Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.
-
@zzt I posted this poll after a meeting where we discussed the design of the kill switch, and there was uncertainty around translations. I want to make sure the community's voice is represented in these discussions.
> I want to make sure the community's voice is represented in these discussions.
if that were true, the poll would have had a "remove all LLM functionality" option.
-
@firefoxwebdevs going through all the other replies and your lack of response to any of them..
“why are there flaming bags of poop on my porch, and why do they all have different postmarks”
-
@funkylab @flxtr @firefoxwebdevs there's no such thing as good machine translation
@typhon @flxtr @firefoxwebdevs oh I'm sure not being able to read what the Japanese seismographic agency's website said about the underwater quake is much better than having had a machine translation of that page for my friend in Alaska. </sarcasm>
-
@eckes @fasterandworse To further the charitable mission, pretty obviously.
q1 - design one
q2 - see the post you responded to -
@twifkak Also notice that Mastodon instances are using LibreTranslate.
Has that been debated as well?
-
Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.
They're not LLMs. They're trained on open data.
Should translation be disabled if the AI 'kill switch' is active?
@firefoxwebdevs I've only got one more firefox uninstall before all my machines and devices are completely free of your AI, ML or whatever you're branding the scraped, stolen data, trained models you see fit to use now and in the future.
-
@zzt I posted this poll after a meeting where we discussed the design of the kill switch, and there was uncertainty around translations. I want to make sure the community's voice is represented in these discussions.
@firefoxwebdevs @zzt As a member of the community: Do not give us a kill switch. Give us a build of Firefox with the AI/ML capabilities removed entirely at compile-time. Then either you can supply a second build with the slop features, or possibly offer your slop features as extensions.
-
@firefoxwebdevs as a user, I like and use translation. Having one app render and translate content makes sense to me.
I like how you do it (incl on-device, on-demand and privacy-preserving, and open data (assuming it means not copyrighted?)).
Because of both, it is clearly different from other “AI” to me, even if it technically would use language models that are large, and this poll makes sense to me.
It's tricky, I voted, but wasn't super sure. I think granular controls would be great.
@hdv @firefoxwebdevs Hidde’s perspective lines up with mine, I think. It feels different to me, though I could see arguments for either UX approach.
-
Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.
They're not LLMs. They're trained on open data.
Should translation be disabled if the AI 'kill switch' is active?
@firefoxwebdevs the language and framing of this poll is misleading. i voted for "re-enable translations", but after reading more of the discussions happening in the comments i would have rather voted for "yes". what kind of open data are we talking about? why does this need to be built into the browser instead of being available as an optional extension?
-
Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.
They're not LLMs. They're trained on open data.
Should translation be disabled if the AI 'kill switch' is active?
There should be an option to enable/disable translation regardless of any other feature.
-
potayto, potahto
-
Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.
They're not LLMs. They're trained on open data.
Should translation be disabled if the AI 'kill switch' is active?
@firefoxwebdevs Is it an off-switch, or isn't it?
"Off-switch except for this PM's pet project" is not an off-switch.
-
@gatesvp @firefoxwebdevs @sil @jmax this is the sort of obfuscatory claim I see from AI marketers. "You say you hate slop, so that means you must hate X-ray scanning! Checkmate, AI hater!" It's not convincing.
Let's assume you're correct.
People only care about AI slop.Why is Firefox even running this survey? Like who cares? Translations aren't "AI slop", they don't need to be covered by the "AI Kill Switch"... why are they even asking this question?
Now take that assumption and read the rest of the comments. From what I'm reading, people care about more than just the AI slop. People are asking questions about the models being used for ML systems, systems that are incapable of generating AI slop.
So we're at a weird spot here. You believe that people care only about AI slop. Firefox obviously believes that people care about more than that, because they're running this survey. And people responding are asking questions that also indicate they care about more than AI slop.
So how do we square this?
What do you think is a better outcome for Firefox and the community? -
@mdavis @firefoxwebdevs I don't like the word "idiot". But a programmer who would use LLM codegen is a programmer with bad judgement. A programmer who has bad judgement cannot spot the errors made by LLM codegen. QED.
