this post was submitted on 31 Dec 2024
80 points (71.5% liked)

Firefox

18626 readers
71 users here now

A place to discuss the news and latest developments on the open-source browser Firefox

founded 5 years ago
MODERATORS
top 50 comments
sorted by: hot top controversial new old
[–] [email protected] 141 points 1 month ago (3 children)
  • no account or login required.
  • it's an addon (and one you have to go get), not baked-in.
  • limited to queries about content you're currently looking at.
    (it's not a general 'search' or queries engine)
  • llm is hosted by mozilla, not a third party.
  • session histories are not retained or shared, not even with mistral (it's their model).
  • user interactions are not used to train.
[–] [email protected] 26 points 1 month ago (3 children)

Thanks for the summary. So it still sends the data to a server, even if it's Mozillas. Then I still can't use it for work, because the data is private and they wouldn't appreciate me sending their data toozilla.

[–] [email protected] 21 points 1 month ago (1 children)

In such scenario you need to host your choice of LLM locally.

[–] [email protected] 5 points 1 month ago (1 children)

does the addon support usage like that?

[–] [email protected] 7 points 1 month ago (1 children)

No, but the “AI” option available on Mozilla Lab tab in settings allows you to integrate with self-hosted LLM.

I have this setup running for a while now.

[–] [email protected] 4 points 1 month ago (1 children)

Which model you are running? Who much ram?

[–] [email protected] 4 points 1 month ago* (last edited 1 month ago)

My (docker based) configuration:

Software stack: Linux > Docker Container > Nvidia Runtime > Open WebUI > Ollama > Llama 3.1

Hardware: i5-13600K, Nvidia 3070 ti (8GB), 32 GB RAM

Docker: https://docs.docker.com/engine/install/

Nvidia Runtime for docker: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

Open WebUI: https://docs.openwebui.com/

Ollama: https://hub.docker.com/r/ollama/ollama

[–] [email protected] 12 points 1 month ago

Technically it's a server operated by Google, leased by Mozilla. Mistral 7b could technically work locally, if Mozilla cared about doing such a thing.

I guess you can basically use the built-in AI chatbot functionality Mozilla rushed out the door, enable a secret setting, and use Mistral locally, but what a missed opportunity from the Privacy Browser Company

load more comments (1 replies)
[–] [email protected] 9 points 1 month ago

That's really cool to see. A trusted hosted open source model is really missing in the ecosystem to me. I really like the idea of web centric integration too.

[–] [email protected] 1 points 1 month ago* (last edited 1 month ago)
[–] [email protected] 56 points 1 month ago

no, i don't want to meet orbit, thank you very much.

[–] [email protected] 45 points 1 month ago (2 children)

I want to point out that by downvoting this, you're reducing the visibility of the post for other people, therefore making less people informed of the change.

[–] [email protected] 5 points 1 month ago

My thoughts exactly.

However, I always found upvotes and downvotes a bit confusing because upvote is almost synonymous with "like" and downvote with "don't like". With upvote, that assumption isn't that problematic but with downvote it is, like in this case where post will have less chance of being seen.

[–] [email protected] 2 points 1 month ago

Textbook shooting the messenger

[–] [email protected] 35 points 1 month ago (2 children)

I will try it. People are too negative about mozilla. They are a hundred times better than Google and we need them to survive.

[–] [email protected] 49 points 1 month ago (1 children)

Yes we need them to survive,
yes they're better than Google.

But no we're not being too negative/hard on them!

Lately Mozilla has been pulling a lot of anti-consumer yet pro shareholder shit.

AI is a perfect example of that,
unwanted by the majority of their community, yet still forced upon us by shareholders, for now through an optional addon, which appears to be a foot in the door, which can quickly grow into a baked in addon which ships with FireFox by default.

Sources:

They blatantly ignore their community,
and for that we're allowed to be angry with them.

[–] [email protected] 14 points 1 month ago (1 children)

Yup.

DuckDuckGo's search engine introduced AI assist and an AI chat as opt-out features, which it repeatedly re-enables at random, with no ability to disable it permanently, even though we've been able for years to set a bookmarklet to make all our other DDG settings persist.

Users are very unhappy, with requests for a way to permanently disable AI features ignored, receiving only patronising responses from DDG.

No matter, DDG's utility for searching has deteriorated these past years so severely, even relative to the deterioration we've seen with many other options, that I wonder will it survive.

It is always unfortunate when a recommended privacy tool shifts away from privacy, but several doing so all at once is alarming.

[–] [email protected] 7 points 1 month ago* (last edited 1 month ago) (1 children)

You may self-host SearxNG (via Docker) and avoid direct interaction with search engines - be it google, bing, Brave or DDG.

SearxNG will act as a privacy front-end for you.

https://github.com/searxng/searxng

[–] [email protected] 2 points 1 month ago (2 children)

Have tried out SearxNG without self-hosting, via different instances, but had to abandon it as it is way, way beyond my mental capabilities to get it to work.

