legal battle between publisher Axel Springer and Eyeo (the maker of Adblock Plus) [...] Until now Eyeo has largely prevailed and the legality of ad blockers has been upheld.
Unfortunately, on July 31, the German Federal Supreme Court partially overturned the decision of the Hamburg court and remanded the case for further proceedings. The BGH (as the Federal Supreme Court is known) called for a new hearing so that the Hamburg court can provide more detail regarding which part of the website (such as bytecode or object code) is altered by ad blockers, whether this code is protected by copyright, and under what conditions the interference might be justified.
I hadn't seen that, but it doesn't sound like it's a done deal just yet.
In any case, I think Mozilla would be able to defend such a lawsuit better than a plugin maker.
Ad-block has to be a plugin
Mozilla is only around because Google is paying them so much
I think there's a connection there that you are not seeing.
It's possible that Google funds Mozilla so they can have an excuse to not be considered a monopoly, call it "browser washing". And if that's true and Google does it against their monetary interests, they would have a good excuse to stop doing it if Firefox blocked ads out of the box. Following this train of though, it's ironic that ad blockers are featured on Firefox add-ons homepage.
By the way, I was there too, and I don't remember Netscape being that much better browser than IE. After their death at Netscape 4, I moved to tabbed IE shells, and then Opera.
Oh I definitely see that connection, and have complained about it here before. IMO it's the biggest reason Firefox won't even try ad-blocking, and why the the tracker blocking they added a while back was so neutered and only blocked ones they deemed "bad".
I don't know if Firefox was ever really light years ahead, especially not compared to Opera, but they were at least willing to try new things. The whole idea of moving things like ad-block into plugins was a new idea that other browsers really hadn't tried too much, IIRC.
Honestly, Opera was the best, but that's a different thing altogether.
Netscape 4 wasn't, but Netscape 6 was a better browser. (But Netscape 6 was mostly just Mozilla with Netscape branding on top.)
I remember Netscape 6 being stupidly unstable and taking forever to ship. Some of the stuff it brought to bear (XUL in particular) had a ton of potential, but took even more years to grow into it.
By my memory, IE 5 was nicer to use on both Mac and Windows for quite a while. The Seamonkey suite eventually got to be OK, but it felt like that took forever. (Some of the offshoots were great, like Camino on Mac OS.)
I don't think anything Gecko-based felt nice to use until Chimera/Camino on Mac and Phoenix/Firefox on Windows/Linux. That said, once XUL got good, I really liked it and was sorry to see Firefox eventually move away from it. The Komodo IDE based on the XULRunner shell was really good, also, IMO.
I’d switched to Mozilla then and I remember that mail, news, and the browser all ran in the same process. And they were all buggy. The same-process thing meant that a bug in the browser would take out the mail client as well. A few years later, the stand-alone browser that became Firefox shipped and I switched back from Opera. I had switched to Opera to have a browser that didn’t crash a few times a day and take out my mail client when it did.
I really think Mozilla is only around because Google is paying them so much.
What? Absolutely this is true. This is precisely what is happening. If this isn't common knowledge, it should be.
To extend/amplify your point.. And why does Google pay them? Because there is anti-trust legal precedent with Microsoft on web browsers. So, this is a small "tax" to help Google to retain for all practical purposes monopoly.
If it's just AI to be AI, no. Not useful or helpful. If it's a feature that is useful and is powered by AI, sure fine, whatever.
Like in Apple land, you can just copy text out of any image anywhere pretty much. It works surprisingly well. It's AI powered, but it's a useful feature first. AI is just the implementation details.
What would be a useful AI powered feature in a browser? Better ad detection for blocking? I dunno.
Like in Apple land, you can just copy text out of any image anywhere pretty much. It works surprisingly well. It's AI powered, but it's a useful feature first. AI is just the implementation details.
And that's an interesting line, IMO. That's a local feature. As far as I'm concerned, Firefox using "AI" in local features doesn't matter to me, as long as I can turn it off if it causes the browser to use too much of my RAM or slows the browser down. I absolutely don't want my browser sending data to some remote service to "augment" my work with AI.
Those "Nobody" are usually us, not the average people. This is like saying "Nobody wants the Facebook TikTok app on their just purchased phone" but then every ad on eBay reassures you that that cheap Android phone will run whatever you need.
And from those who were part of the study and are against the feature, how can we know they are not just not gatekeeping what a browser should be and are anti AI from the very beginning? Also, it's more common for people to post a review when there's something remarkable bad than when the service was ok to average.
I am profoundly sceptical that there is a secret groundswell of Firefox users who frickin' love all the AI stuff, and you just never see any evidence of them in the wider world because they're so quiet about it, as AI fans notoriously never talk it up, see.
The "clankers support me in email" theory. Or perhaps "the clankers of the gaps". You can't prove they don't exist!!
I think it's silly to say anyone "wants" ai features, because ai is just a style of algorithm that implements a feature. As an end user, it's a black box. It's like "fuzzy logic" on a rice cooker box, I don't care how the machine does math to cook the rice, I just care that it's better.
