Share your feedback on the AI services experiment in Nightly

[asafko]

Hi folks, 

In the next few days, we will start the Nightly experiment which provides easy access to AI services from the sidebar. This functionality is entirely optional, and it’s there to see if it’s a helpful addition to Firefox. It is not built into any core functionality and needs to be turned on by you to see it. 

If you want to try the experiment, activate it via Nightly Settings > Firefox Labs (please see full instructions here). 

We’d love to hear your feedback once you try out the feature, and we’re open to all your ideas and thoughts, whether it’s small tweaks to the current experience or big, creative suggestions that could boost your productivity and make accessing your favorite tools and services in Firefox even easier.

Thanks so much for helping us improve Firefox!

[Sztefa001]

It's not optional or opt-in or whatever, that would mean it's like an extension you can add, not an already-in-just-choose-from-menu thing. Even then it would be unethical as it would be Mozilla saying "hey, if you wanna steal from others and waste resources there's a handy littel AI addon we have!"

What we're being proposed rn is the exact same tbh, just worse 'cuz it's not a true opt-in.

[Liz]

**bleep** dude you are really hell bent on your bootlicking, huh?

[SoliTheFox]

I want this.

The feature is basically a "fork" of the official sidebar extension, so it is the same as another tab open. There's no "AI" "wasting resources" in the background.

The feature is opt-in, and even if opted-in, is only used when you actually want to use it (select text and click on a button, then it opens the sidebar on the selected provider). If you don't opt-in, select a provider, select some text and click on a button, it is not using a single bit of extra ram, or CPU.

Really recommend actually reading about and maybe exploring stuff before hating on it for free, buddy

[Sztefa001]

Pretty sure the issue about resources is not about one's specific PC but rather AI engine servers that waste a lot of water to cool down and being generally bad for environment. It's not about CPU, it's about Earth.

[llliiimmmeee]

Hi! I love this feature. It's insanely helpful. Things I think could be improved:

• ChatGPT, which I've been using, starts a new chat every time I use the button that appears when you highlight text. I would prefer if there was a toggle to change this behavior.
• Although a complete overhaul, I think it would be nice to have an option for people with OpenAI API keys to use a more firefox-like user interface.
• I would prefer if I could add my own custom prompts to the highlight menu from preferences.
• I think that a button to minimize the chat until I press a sparkles button next to Firefox View would be amazing.

Here are some things I hope you keep the same:

• Opt-in rather than opt-out.
• Using the sidebar rather than opening a new tab.
• The option to keep it in the context menu rather than when I highlight text.
• The variety of different chatbots (I hope that this will expand to allow more healthy competition to Open(Closed)AI!)

[timbro]

the gemini ai tab is using the wrong google account and wont let me switch it

[Andreslav]

Hello everyone

I liked this feature, in particular, the buttons that appear when you select text.

It is necessary to provide for cases when users use plug-ins (1) with similar pop-up windows. It may be necessary to provide developers with an API to add their icon near the cursor, but for Firefox to decide where it will be displayed.

You can also add the option to disable this pop-up window to the AI settings.

(I haven't read all the messages, so I don't know if it was written about earlier. I'm sorry).

[Unrealgeek24]

Hey, quick suggestion if an AI sidebar ever makes it into an official release of Firefox. For me, it would be helpful to be able to completely turn it off, as other browsers like Edge give you a hard time when trying to turn it off. I personally never use AI, so having a sidebar of some sort taking up space for no reason would be counter intuitive to me. It would be nice and helpful to know that you won't do what Edge did and shove AI down my throat. Note: I'm talking about an official release, I know that right now this sidebar is only an experimental feature and can be turned off or on at any time.

[Starlightwars]

It's already known that these generative "AIs" use massive and detrimental energy usage, the cooling systems also use gallons of water per prompt leading to mass evaporation!
all for a crappy chat-bot
it's just not worth it.