I doubt I could manage to self-host, having looked into Docker for some other matter.

Using Mojeek currently, which isn't great but not too terrible.

[–] [email protected] 1 points 1 month ago

You can send us in results via the submit feedback button or contact page when they're not-so good, if you'd like to help

[–] [email protected] 1 points 1 month ago (1 children)

Though it is an off-topic but what exact issues you faced with SearxNG?

[–] [email protected] 1 points 1 month ago (1 children)

Well, it worked initially, then more often than not my searches produced no results or confusing error messages.

Experimented a lot with the SearxNG settings, and also with my browser and firewall settings in case there was some issue there, and eventually gave up.

I was unable to find information online about the issues I experienced, in part because I had no idea how to describe them in order to find help.

Think I tried it in three different browsers, over the course of a month or so, but primarily in Firefox.

[–] [email protected] 1 points 1 month ago

To be honest, I never tried publicly available instances of any privacy front-ends (SearxNG, Nitter, Redlib etc.). I always self-host and route all such traffic via VPN.

My initial issue with SearxNG was with the default selection of search engines. Default inclusion of Qwant engine caused irrelevant and non-english results to return. Currently my selection is limited to Google, Bing and Brave as DDG takes around 2 sec to return result (based on the VPN server location I'm using).

If you still remember the error messages, I might help to help fix that.

[–] [email protected] 6 points 1 month ago

The bar is in hell

[–] [email protected] 32 points 1 month ago

No thanks, I'll pass.

[–] [email protected] 32 points 1 month ago (1 children)

I appreciate the option to not install it.

Now if only Mozilla could migrate their built-in AI stuff to this optional extension so it doesn't come pre-installed, that'd be great

[–] [email protected] 10 points 1 month ago (1 children)

The built-in AI staff,you referred to, is nothing but an accelerator to integrate with 3rd-party or self-hosted LLMs. It's quite similar to choosing a search engine in settings. This feature itself is lightweight and can be disabled in settings if not required.

[–] [email protected] 3 points 1 month ago (1 children)

The built-in AI staff [sic]... is... an accelerator to integrate with 3rd-party or self-hosted LLMs.

Users are only shown Big Tech "3rd-party" options. Mozilla made this choice intentionally.

Since Mozilla is clearly capable of developing an add-on that is not forcefully installed on user's devices, they should remove the built-in thing that endorses the highly unethical chatbots run by Google, OpenAI, etc.

[–] [email protected] 1 points 1 month ago* (last edited 1 month ago) (1 children)

Users are only shown Big Tech “3rd-party” options. Mozilla made this choice intentionally.

Well, how many users really have LLM local-hosted?

load more comments (1 replies)
[–] [email protected] 31 points 1 month ago

"Don't worry, we're going to open source Pocket and make it optional any day now."

  • Mozilla Corporation (note: not Foundation), 2017
[–] [email protected] 29 points 1 month ago (1 children)

Firefox, tell your creepy little friend he can get the fuck off my property!

[–] [email protected] 27 points 1 month ago (1 children)

It's an add-on, not something baked-in the browser. It's not on your property at the first place, unless you choose to install it 🙂

[–] [email protected] 24 points 1 month ago (2 children)

For now, but one day Firefox will try sneak him into my house and hide him in a cupboard untill squatters rights kick in.

[–] [email protected] 5 points 1 month ago

Even they choose to do so in future, usually there always is a about:config entry to disable it.

load more comments (1 replies)
[–] [email protected] 17 points 1 month ago (2 children)

Man, I hope this is not the new Pocket.

[–] [email protected] 4 points 1 month ago

I'm still using Pocket 👀 (though an unofficial version)

[–] [email protected] 3 points 1 month ago

Can't be. Integrating Pocket fucked up something I already loved.

[–] [email protected] 15 points 1 month ago* (last edited 1 month ago) (2 children)

I'm starting to warm up to this stuff. There is a future rapidly hurtling towards us where, if you take the time to read and think for yourself, you will become a genius. It was happening already in some stem fields where people used GUI tools without ever reading what the buttons did, and if you took the time to read the manuals and the underlying methods, you could become vastly more competent than anybody else in your team. This "AI" bullshit is just extending the lazy culture out to every piece of information on the web, where average Joe is already unable to concentrate beyond 140 characters. Those that take the time to learn the fundamentals and read deeply will have vastly superior knowledge of any subject, while the majority will be spoon fed superficial summaries filled with errors and no way of realising.

load more comments (2 replies)
[–] [email protected] 12 points 1 month ago

Bugsbunnyno.jpg

[–] [email protected] 11 points 1 month ago

No thanks im good

[–] [email protected] 8 points 1 month ago (1 children)

Cab you use it to look up instructions on how to disable it?

[–] [email protected] 7 points 1 month ago

I've got them here: you're done.

You have to go out of your way to install ("enable") it.

load more comments
view more: next ›