Yes, Translate is the only one I want - and we already have that!

The worst is anything that tries to suggest stuff in text fields or puts buttons etc. to try and get you to "rewrite with AI" or any nonsense like that - makes me just want to burn anything like that to the ground.

  > Image search? I have a search engine for that.

Plus, Search by Image[0] is one of the most popular extensions, with 3x as many people using it as tree-style tabs.

I don't use it but a grammar tool is the next most popular[1], so I could see this being quite a useful feature.

But the other stuff, I'm with you. I like translate but I personally don't care for dubbing, summarizing, or anything else.

[0] https://addons.mozilla.org/en-US/firefox/addon/search_by_ima...

[1] https://addons.mozilla.org/en-US/firefox/addon/languagetool/

Local RAG on your browsed pages (either automatically, manually or a mix (allow/disallow domains/url) ?

You personally wouldn't use live captions and dubbing, so there's no point building it for the millions of people who need it as an accessibility feature?

Live captions and dubbing can be a game changer for:

- non native speakers - moving away from the english-centric web - impaired people

local LLM assisted 'tampermonkey' userscript generation?

>"at least 900 W" when sorting through a list of microwave ovens on Walmart.

Newegg has that as a built in filter.

Why do you people keep insisting I "need" an LLM to do things that are standard features?

I find shopping online for clothes to suck, but there's nothing an LLM can do to fix that because it's not a magic machine and I cannot try on clothes at home. So instead, I just sucked it up and went to Old Navy.

Like, these things are still lying to my face every single day. I only use them when there's no alternative, like quickly porting code from python to Java for an emergency project. Was the code correctly ported? Nope, it silently dropped things of course, but "it doesn't need to be perfect" was the spec.

>Or searching for text in images with OCR.

That thing that was a mainline feature of Microsoft OneNote in 2007 and worked just fine and I STILL never used? I thought it was the neatest feature but even my friend who runs everything out of OneNote doesn't use it much. Back in middle school we had a very similar Digital Notebook application that predates OneNote with a similar feature set, including the teachers being able to distribute Master copies of notes for their students, and I also did not use OCR there.

The ONE actual good use case of LLMs that anyone has offered me did not come from techbros who think "Tesla has good software" is not only an accurate statement but an important point for a car, it came from my mom. Turns out, the text generation machine is pretty good at generating text in French to make tests! Her moronic (really rich of course, one of the richest in the state) school district refused to buy her any materials at all for her French classes, so she's been using ChatGPT. It does a great job, because that's what these machines are actually built for, and she only has to fix up the output occasionally, but that task is ACTUALLY easy to verify, unlike most of the things people use these LLMs for.

She STILL wouldn't pay $20 monthly for it. That shouldn't be surprising, because "Test generator" for a high school class is a one time payment of $300 historically, and came with your textbook purchase. If she wasn't planning on retiring she would probably just do it the long way. A course like that is a durable good.

So you're not going to get it until your OS decides to, and if its implementation is poor you're SOL?

The mindset of every browser vendor is that they are the OS now, and all that kernel and userland guff merely supporting infrastructure.

> Sounds like a great OS feature.

Cool, and some DEs make it possible to start implementing this for most applications today. But Mozilla is not KDE or Gnome, so the most they can do is to make this on their software, and make it easy to copy for the entire system.

> Sounds like a web site, not a browser feature.

Sounds like a bit of lack of imagination on your part. Do you think the same for text search?

>

Most of those things weren't very good before AI was applied.

Translation specifically was pretty bad before Google applied machine learning methods to it around 2007 when it became very good almost overnight.

Technically, many of those things often were AI.

They just existed before the GenAI craze and no one cared because AI wasn't a buzzword at the time. Google Translate absolutely was based on ML before OpenAI made it a big deal to have things "based on AI".

But just putting stuff in your browser that hooks into third-party services that use ML isn't enough anymore. It has to be front and center otherwise, you're losing the interest of... well, someone. I'm not sure who at this point. I don't care, personally.

Many of these things were "AI" but the marketing hype hadn't gotten there yet. E.g. the local translation in FF is a transformer model, as was Google translate in the cloud since 2018 (and still "AI" looong before that, just not transformer based).

Firefox uses local models for translation, summarisation and possibly other stuff. As it is not restricted on one platform, I guess that it has to use its own tools, while apple (or macos/ios focused software in general) can use system level APIs. But the logic I guess is the same.

I’m not exactly surprised that AI grifters that have probably bet all their life savings on nvidia “love” their AI browsers.