[wutongtaiwan]

AI features should be turned off by default and opted in by the user.

[Dandelion44]

firefox just follows what's trend on the internet instead of actually making useful update. What's next update, skibidi in firefox? because it's current trend

[Dandelion44]

Reading replies... I just noticed how firefox employee replies to those who like AI near instant. But when someone disagrees with AI, no response from them.

[SoliTheFox]

When most of the replies show people see the word "AI" and instantly hate on it without even understanding how the feature actually works, yeah, that's what's gonna happen...

I've seen so many replies that are simply "I hate AI, take it out or else gonna leave Firefox" and replies that clearly show that people don't even know **bleep** about what they are actually talking about, I would be surprised if he is even reading the comments anymore.

[Dandelion44]

You just trying to make AI haters look like stupids. Yeah we exactly know what we are talking about. Did ChatGPT wrote this reply for you?

[SoliTheFox]

Then explain me how the feature works.

Or show me the source code of it, and tell me where exactly "AI" is being trained, or used for inference using the user's resources. I already found the source code for you: https://searchfox.org/mozilla-central/source/browser/components/genai/GenAI.sys.mjs

[nokko]

The complaint about using up system resources is incorrect. However, the ethical concern over tacitly approving the unauthorized scraping of web pages on a mass scale is what most people are riled up about.

[SoliTheFox]

I can completely see this point and also the ambiental aspect point. Somewhat agree myself.

But I really do not think most people understand how the feature actually works. Really feels like people think this feature is actually scraping your data, from the browser, unless you completely remove it from the source code. Feels like people are just putting everything "AI" on the same box and hating on it for the sake of hating, when it can be good specially when it is not used as a source of information, and mindfully. For example, I use open weights models to summarize stuff or check if a certain page contains what I actually want to know about. I do not trust it 90% of the time, as it can hallucinate.

It also can also be improved ethically. For example, The Stack dataset allows you to check if your repo was scraped, and to fill in a form to remove your repository from it.

[doodlemancy]

---

@SoliTheFox wrote:

For example, I use open weights models to summarize stuff or check if a certain page contains what I actually want to know about. I do not trust it 90% of the time, as it can hallucinate.

---

I'm sorry, but why are you using a "tool" that you apparently can't trust 90% of the time? You're double-checking its work anyway. 🙃 It sounds like you're actually making more work for yourself.

You know what can't hallucinate? Ctrl + F.

[SoliTheFox]

Can ctrl-f summarize entire pages?

Can ctrl-f read an entire page in 2 seconds and tell me if my super specific problem is covered inside the page? Or should i prefer to ctrl+f through all of the 72 ocurrences of the word (8 of them on the same paragraph) related to what i'm looking for in order to get what i want?

Can ctrl+f merge all of the code blocks into one so i can actually figure out in 2 seconds if i'm reading the pytorch documentation all over again, written with different words, or something that will actually be new for me?

Idk, ctrl+f seems really easier to use and more powerful when you know exactly what phrase to look for to go straight to where it covers what you want. Also, 90% of the times is not even necessary to check anything if you are using 8b models instead of 2b models. Smaller models hallucinate more frequently, but they also give me an instant answer.

And honestly, there's a reason why people prefer rendering html instead of reading the raw file, even though rendering a webpage is extra work. Ease of use is important.

[SoliTheFox]

And it really do seems like your own words apply to you:

"The dismissiveness and arrogance with which you responded to this [...] person, who gave you their ...

[Dandelion44]

I do not have to tell you anything but if this means you shut up afterwards, then here we go:

What AI (ChatGPT, Cladue, others) is just an improved version of autocompletion.

Where exactly AI being trained: Many text on internet, including copyrighted content and code, it's used as soon as OpenAI is able to web-scrape it.

Also, you don't scare me when you show me code snippet of what you find over open-source projects, another funny thing is that the code you show is UI of Firefox's AI integration, nothing actually related to AI itself.

