Technology

There's nothing wrong with using an LLM for offline private language translation. It literally preserves privacy by not simply sending all that data to a Google translation server.

There's nothing wrong with using offline image recognition to aid in helping blind people know what's on their screen.

As for their "advertising" - you should look up what they actually did. It completely preserves privacy while at the same time not completely destroying the economic model that content creators rely on. It's a good thing. With any luck, regulators will enforce it.

Well, you have the option to elaborate otherwise. Huge effort to normalize this survey.

right? mozilla, you gotta focus on making a good web browser right now. not a more gimmicky web browser

I agree that's basically what I out in the text box underneath the AI multi-select options. "We don't want yet another annoying AI search feature or chatbot! We want a focus on useable features and security!"

Private, offline language translation is not "no tangible benefit".

Neither is alt-text generation for images to assist blind people in searching the web. That's a massive feature.

E: idk whether you're down voting because you don't want privacy or because you don't like blind people lol

An engine component separable from the UI (which was XUL and thus Firefox initial advantage that gave it popularity), deeply extensible via plugins, tunable (it would be so frigging cool to be able to turn off sections of EDIT: ... what's currently called web standards, say, drop HTML5 or JS).

What it was needed for when it was popular.

Not a Chrome alternative with a different engine.

Somehow every time I mention XUL and XULRunner people mention that one can use PaleMoon or that XUL is incompatible with some security and stability changes and so on.

I know that. I don't mean literally XUL, I mean low-level access to the engine. Allowing it to be used for things like old Conkeror and such, or just customizing Firefox as deeply as it was possible in olden days.

You're free to send your data to google or deepl instead of using Firefox's included AI translate. You know, privacy, no AI in the browser, choose one.

But their AI helps protect privacy? The main thing it's currently used for is offline private translation that doesn't send data to Google's servers.

The other main AI feature they're working on is AI-generated alt-text for untagged images, so that blind people can better use the web.

I feel like you're doing the classic Lemmy/Reddit thing of seeing the letters "AI" and automatically freaking out, before looking into what they're actually doing. We aren't talking about ChatGPT integration here...

Helping blind people use computers is a good thing.

Private, offline translation is a good thing.

If they had called these features "machine learning" instead of "AI", it would make zero function difference, but you wouldn't be reacting in this manner.

It's opt-in already, in fact you have to go out of your way to do it. And it's currently only used for offline, private language translation, to my knowledge.

That is a very good usecase considering the alternative is to send it to a Google translation server.

I feel like people need to actually read beyond the "Mozilla adds AI to Firefox" headlines.

My brother in Lemmy, how do you think they pay their engineers?

Would you rather them try and get revenue through advertising means? Because that's what it sounds like, no decision is a decision.

"Responsible use of AI" could mean things like providing small offline models for client-side translation. They're actually building that feature and the preview is already amazing.

I mean, generally, it is.

It's just that the uneducated masses don't realize that "AI" outside of today's LLMs has been improving our technological life for well over a decade now.

And so abused and misused for just as long. LLms and the hype and slop is a relatively new thing, this is old, useful, technology.

"Eradicated" is literally impossible, entire swathes of industries can only operate at the levels of efficiency they have come to rely on because of specialized models. And have for ages now, long before the hype and slop started.

Not every model is an LLM 🤦