Anyway I already got what I wanted: Servo, the web browser which will replace Firefox, has *already* banned "AI" code contributions. So it's only a matter of time before Servo is complete enough for day to day use, and I can delete the AI-infected Firefox from my computer.
Points for "AI infected". Treating AI like a computer virus is a helpful concept.
-
Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.
They're not LLMs. They're trained on open data.
Should translation be disabled if the AI 'kill switch' is active?
The translation models are opt-in, because each language must be individually loaded. The same approach should apply to every other AI-adjacent function - those using remote services included. Especially those.
@firefoxwebdevs -
Don‘t „design a kill switch“. Just put all the slop features into seperate extensions.
Then see how many people will bother to install them, so you get a realistic idea for the actual demand.@duke_of_germany @firefoxwebdevs @zzt how can we cook the books by showing rational demand? too rigorous for the valley
-
@chillicampari @joepie91 fwiw I asked about translation because we're figuring out what to do specifically about translation.
@firefoxwebdevs @chillicampari @joepie91 I don’t think ML translation is what most people are thinking of when they’re complaining about AI. Machine translation has been around for over 20 years at this point, is fairly efficient, and while it makes mistakes (and those mistakes keep real translators if business for things that matter), it’s not the carbon spewing plagiarism machine that generative AI is. When I want an AI kill switch, I mean I don’t want my queries to create “summary” responses, or to add to a corpus that leaks my private information. Similarly, I want my radiologist’s CT software to flag potential issues, but I don’t want it to make up phantom blood clots, either.
-
@raymaccarthy @firefoxwebdevs @zzt I don't want a "browser experience". If it's doing its job, I won't be aware of it at all. I only use a browser as a viewer of content, period.
A browser should make websites viewable and allow the user to store locations in a way that makes sense to the *user*. Not a designer, not a bonehead CEO who thinks AI is really spiffy.
That's all it should do. It's very clear that browser execs never use tools. They have no idea what "tool" means.
@W6KME @raymaccarthy @firefoxwebdevs @zzt Exactly this. When you're eating, you don't want to be aware of your cutlery.
-
@mdavis @firefoxwebdevs I don't like the word "idiot". But a programmer who would use LLM codegen is a programmer with bad judgement. A programmer who has bad judgement cannot spot the errors made by LLM codegen. QED.
Anyway I already got what I wanted: Servo, the web browser which will replace Firefox, has *already* banned "AI" code contributions. So it's only a matter of time before Servo is complete enough for day to day use, and I can delete the AI-infected Firefox from my computer.
@mcc @firefoxwebdevs It is a shame that we’ve come to having to ban the use of some tools.
I used an unfortunate word choice, despite an apropos meaning in this context: an idiot is an utterly foolish or senseless person. Programmers should know how to properly use the tools they have. That’s why I’m not all against AI codegen. In the right hands, a tool can create something beautiful and useful. In foolish hands, it can damage.
Learn your craft first. Then use tools properly to enhance it.
-
Let's assume you're correct.
People only care about AI slop.Why is Firefox even running this survey? Like who cares? Translations aren't "AI slop", they don't need to be covered by the "AI Kill Switch"... why are they even asking this question?
Now take that assumption and read the rest of the comments. From what I'm reading, people care about more than just the AI slop. People are asking questions about the models being used for ML systems, systems that are incapable of generating AI slop.
So we're at a weird spot here. You believe that people care only about AI slop. Firefox obviously believes that people care about more than that, because they're running this survey. And people responding are asking questions that also indicate they care about more than AI slop.
So how do we square this?
What do you think is a better outcome for Firefox and the community?Why is Firefox even running this survey?
Because the people in charge genuinely believe that AI slop is The Future
and believe that, in order to stay relevant, Firefox must become an AI Browser.But somehow users inexplicably dislike AI slop?! How can this be?!
Embedding AI slop in Firefox as deeply and pervasively as possible is thus a critical goal. But this risks reputational damage with its actual users! To mitigate the risk, bundle features that were not controversial into the discussion of the controversial features; this serves to average the controversy across the (previously uncontroversial, existing) translation feature and highly controversial new slop features, hopefully reducing it below an ignorable threshold.
