- cross-posted to:
- [email protected]
- [email protected]
- technology
- cross-posted to:
- [email protected]
- [email protected]
- technology
You must log in or register to comment.
What purpose is served by having AI built-in to the browser?
Baked in AI makes C Suite and shareholders happy. That’s about it.
The implementation doesn’t sound terrible.
- It’s opt-in
- It’s basically a sidebar chat window
So if you already use GPT for day-to-day, it may be a welcome experience. If you don’t, don’t opt in.
I’m skeptical of GPT add-ons, bit at least this was done in a low-bloat opt-in way (which allows Mozilla to bring in revenue (probably)).
Know what’s even more opt in? An official extension. Installed only if someone wants it.
I switched to LibreWolf and Mull a few months ago in preparation for this. I’ll come back to Firefox if the investors pull their collective heads from their asses.
“Move all my rf resesrch tabs to a new window”
You can skip search engines sometimes
I can skip the search engine and go right to the bad suggestions, finally!
Ah yes, the update nobody actually wants…
nobody
I think you might be in for a rude awakening.
For me it’s useful, depending how it’s implemented. Being able to say “summarize this article” or “summarize this ToS and call out anything that’s anti consumer” is how I use chatgpt
Here’s me asking it to do 1 thing in Python and it halucinating and repeating itself incorrectly every time.
Looks like the “local AI only” idea was purged in favor of some Big Tech stuff that can give Mozilla some fat cash for promoting their services! Mozilla’s second (or third idk at this point lol) downfall is looking really strong with all their recent decisions. WebKit is another independent engine that still doesn’t seem to suck in terms of enshittification but it’s basically not used anywhere except Apple ecosystem. Chromium is getting a full monopoly yay.
I do self host several AI applications for myself on a low end device and I think for most lowend even mid devices local AI is unfeasible. Nowadays is too much resource heavy and times are too long without high end devices.
For my computer generating a description of a picture (one of the firefox new features) could easily take up to 5-10 minutes with the cpu at 100%. That’s just not viable for doing while browsing.
Anyway I would love for firefox to open source the server side of this. So in case someone have s computer powerful enough they could do it locally if they want to.
Still adding proprietary and actually evil AI providers is a questionable decision.
Well I’m guessing they actually did testing on local AI using a 4GB and 8GB RAM laptop and realized it would be an awful user experience. It’s just too slow.
I wish they rolled it in as an option though.
They wanted to use fast small language models, not LLMs like Llama
Llamafile with tinyllama model is 640mb. It could be a flag to enable or an extension
WebKit does exist for Linux, Gnome Web has been quite a nice experience however it still lacks support for most extensions (however some Firefox extensions do work). The real world performance is still a bit lacking but its close to Firefox on paper and as it continues to update I will probably swap to it. For now its a nice way for me to test if my websites will break on macs (spoiler, WebKit still lacks some stuff).
I know about that. I used to use Epiphany myself. The problem is that it’s unpopular, still not nearly as good as the other options and there’s no cross-platform support. The last one is a big problem because 90% of the market uses Android or Windows.
Cross platform and popular I agree with. Having it in a state where it could be the default for gnome distros would help with popularity. However I think at least in latest versions its pretty comparable to other browsers at least Firefox. Main issue is there isn’t as much extensions that work with it. Considering the pace it is improving though I think it won’t be long till it could be viable alternative at least on Linux, maybe it might get ported some day idk.
I have a pretty slow machine and GNOME Web is unusable on it. The performance is not comparable to Firefox or Chromium. Extensions are very important for regular people apparently (judging from old Firefox Play Store reviews). Also Firefox and its derivatives are known for advanced privacy features that GNOME Web doesn’t have and likely will never have because GNOME is about extreme simplicity by all costs.
I like GNOME Web and I really can see myself installing it on an old person’s computer because of awesomely simple UI. But it’s not for most people and I’m afraid that without commercial support there won’t be any good regular browsers based on WebKit. I wish engine-specific features didn’t exist. Everything would be so simple without them.
That’s fair, I haven’t tried it on low end/older hardware. I only just found performance good enough in the 46 release so I’ve only tested on my high and middle end system. I have some n200 hardware arriving soon and I might give it a go on that.
Advanced privacy and security I agree with and that’s the main reason I don’t use it daily personally. I think better extension support would be a good step in enhancing that even if they keep the base simple. There is also non trivial issues such as fingerprinting which is going to be a lot easier on a browser with so little users.
Firefox does currently have a few more options and I don’t see Gnome Web getting that ootb any time soon. Granted half of firefox’s options these days is to disable telemetry from Mozilla, the actual user exposed options isn’t huge (outside of about:config). Gnome does have gsettings which could serve a similar usage as already seen with enabling web extensions.
I don’t think it will be mainstream any time soon not until Linux is or they support other oses. But I want to be optimistic on how it will be for Linux usage especially with the tablet and mobile scene starting to take shape and Gnome Web being one of the most viewport responsive browsers.
Their new CEO is a McKinsey consultant so this is pretty much guaranteed.
Who is McKinsey?
So it isn’t even local private AI but rather just an Interface for NOT-private LLMs like ChatGPT (which specifically stated, at least at first, that all your queries to it and their responses are being monitored and saved by OpenAI)
Why are they not using their own llamafile ?