[SoliTheFox]

Ok bozo, make the argument about "AI" in general when we are discussing the Firefox feature here. Considering your response, it seems you either have serious problems of text interpretation, or you are misinterpreting my response in bad faith. Any chatbot can help you with that tho, unfortunately we are left with just your intelligence here.

I repeat my question again: where is the new Firefox feature being trained on user's data? Does the inference use any computer resources from the user? What are we discussing on this thread: "AI" in general, or the new Firefox feature?

Like i said, all i want is you to show me how this feature is actually bad other than the "moral high ground" arguments. After all, you guys want this removed, so i wonder how it is actually a bad feature that worsens the user's experience. Does it make your computer any slower? Does it have any glitch or bug that affects other areas of the browser? Is your experience less smooth now in terms of UI/UX? Does it pop up everytime you use ur browser? I ask this because, like i said in my initial comment, i've seen many people claim this.

[tfunken]

Just no!

[Bersl]

• Never try turning it on by default.
• Never prompt me to turn it on.
• If turned off, leave absolutely no trace elsewhere in the UI.

My continued and continuous use of Firefox, which dates back to building from source before the software was even called "Firefox", depends strongly on never being reminded that this is a thing that exists, and I do not recommend pushing your luck any further than you've already done. I already terminated financial support over the fact that you're even wasting resources on this particular hype bubble (and will consider re-instating it if I feel that the madness has stopped), and I'm asking you with all the politeness I can muster not to gamble more aggressively or passive-aggressively with my tolerance in this matter.

Thank you.

[mcc]

I demand a version of Firefox that does not have this feature. It makes no difference to me that the feature can be disabled, or even whether it is disabled by default. The presence of the code to run this feature on my computer is unacceptable. If you can't provide me a version of Firefox that has the AI chatbot feature *removed* (not just disabled), or in which the AI chatbot feature is an uninstallable plugin, then I am going to switch to LibreWolf or whatever fork of Firefox will remove the AI features for me. 

I. Do. Not. Consent. To AI software being installed on my computer. Period. It wasn't okay when Microsoft forced an AI client onto my computer, and it's not okay when you do it. I switched from SwiftKey to GBoard to remove the AI client, I am switching from Windows to Linux to avoid the Windows 11 AI client, I will switch from Firefox to [whatever I have to] to remove the AI client.

You should not have done this, and I cannot support you as a company until you stop.

[PleaseMozilla]

I believe LibreWolf (Firefox fork that focus' on being more private by default) might not have this feature, not entirely sure but I know I will probably look into it!

[tfunken]

As of right now, this feature is also present in Librewolf and Zen Browser.
Not sure if they will remove it in the future.

Kinda wish we had a "Unmozillad Firefox", similar to the Ungoogled Chrome.
But sad that we need that in the first place 😥

[yayroos]

librewolf came to a decision on removing it today, it'll be gone soon i expect

[mariduv]

Yeah, this sucks.  People who want this could already have used extensions, but Firefox integrating this even as a lab, and letting one guy out here carry water for it, for months, is yet another move that undermines trust in Mozilla.  Someone could have been working on bug fixes or features that don't confidently misinform children.

[frenshape]

Please, no more AI chatbot stuff. I do not want this in my browser.

[karakara]

I am vehemently opposed to a "feature" of this nature in Firefox. Please, if you care about the users of Firefox at all. Drop this immediately.

[Cipscis]

I'm incredibly disappointed that this feature has been added to Firefox. I hope it will be removed immediately in the next build, and that Mozilla won't go down this route again.

One of the reasons I changed my primary browser to Firefox is that its major competitors available on Windows - Chrome and Edge, owned by Google and Microsoft - were going all-in on LLM-style "AI" en**bleep**tification. I'm certain I'm far from the only Firefox user who can say this.

