Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Mozilla's open source AI strategy

blog.mozilla.org

Post reply on HN
Print view View original

131–140 of 213 posts

Re: Mozilla's open source AI strategy

#131

Post by giancarlostoro »

blibble wrote: post #100
dralley wrote: post #54

Earlier quoted context omitted.

> I think the problem many of us have is it feels like Mozilla invests more effort into everything other than its browser, we see the things they do outside of Mozilla, I mean it was so bad Thunderbird had to become its own foundation due to lack of funding! And then people simultaneously complain that Mozilla is reliant on Google for funding. There are not many good revenue options available for a browser other than…

> There are not many good revenue options available for a browser other than selling search defaults if they had set up an endowment instead of blowing it on unrelated pointless crap for decades they would have been self-funded indefinitely they were pulling in over $500 MILLION a year

500 million could net you about 50 to 70 million annually if you put it all on the S&P 500... A few years of this and you're a self-funded non-profit...

Re: Mozilla's open source AI strategy

#132

Post by capitainenemo »

mkroman wrote: post #129
capitainenemo wrote: post #124

Earlier quoted context omitted.

Mozilla did integrate TTS into Firefox - in fact, one of the better FOSS TTS AI models out their was their initiative. https://github.com/mozilla/TTS

Ah yeah, there's this: https://github.com/mozilla/TTS I can't seem to find anything that mentions a Firefox integration though?

Click Reader mode on a web page, then the read aloud option in the sidebar.

Note that how well it works on Linux will depend on your distro and default settings, as is common for Linux world. They do try to provide setup instructions if your linux distro has issues.

... now whether that model is integrated by default, no idea. I imagine that depends on size.

Oh, and mozilla's off-line translate for private translation of web pages... that's another neat AI thing they added that I've found super helpful. Chrome still requires sending the content to their servers.

Re: Mozilla's open source AI strategy

#133

Post by Yoric »

encom wrote: post #115
Yoric wrote: post #99

Earlier quoted context omitted.

I feel that, at some point around the Brendan Eich-gate, the Internet decided that Mozilla was always wrong. Change the shape of tabs? We received rape threats. Change it back? Bomb threats. Bringing in new APIs for add-ons that make Firefox faster, more secure, more stable and doesn't break all the time? No, we want addon $X, we don't care about security. I'm not going to claim that everything Mozilla has done is ri…

As a former Firefox user, I got fed up with the constant change for the sake of change. Why change the tabs? They were fine the way they were. People got mad about the addon situation since it broke their workflows because of vague technical reasons. And Mozilla usually ignored user protests while pointing at telemetry, and did whatever they wanted to, users be damned. At least that's how it looked from this side. I…

> And Mozilla usually ignored user protests while pointing at telemetry, and did whatever they wanted to, users be damned.

When I worked on Firefox, most of the changes happened exactly because user research determined that users wanted them and/or that not having them hurt the product. We changed the tabs at least once because users thought that the old shape of tabs made the browser feel slow (true story, sadly). We changed the add-on API (after having warned add-on developers for at least 6 years) because the old API was incompatible with multi-threading, multi-process, sandboxing, which in turn was really bad for both performance and security.

I'll absolutely grant you that Mozilla hasn't been very good at communicating these choices, but again, the sheer hostility of tech crowds is exhausting.

Re: Mozilla's open source AI strategy

#134

Post by Yoric »

encom wrote: post #111
Yoric wrote: post #97

Earlier quoted context omitted.

It's a really hard line to walk. If you put too much in your Telemetry/crash reports, yeah, users become fingerprintable. On the other hand, if you return spoofed values, it means that Firefox developers cannot debug platform/hardware-specific crashes. If you disable Telemetry, improving performance becomes impossible, because you're suddenly unable to determine where your users suffer. If you remove WebGL, plenty of…

Sure, but then don't go grandstanding about privacy. You can't have both. And saying that improving performance is impossible without it is hyperbolic. Developers did that before every major application turned into actual spyware. Profilers still work without it.

