You must log in or register to comment.
The worst part about the Firefox AI stuff is their provider selection is shit. No Deepseek, no OpenRouter, so I have a stupid pane and a stupid popup button every time I select text that only works with models that are either inferior or closed off.
I don’t understand why they don’t let you add any LLM you want like you can with search engines.
To be fair, Firefox on desktop doesn’t let you add custom search engines, by default. Unless you flip magic key
browser.urlbar.update2.engineAliasRefreshtotrue.Wow that’s depressing. Apparently also Firefox makes it very difficult to install extensions that aren’t signed by Mozilla, which I learned when a pro-Palestine extension got removed from the Mozilla store and we were discussing alternative ways to distribute it. It’s an issue in LibreWolf as well, although there might have been less hoops I had to jump through to disable that.
I wouldn’t mind a decent LOCAL open source AI helping
Firefox can use a local llamafile model, but you have to enable it in about:config first.
Honestly it’s easier to find an addon that’ll hook to ollama instead, fire fox’s inbuilt support is shit
DeepSeek’s model is open-sourced and can be run locally; though I think there some bits related to its training data they have been kept obscured (if I remember correctly) - likely due to the dubious nature of how it was acquired.
Unless training data is made available, a model is not open source. DeepSeek is better described as “open weight”.
some bits related to its training data
AKA ANY details about its training data, and its training hyperparameters, and literally any other details about its training. An ‘open’ secret among LLM tinkerers is that the Chinese companies seem to have particularly strong English/Chinese training data (not so much other languages though), and I’ll give you one guess on how.
Deepseek is unusal in that they are open sourcing the general techniques they used and even some (not all) of the software frameworks they use.
Don’t get me wrong, I think any level of openness should be encouraged (unlike OpenAI being as closed as physically possible), but they are still very closed. Unlike, say, IBM Granite models which should be reproducible.
Large X models lack a crucial component of “open-source”. Freely redistributable and modifiable for any purpose, sure, but there’s no chance in hell of auditing one, let alone if the training data is kept a secret. It’s literally impossible; human beings cannot look at a trillion weights and biases representing a single highly chaotic, unfathomably complex nonlinear function whose input and output space are the totality of human language/images/etc. and say “yup, looks good to me.” Deep learning models – contrasted with traditional machine learning models – learn their own features which almost 100% of the time would be nonsense to a human. You just have a blob of shareware when you run DeepSeek.
(They also just outright steal from billions of copyright-protected sources to create it, so calling it “open-source” is pretty funny.)
Auditing for bias purposes, yea true. But my primary concern is it having the capability to “phone home” which you don’t really need to audit the model itself to be able to detect or prevent
There are a few that are “truly” open like IBM Granite, and a handful of others over the 7B range.
Generally agree, I do appreciate Firefox’ built-in translation tool though, that also falls under “AI” I guess.
Idk if that uses AI i am unsure
A bajillion things are “AI” now, and weren’t before recently. It’s so frustrating to see people hate them all equally. It’s like when everything started to get called an “app” but worse.
AI has so many uses and it has been employed in scientific research for years, Google’s DeepMind event got the Nobel Prize for that. It’s sad seeing people hating AI and claiming it has nothing to offer. But what else can you expect from haters.
Says the AI post. Okay ChatGPT. You can go to bed now. The humans have it from here.
Go touch grass, that’s obviously not an AI post lmao.
Oh, fuck off, idiot.
I have Firefox on my PC but I gotta say, Safari on my MacBook and iPhone hase been solid. It has, so far, done exactly what the post wants. Safari doesn’t just stay the hell out of my business but it also seamlessly shares tab groups with my phone and that’s super nice, too.
I’m sure there are many more hidden things that I will learn are bad about it after posting this comment but on the surface it has been a perfectly unexciting, simple, and easy to use browser. I didn’t even think about it right away and had to come back to this post because of how delightfully boring it is despite using it every day.
I’m a web developer and I always get shit on for actually loving Safari. I don’t know why it’s a crime to love a web browser that stays out of the way.
If you need Chrome or Firefox-style extensions there’s always Orion.
Orion my beloved
How do I 10x my upvote for a post?
As much as I hate Firefox having AI, they really don’t have a choice. If the majority of people are already using it and don’t give a rats ass, they’re absolutely gonna switch to the AI integrated browsers ( chrome, edge, probably safari if they already have AI in it or are working on it, etcetera ).
