However, a strong locally-executed AI would have potential to vastly improve our experience of web! So much work is done in browsers could be enhanced or automated with custom agents. You'd no longer need any browser extensions (which are privacy nightmare when the ownership secretly changes hands). Your agents could browse local shops for personalized gifts or discounts, you could set up very complex watches on classified ads. You could work around any lacking features of any website or a combination of several websites, to get exactly what you seek and to filter out anything that is noise to you. You would be able to seamlessly communicate with the Polish internet subculture, or with Gen Alpha, all without feeling the physical pain. With an AGI-level AI maybe even the Reddit could be made usable again.
Of course this is all assuming that the web doesn't adapt to become even more closed and hostile.
Image search?
Live captions?
Dubbing?
Summary?
Rewrite text better?
Image search? I have a search engine for that.
Live captions? Didn’t ask for that, wouldn’t use it.
Dubbing? Ditto.
Summary? Wouldn’t trust an AI for that, plus it’s just more tik-tokification. No fucking thanks. I don’t need to experience life as short blips of everything.
Rewrite text better? Might as well kill myself once I’m ready to let a predictive text bot write shit in my place.
So… no thanks.
The worst is anything that tries to suggest stuff in text fields or puts buttons etc. to try and get you to "rewrite with AI" or any nonsense like that - makes me just want to burn anything like that to the ground.
> Image search? I have a search engine for that.
Plus, Search by Image[0] is one of the most popular extensions, with 3x as many people using it as tree-style tabs.
I don't use it but a grammar tool is the next most popular[1], so I could see this being quite a useful feature.
But the other stuff, I'm with you. I like translate but I personally don't care for dubbing, summarizing, or anything else.
[0] https://addons.mozilla.org/en-US/firefox/addon/search_by_ima...
[1] https://addons.mozilla.org/en-US/firefox/addon/languagetool/
- non native speakers - moving away from the english-centric web - impaired people
Or searching for text in images with OCR. Or searching my own browsing history for that article about that thing.
Newegg has that as a built in filter.
Why do you people keep insisting I "need" an LLM to do things that are standard features?
I find shopping online for clothes to suck, but there's nothing an LLM can do to fix that because it's not a magic machine and I cannot try on clothes at home. So instead, I just sucked it up and went to Old Navy.
Like, these things are still lying to my face every single day. I only use them when there's no alternative, like quickly porting code from python to Java for an emergency project. Was the code correctly ported? Nope, it silently dropped things of course, but "it doesn't need to be perfect" was the spec.
>Or searching for text in images with OCR.
That thing that was a mainline feature of Microsoft OneNote in 2007 and worked just fine and I STILL never used? I thought it was the neatest feature but even my friend who runs everything out of OneNote doesn't use it much. Back in middle school we had a very similar Digital Notebook application that predates OneNote with a similar feature set, including the teachers being able to distribute Master copies of notes for their students, and I also did not use OCR there.
The ONE actual good use case of LLMs that anyone has offered me did not come from techbros who think "Tesla has good software" is not only an accurate statement but an important point for a car, it came from my mom. Turns out, the text generation machine is pretty good at generating text in French to make tests! Her moronic (really rich of course, one of the richest in the state) school district refused to buy her any materials at all for her French classes, so she's been using ChatGPT. It does a great job, because that's what these machines are actually built for, and she only has to fix up the output occasionally, but that task is ACTUALLY easy to verify, unlike most of the things people use these LLMs for.
She STILL wouldn't pay $20 monthly for it. That shouldn't be surprising, because "Test generator" for a high school class is a one time payment of $300 historically, and came with your textbook purchase. If she wasn't planning on retiring she would probably just do it the long way. A course like that is a durable good.
> Translation?
> Image search?
> Live captions?
> Dubbing?
> Summary?
> Rewrite text better?
Cool, and some DEs make it possible to start implementing this for most applications today. But Mozilla is not KDE or Gnome, so the most they can do is to make this on their software, and make it easy to copy for the entire system.
> Sounds like a web site, not a browser feature.
Sounds like a bit of lack of imagination on your part. Do you think the same for text search?
>
Translation specifically was pretty bad before Google applied machine learning methods to it around 2007 when it became very good almost overnight.
They just existed before the GenAI craze and no one cared because AI wasn't a buzzword at the time. Google Translate absolutely was based on ML before OpenAI made it a big deal to have things "based on AI".
But just putting stuff in your browser that hooks into third-party services that use ML isn't enough anymore. It has to be front and center otherwise, you're losing the interest of... well, someone. I'm not sure who at this point. I don't care, personally.
Mozilla having unique features is what made it popular in the first place (tabbed browsing versus IE6).
These stories just look compelling and obvious in retrospect, when we can see how the dice landed.