If firefox offers offline machine translation, that's a good feature, whether or not it uses an llm. But most of the "add AI to it" crap is so half baked, just assuming users hate your ux so much they'd rather address it conversationally.
It's not that secret... Most regular people I meet love this stuff. It's only in tech I encounter both sides.
Most people I meet who love this stuff also use google chrome
clankers of the gaps again
You see, what I observe is objective truth, while what you observe is just "clankers in the gaps"!
There isnt. Every single Firefox user I know is against LLMs. My guess is that the current leadership, being from SV culture is seeking growth by changing the product to fit the demographic they are targeting. They are building the browser their imaginary user base would like, eschewing their userbase. Similar to what Nintendo went with the Switch.
But the logic of changing yourself in search of a product market fit is innapropiate for a non-profit, whose mission is to proselitize a specific set of values.
I'll bite. I use Firefox's integration with AI chatbots, simply because it's so damned convenient. I play a Chinese mobile game, which has many more players (and therefore content creators) in China than it has in the West. But I don't really know Chinese beyond the simplest words, so I use AI to help me translate what they say. Unlike Google Translate, which only translates whole phrases, an AI chatbot can also analyze their syntactic structure, and explain slang and memes. It's not perfect, of course, but it's “good enough” for entertainment purposes, as long as you can manage / tolerate the risk of it being wrong.
That being said, I still wouldn't trust an AI with my paid job, because I'm not always authorized to share my employer's information with third companies, and even when I am, the negative consequences of being wrong are much more serious than in a gaming context.
I do not think that Firefox would benefit from being "the browser without AI"; I already have enough headaches when I'm told by non technicals friends that some site won't work on Firefox, and had to move to Chrome. It's even more frustratingly ironic (or ironically frustrating) when it's because ad blocking broke a script.
For another example, recently iTerm added a chatbot panel, and it received so much backslash it had to be moved to an external plugin. Meanwhile, VSCode includes Copilot out of the box and the worst I've heard it's that sometimes it's too quick.
users who frickin' love all the AI stuff
You are forgetting those people who just turn it off and forget about it.
I really hate how they are jamming copilot stuff into VS Code for the record. I think the only reason I can somewhat put it out of my mind is that I disabled it a long time ago, and have no idea what the first-run experience looks like these days.
I think the criticism would disappear immediately if they made it a plugin.
Let people decide if they want AI for themselves. That was the point of plugins, and it's the rule everything else has to follow.
Regular, non-expert internet users find it interesting, or even amusing, to generate images or videos using AI and to send that media to their friends. While sophisticated media aesthetics find those creations gauche or even offensive, a lot of other cultures find them perfectly acceptable. And it’s an inarguable reality that millions of people find AI-generated media images emotionally moving. Most people that see AI-generated content as tolerable folk art belong to demographics that are dismissed by those who shape the technology platforms that billions of people use every day.
It's interesting that this is the example of what "normal" people use AI for, and I think it's pretty accurate. It also reinforces my hope that the AI bubble will crash before it can cause much more damage. I believe it was Cory Doctorow that pointed out that the things that LLMs (etc) are good at can't be profitable, and the things that could be profitable, they're not good at.
The article as a whole... it's not good. There's a concept in Marxist theory called tailism. Tailism is when the leadership of the party caters to the beliefs of the least politically educated members of the movement, rather than educating them. I think something similar applies here: computer users who are technically educated, politically aware, and who take time to communicate on sociotechnic subjects like this are the vanguard of the user-focused software movement. The Mozilla Foundation has positioned itself as a key element of that vanguard. Integrating AI into Firefox is a betrayal of the users, and this article tries to justify itself in tailist terms.
rather than educating them
Anil specifically tells Mozilla to do a better job at educating users on how creepy Big AI is. But still, you can't do that without offering alternatives. It's not either-or.
I want AI in Firefox 👋 as long as it's opt-in. And this is good for normies, as they'll be flocking to Perplexity and other browsers that will be way more privacy invading and insecure.
If anything, my problem with Firefox is that I can't bring my own offline AI or my own API keys, and because it appears to be loading a website each time you open it, latency is high. That chat sidebar is half-baked. If people aren't using it much, it's no wonder.
I understand the reluctance, but I think that Mozilla appealing to tech luddism is a losing proposition. Firefox has been bleeding market share partly because, on the tech side, they've been too conservative, and normies simply don't care about all the reasons for which Firefox doesn't do what Chrome already does. When push comes to shove, people will pick the standard, and featureful solution that does what they want.
As a data point, the Brave browser passed 100 millions monthly active users, while Firefox is at 150 millions MAU and still hemorrhaging users. And I'm pretty sure that's Firefox's market share they grabbed, not Chrome's. Let that sink in, a Chrome fork, with minor customizations, ad-blocking, AI integration and freaking cryptocurrency baked in, is poised to be more popular than Firefox. And when you ask its users, most will just say “oh, I just don't enable the crypto crap”. Apparently the users of other browsers realize they can just not opt-into features they don't want.
I don’t want it anywhere, I just recently updated vscode and it keeps putting chat and ai and copilot and other nonsense everywhere. It keeps adding yet more preferences around slop, everyone defaulting to enabled when added, and I would swear that even with a user/global setting it defaults every workspace to having it enabled again.