Firefox is inbetween a rock and a hard place right now. They either not add AI integration and attract less users or they do and risk alienating their current small userbase and becoming irrelevant enough to become unusable because big tech 100% enforces a new web standard that non-AI Firefox cannot handle.
For now, I’m siding with Mozilla on this because I can almost 100% guarantee if Firefox falls, the free web will die in less than a year. No more Librewolf, Firedragon ( floorp w/ Librewolf settings/patches IIRC ), etcetera, because if we’re being honest, what open source company/rando volunteer has the time, drive, and money to keep the Gecko rendering engine alive? And that’s just a start to keeping Firefox alive.
I’m far from an AI hater, but I fully agree with this.
I think there’s a distinct business oppotunity coming up for two things: Hassle-free self-hosting and back-to-basics apps and services.
Nobody is tapping into those correctly (you’re going to want to give me examples of self-hosted things, and you’re wrong), and it’s extremely hard to do either right, but if you can figure it out and are ballsy enough to build a proper business around it I may be interested in your pitch deck.
Can you elaborate on “Hassle-free self-hosting” & “and you’re wrong”
genuinely curious to see what your argument is here.
Kinda not the point, but at the risk of starting a huge tangent: yes, there are a bunch of self-hosted applications that are reasonably practical and easy to install, but there’s still the layer of having to understand how to access a thing in your LAN from each device, and ideally you’d want some sort of dedicated server running at all times and a bunch of this stuff is provided in multiple formats, including containerized versions or versions for virtual machines, all of which is way over the heads of normie users.
The closest to a fire-and-forget self-hosting platform is maybe Home Assistant or perhaps some of the commercial NAS sellers, like the Synology suite of apps that will mooostly set themselves up. Maybe Plex. But even those don’t work in quite the way mainstream users think about applications working. You really need something you plug in and it goes. Maybe the branded Home Assistant hardware is closest to that, but HA itself is so overengineered and customizable it’s not so much the start of a commercial self-hosting revolution as a relatively accessible hobby project rabbit hole.
Have you heard of YUNOHOST? Thats all I’ll ask I dont want to like waste your time if you have and you already have an opinion.
I hadn’t, but at a glance, while well intentioned that’s pretty much exactly the “still a bridge too far” thing I’m talking about.
Effectively that mimics the interface (bit uglier, but same idea) you get in a Synology NAS or other commercial home server services.
Here’s the problem, Jellyfin itself might already be alien tech. The type of solution they’re proposing is trying to streamline something end users don’t even know exists.
And I’d be moderately interested on it at my level of awareness, but now I am looking at redoing my own self hosting machine from scratch and wondering if some of the things I’m doing with it will be doable with this, so as of right now, moving to it is more complicated, not less.
The bar self hosting needs to be mainstream is this: I click a button on a Windows PC and it downloads a piece of software. I click “install” and said software installs itself like a normal application.
There is now an application I can use to do a thing everywhere.
Alternately, I buy a little box, plug it in and there is now an application I can use to do a thing everywhere.
The only examples that approximate this in my view are Plex (NOT Jellyfin) for scenario one and HA Yellow/Green for scenario two. And even those two will set up the hardware and software but you’ll still be pointing at a LAN IP for access. They both will only do remote access via a subscription and a connection to an external could-based service, so they aren’t even a fully self hosted solution if you want to go with the “easy” proper external access.
While I see where you’re coming from, I feel that given the current economy (subscriptions & enshitification) what youre describing functionally can not happen without a major shift in consumer habits, or a whole shit ton of regulation. I love the idea of making Everything accessible, but there is always a catch 22 with these things. Your example for instance being Plex, dont they charge you to encode and stream your own media? Thats a bit ridiculous, and should be free. If you pay for software, it should be a one and done. I’m not subscribing to use features of the hardware Ive already purchased. To end though, I do understand where your coming from, and in an ideal world would agree with you. I dont think that ideal world is going to happen without some sort of insane turning point in consumer consciousness or a major gov. enforcing a new law.
I think there’s a bit of misunderstanding there. I’m not saying we should force self-hosting. I’m saying that when you get enshittification to a certain point, the idea of a non-shitty service becomes a selling point and you can compete on that as a feature.
You see that in commercial software all the time. Davinci Resolve exists because nobody wants to deal with Adobe, ClipStudio grew for the same reason, then went around that loop and now Affinity is getting some attention, and so on.