Other users have already patiently explained the problems with features such as this. From the complete lack of respect for users' privacy that comes with using models that scraped their learning data without permission, to the complete lack of ethical concern that comes with shipping a feature that will undoubtedly lie to your users. This is a bad idea, both as a feature in general and in terms of the effect it will likely have on Mozilla's reputation as a trusted and privacy respecting alternative to other major browsers. Being able to disable it is not enough, this feature should not be shipped with Firefox.

[isopod]

I'd be very interested to know in what way Mozilla expects people to find this useful.

[mcc]

(The comment you are currently reading was posted by accident. If the moderators can delete it, they should.)

[PrismLizard]

-1. I am vehemently opposed to the addition of any form of LLM or chat bot in Firefox.

The current implementation raises several major concerns.

1. There is zero information about the privacy implications of using a third-party service. The omission of a privacy statement is a show-stopping red flag.
2. The way this is built severely limits user choice. Among other issues, you cannot choose to use the chat bot in a container dedicated to its use.
3. Because this runs as a privileged feature of the browser, it's difficult (if not impossible) to examine what requests and data are being sent. You are asking users to trust that nothing shady is going on without giving them the means to verify.

The recent update that adds weather information to Firefox already demonstrates that Mozilla no longer can be trusted to keep my personal information private. Sharing my location with a third party without my explicit consent has already broken my trust. If an LLM has privileged access in the browser, especially an LLM operated by a corporation with a profit motive, then there is nothing stopping them from stealing and using my information aside from easily-broken promises.

If this experiment is continued, I will switch to another browser that actually provides the level of privacy and security that Firefox was once known for. I am not comfortable with it being optional; there is already precedent for new options being enabled without user consent. I want the functionality completely stripped from the browser. I want Mozilla to go back to Firefox's roots: A minimal web browser that users can customize with extensions.

[mgztUcWsVXZNptW]

Many people have made it clear AI should never be added to Firefox, and I want to add my two cents: I believe anyone who works for Firefox / Mozilla in any capacity that believes AI belongs in Firefox (or any Mozilla product) should resign immediately. The history of Firefox is all about privacy and that's why many people continue to use it. AI is not private and should never exist. If AI is integrated any further, I expect Firefox to lose tons of users and their respective donations.

As a website owner, I believe AI should never exist and all the data that any models has collected has been collected illegally.

[ephemeris]

The environmental effects of "AI services" content is appalling - it's cryptocurrency all over again, but worse. Mozilla's effort on this feature in Firefox is profoundly depressing. Please stop.

[doodlemancy]

So-called "AI" technology is unreliable, wasteful, and unethically built on the stolen work of smart, experienced and creative humans who never consented. It does not, and can not ever, do what it claims to do, other than give overpaid sentient office chairs a reason to lay off humans from their jobs and then re-hire them as contractors later. I don't care if it's optional. I don't care if I can turn it off or ignore it. AI is an absolute no-go. The very fact that you are treating this garbage as if it is worth any legitimate or serious interest outside the realm of legitimized corporate fraud is a testament to how far Mozilla has fallen in recent years. It's Herbalife for mediocre tech bros, and you should be ashamed to be associated with any of it.

I've ushered many people away from crappier browsers to Firefox over the years. I've also been been glad to financially support Mozilla in the past, most recently through paying for Mozilla VPN. That will end if you decide to engage with this grifter nonsense. End of discussion. You will not see another dollar or a single word of support from me ever again if AI integration makes it out of the nightly build. 👎

P.S. Don't play around and pretend that "spellcheck is actually AI" or whatever. You know exactly what people are calling you out for here. You're just trying to see what you can get away with. It's pathetic, it's transparent, and it won't absolve you of responsibility.

[Rangal]

Yeah this sucks lmfao

[Shadlock0133]

don't waste time with AI gimmick that's gonna run out in ~2 years, just like nfts and blockchain before. AI is objectively bad at most tasks, bad for environment, bad for artists and writers and other creative folks, and it's gonna be bad marketing when it flops.