I do not want a chat bot in my editor, I do not want source randomizers, and I do not want trivial and deterministic refactoring algorithms replaced with expensive non deterministic slop engines.
I also don’t want to steal the work of others or ignore the gpl and similar licensing of the stolen code that is being laundered through the idea of these slop engines being anything other than text predictors being used to avoid fulfilling OSS licensing requirements.
Just give people the “shut off all AI features” button. It’s a tiny percentage of people... [and the paragrah continues]
No need to be dismissive. There's a (respectable) percentage that just don't want AI features to appear out of thin air like Clippy.
tiny
As if Firefox had the biggest market share /s. Both side can play dismissive :).
They could show the telemetry of how many people are actually using the current AI features, but I also disable telemetry so would not count.
On the other hand, Github justifies the features with "an increase on how many people use the included Copilot on the first 2 weeks from creating the account means thate everyone expects and wants AI", without any numbers of how many people use it again after first use, or after one month
This whole thing reminds me of the Stallmanite tactic of just shouting "software is bad actually" at everyone they meet and hoping the result will be software freedom.
Telling people the tools they want or need are bad doesn't make them seem better tools (on average, obviously everyone is different) it makes them grumpy at you. The only way forward, then and now, is to say "hey I've been using tool X and it's pretty great you should check it out"
shouting "software is bad actually" at everyone they meet and hoping the result will be software freedom.
Stallman wrote a compiler, tools, a license and organised a political movement that has allowed me to post this comment with Free Software. Why would you be so dismissive about it like Free Software didn't achieve anything?
The Free Software movement has achieved many things and remains one of the most important movements and most important communities in my life.
But none of that was accomplished by attacking software users for their choices.
I only started prioritising Free Software because I felt ashamed and embarrassed using proprietary software.
use an AI‑free Firefox fork such as LibreWolf, Waterfox, or Zen Browser
Funnily enough if you go to the Zen Browser website now, the big demo video shows ChatGPT in the tab list.. x_x
I wrote against AI in Firefox some time ago as well but moderators here removed the story saying "Lobsters is not a place for Mozilla support requests". Happy to see this post stick. We should be talking about this kind of stuff here.
I too hope that Mozilla could position itself as an alternative to "Big AI", but as of now it's very far away from that. Besides local translation I haven't seen anything coming out of there that is fundamentally "more ethical" let alone competitive. Case in point Mozilla is working on what appears to be an Oxylabs competitor.
Tangentially related: Tor Browser itself announced the "removal of various AI features" in an alpha release a month ago.
Over the past year Mozilla has been working on integrating various AI features and integrations into Firefox (e.g. the AI chatbot sidebar). Such machine learning systems and platforms are inherently un-auditable from a security and privacy perspective. We also do not want to imply recommendation or promotion of such systems by including them in Tor Browser. Therefore, we have done what we can to remove such features from the browser.
https://blog.torproject.org/new-alpha-release-tor-browser-150a4/
As someone who worked on a natural language interface for Firefox I really want LLNs integrated into Firefox! The tech was way to early for what users expected, but I think LLMs are a way to bring the CLI to the masses. The URL bar is about the only text based interface most people interacted with regularly before LLMs came on the scene. There is a lot of potential here.
The way to bring CLI to the masses is through education and teaching them in my opinion. Hoping a prompt will lead to the correct outcome full of potential for errors as much if not more than for success. Mozilla Foundation used to have an education department that did good work teaching the web and how to use it. I wish Mozilla was doing more of that than somehow trying to add LLMs.
@pushcx This seems like it should be merged into the discussion for the earlier piece it's responding to.
Mozilla is probably in this weird spot where not providing what everyone does (AI integration) would make them even less appealing to casual users, while also being niche enough to have a loud minority of power users which might be the ones bringing in more casual users in the first place.
I certainly don't need AI in my browser - but I already use deepl for translations. So I can see how simply asking a browser to summarize certain aspects of a website will be de facto standard in 5 years.
It's the same with DRM: Some people might totally be fine never having any netflix in their browser - but for a lot of the casual users it is probably a decision between that or always moving to another Browser when opening streaming sites. At which point they might as well switch completely to chrome, as it's already doing everything and has all of their settings.
I largely agree with the other commenters here. However, I see it as this: if Mozilla is being paid to implement LLMs into their browser (I don't know if they are, but considering they rely on Google, perhaps they'd accept funding from others too), then fine, so be it, so long as it can be fully and entirely disabled. I've disabled it on my browser and haven't seen a single whisper of any of the features.
If stupid features like this keep the browser alive, maybe it's the lesser evil. Still a complete waste of disk space and compute though.
I want browser to have the “handles” for it to be fully controllable by external means. But i want to have extremely elaborate ACL for this. And then I want something like Claude Code to interact with it using skills. I want to have a single AI which acts on my behalf not only in browser but pretty much everywhere. Having a separate “AI” in browser sounds like a dead-end to me.
The problem is, that we actually need a whole new set of web-apis to implement a level of control that I imagine. And I didn’t hear about any projects about that yet.