So what I’m saying is a self-contained package/service for self-hosting has a good chance to compete on price and features with enshittified services. The problem with getting that out of the OSS community is that they typically have more decisionmaking power on the engineering side and you end up with overly flexible, customizable software no mom and pop normie would ever get into unless they’re making a project out of it.
See, Jellyfin should be a hit. Everybody should have a Jellyfin server. But instead they have an overly powerful thing that is trying to allow you to customize the UI and incorporate every single piece of media and do everything Plex does except for the one useful thing Plex does which is give you Internet access to your library.
That’s the opposite of what an eventually successful self-hosted thing would be. You want one thing that does one thing with zero hassle and has the hard feature but none of the superfluous easy features. That’s why I’m saying HA, Plex and Synology are best positioned.
I think Synology is going the Plex route, where they are starting to enshittify their hardwareto sell you more hard drives. Their software is a better version of Yunohost already, though. And crucially they do provide a one click OpenVPN install, which is the still-too-complicated version of how all of this should work.
But if you really wanted to make some money one can envision a world in which a ISP (particularly a Starlink-style connect-anywhere ISP) sells you a one time stop package with a box that does your routing and also has a big app manager thing that sets you up for what you want. “It works just like Gmail but it’s at your place” is the pitch, not “ironclad security and full access to set it up just like you want”. That’s for nerds.
And then you charge them for cloud backups, if you’re clever.
Thanks for coming to my pitch, I’ll be in meeting room 4 all week.
I guess what I am not understanding is why you would self host if youre not doing it for privacy and security. Who cares where the information is if the info gets to where it needs to be? If the user doesnt set it up just how they want and all decisions are made by a 3rd party, why not just centralize the information? That 3rd party can still implement all the spyware, telemetry and backdoors they want to into the software so to the end consumer it makes literally no difference other than the fact that now instead of everything being in “the cloud” you have to spend $100+ on a box that does the same thing the cloud did.
back-to-basics apps and services.
I think these do exist, but they’re in such a sea of shit that most users scrolling on their phones can’t find them. Shameless apps have an intractable engagement/marketing advantage over them, as do the ‘lets get acquired by Big Tech’ ones.
I guess big companies could engage in this, but… shrug.
Hassle-free self hosting is hard, yeah, AI or not. Not going to argue with that one bit.
Accept cookies?
I want my web browser to actively defend me against tracking/enshittification/exploitation/hostile design, then show me cleaned-up web pages with all the ads and shit removed, then get out of the way.
I want it to show me the information (which is not same thing as the “page” as a whole) that I’m looking for without modifying it or hallucinating some kind of AI summary, but I want it to aggressively get rid of as much of the extraneous crap obfuscating said information as possible.
Thank fuck for uBlock origin
yeah, whenever I have to look at someone else’s browser and it’s an ad-filled hellscape I’m really grateful for uBlock. The internet would be completely unusable for me without it.
Same when people talk about how creepily the ads target them based on circumstantial stuff* it feels like an alien experience bc even if I get targeted despite employing quite a few tracking blockers, I never actually see the ads lol.
(* like that story of the father hearing about the daughter’s pregnancy because he got spammed with baby care ads after the daughter googled some medical symptoms)
+ bonus recommendation for those of us who still have to use Facebook: F.B. Purity is great
And Firefox’s reader mode, and noscript.
Actual proof of good in the world
Reader mode is great, though I’ve seen some sites that seem to have taken deliberate steps to make their articles unviable with it by making all the text disappear as soon as you turn it on
also consent-o-matic and canvas blocker
Reader mode on mobile is a let down because there’s no way to force it.
i had 2-3 other ones for extra protection.
and ironically, LLMs could be great for this! recognizing what’s ads and what’s content, what’s slop and what’s high-effort, wading through the cesspool of feeds and dark patterns to find the stuff that’s relevant to you.
unfortunately, the money is in using LLMs to generate more slop and make things even worse, not make it better.
I could believe it for advertising versus content (to an extent), but I think it would not be useful in ‘slop’ versus content, for the same reason it’s output is slop. If an AI approach can detect slop, then a related AI approach can generate better slop that it could no longer detect.
But it could also make advertising more baked into a content that is hard to extricate.
I actually would be pretty happy if my browser could detect and block ads.