Profilers only work once you have identified the problem. Telemetry lets you find out about it in the first place.

Re: Mozilla's open source AI strategy

#136

Post by yogthos »

In my view, this is the exact right approach. LLMs aren’t going anywhere, these tools are here to stay. The only question is how they will be developed going forward, and who controls them. Boycotting AI is a really naive idea that’s just a way for people to signal group membership.

Saying I hate AI and I’m not going to use it is really trending and makes people feel like they’re doing something meaningful, but it’s just another version of trying to vote the problem away. It doesn’t work. The real solution is to roll up the sleeves and built an a version of this technology that’s open, transparent, and community driven.

Re: Mozilla's open source AI strategy

#137

Post by drnick1 »

Yoric wrote: post #97
drnick1 wrote: post #38

The open source community will start taking Firefox seriously again when all the AI shit is removed for good and real improvements to performance and privacy are made. Despite all the posturing about "respecting your privacy and freedom," the stock configuration of Firefox is trivially fingerprintable. At the very least, a privacy-focused browser should adopt the Tor patches and report standardized spoofed values for…

It's a really hard line to walk. If you put too much in your Telemetry/crash reports, yeah, users become fingerprintable. On the other hand, if you return spoofed values, it means that Firefox developers cannot debug platform/hardware-specific crashes. If you disable Telemetry, improving performance becomes impossible, because you're suddenly unable to determine where your users suffer. If you remove WebGL, plenty of…

> If you put too much in your Telemetry/crash reports, yeah, users become fingerprintable.

It's not only what gets send to Mozilla as telemetry or crash reports that is a problem. That can be turned off (many Linux distros do), or firewalled.

The main issue is that websites can more or less accurately identify users uniquely by extracting information that they should not have access to if the browser was designed with privacy in mind.

This includes, but is not limited to, fonts installed, system language, time zone, window size, browser version, hardware information (number of cores, device memory), canvas fingerprint, and many others attributes. When you combine all of that with the originating IP address, you can reliably determine who visited a website, because that information is shared and correlated with services where people identify themselves (Google accounts, Facebook, Amazon, etc.) Even masking your IP may not be enough because typically there is enough information in the other data points to track you already.

Re: Mozilla's open source AI strategy

#138

Post by mkroman »

capitainenemo wrote: post #132
mkroman wrote: post #129

Earlier quoted context omitted.

Ah yeah, there's this: https://github.com/mozilla/TTS I can't seem to find anything that mentions a Firefox integration though?

Click Reader mode on a web page, then the read aloud option in the sidebar. Note that how well it works on Linux will depend on your distro and default settings, as is common for Linux world. They do try to provide setup instructions if your linux distro has issues. ... now whether that model is integrated by default, no idea. I imagine that depends on size. Oh, and mozilla's off-line translate for private translatio…

Ah cool, thanks, didn't know this existed. I just get a dummy message when playing audio, so I'll play around with some speech dispatcher[1] solutions later!

> Oh, and mozilla's off-line translate for private translation of web pages... that's another neat AI thing they added that I've found super helpful.

Yes, it's awesome! And one of my favorite additions to Firefox in many years, it's stuff like that they should focus on if they want AI, imo.

[1]: https://wiki.archlinux.org/title/Speech_dispatcher

Re: Mozilla's open source AI strategy

#139

Post by drnick1 »

giancarlostoro wrote: post #131
blibble wrote: post #100

Earlier quoted context omitted.

> There are not many good revenue options available for a browser other than selling search defaults if they had set up an endowment instead of blowing it on unrelated pointless crap for decades they would have been self-funded indefinitely they were pulling in over $500 MILLION a year

500 million could net you about 50 to 70 million annually if you put it all on the S&P 500... A few years of this and you're a self-funded non-profit...

Reminder: Ladybird is being developed by a handful of people with contributions from the community. It's far from being complete, but it clearly shows that you don't need an enormous budget to build a Web browser entirely from scratch, let alone maintain one.
Post reply on HN