But they put a fuck ton of work in to not only NOT do that, they expend material efforts fucking with extensions and other tooling that provide that functionality.
Blocklists are a much more efficient way to do this, and TBH many “traditional” adblockers are still huge performance hogs. Ublock is an exception in this regard due to webassembly and its explicit dedication to lightness.
Vision models are a pretty good way to build sponsorblock/adblock databases though, and maybe even engineer HTML workarounds automatically. It would be cool if you, say, encounter an ad or a dysfunctional web page, and you can opt-in to automatically contribute a fix with your own compute.
I always assumed adblocks already were first-passing against known-advertizing patterns and then rewriting the DOM on the fly. I’m surprised that a vision model would be more performant given that it’s still going to have to adjust the DOM anyways.
I’m talking theoretically, heh, I don’t think anyone actually does that yet.
And I am just talking edge cases where existing blockers fail and there’s no manpower to figure out a customization.
deleted by creator
Firefox offered me a survey the other day on this exact topic. I said I don’t want it in my browser for all questions.
“What if your browser…”
No, just no. Please stop shoving new features in that I won’t use.
It’ll get to a point where you just have to work on your critical thinking skills and just be a pessimist because everything that’s going to be presented to you is just bullshit lies. So just acknowledge that this relationship is adversarial. Listen to other people talk about work cited, maybe dig into the unknown, the abyss. They will take everything away from you. And they’ll make you feel bad for being angry. You are the product. There is no escaping capitalism until you’re ready to do something about it. At this point it’s just the game of cat and mouse and you’re getting closer to the corner. Please, I know, I’m super fucking negative. Don’t stop doing things. I’m just saying. Half of the battle is being aware.
Prompt: please summarize this meme for me
Browser ai bad
Browser go to webpages good
Jarvis, summarise this comment, I ain’t reading allat
If i have a question i want an answer not a bunch of links where i might find the answer to my question if i read all the pages and try to connect the dots. So yes, i want all of it.
I think its fine as long its not forced on users that don’t want it
There are always options 🤷🏼
This is honestly kinda scary to read. You want an intransparent software that can by definition not think to try and check what facts are correct instead of doing it yourself? And that’s if we’re assuming there’s no intentional fact skewing in the software.
It is certainly the most convenient interface, and that’s what makes it enticing.
I don’t think I’ll ever trust one source enough to use it like that, though.
So you also never use google for instance or do you first compare the results of google, DuckDuckGo, ecosia,… before actually open a page? Interesting, i wonder how long it takes before you find something on the web.
Where did i say i want intransparent software that can by definition not think to try and check what facts are correct instead of doing it yourself?
Thats unfortunately the only way to get what you say you want. Unless you’re paying a human to do the web searching for you.
So you don’t use any search engine at all i understand. For instance, are you confident that google is fully transparent and gives you only checked facts? No intentional skewing towards favouring websites that pay for a high ranking?
Maybe the difference between me and you is that i always check the facts …for example where they come from. If the answer is given by a human or machine makes no difference for me. The machine though gives me the links where it derived the information from… not many humans do that.
So you either crawl back in fear fuelled by a lack of understanding or you embrace new tools when they come and learn how to work with them, what can be trusted and not, what improvements can be made. 🤷🏼
Example:
I get an answer AND a list where the answer is based on. Personally, i don’t understand your issue at all.
Imma ignore your blatant rudeness and strawman based ad hominems in the above comment for a sec
So, if you’re going to check each of those sources, what’s the advantage of those over using searx? Basically, if you’re going to do your due diligence, you’re not even going to have to look at the generated summary at all. Searx has the additional advantage of being open source, so you can go check how it does what it does. That’s impossible to do with AI by its very definition- even the devs can’t know why it does what it does.
Just responding to you with examples. The rudeness is yours.
I get what they are trying to say, but I definitely don’t want my browser to just facilitate me raw-dogging the internet. I had to use someone else’s computer at work the other day, and they don’t have any ad block and have apparently clicked “yes” to every dialog box for years. It was a fucking nightmare. Every web page was so full of ads, pop-ups, notifications, banners, auto-playing videos, etc. Jesus christ, I just needed to check the weather on a local news website and the internet skull-fucked me until I had ocular hepatitis. Decided the safest course of action was to just stand outside and look for tornadoes myself.
I want my browser to let me raw-dog the internet until I tell it otherwise.
