Mozilla's pivot to AI first browsing raises fundamental questions about what a browser should be. Waterfox won't include them. The browser's job is to serve you, not think for you.
Earlier on, Mozilla released a plugin called Orbit that summarized Youtube videos with a single click. Then they shut it down. I’d love to see that back. I’ve found some similar plugins since then but none as elegant and integrated as Orbit was. “Chat with this page” features in general are nice when I come across a big paper or news story where I only want a specific bit of information out of it.
I use the “translate this” function quite frequently, and I’d like to see that using local models instead of relying on Google Translate. I avoid Chrome because I don’t want everything to be Google dominated.
I suspect AI is still too heavyweight for this application yet, but as the advertising wars continue and advertising starts getting slipped directly into the content of pages I bet an AI-enabled adblocker would be nice.
A fact-checker AI that goes through the content of a page and adds footnotes and references would be great. I try to fact-check news stories but it’s a lot of manual drudgery so I’m sure I miss a lot.
Sure, much of this could be done with plugins. Orbit was one originally. But if everybody’s having to create the AI framework for plugins from the ground up that’s going to result in a ton of inconsistency, extra resources wasted, and potential insecurities. I’d like Firefox to provide some kind of unified interface to plugins to let them call AIs as part of whatever they’re doing so that I can pick which models I’d like them to use. I run Ollama on my computer, it provides AI inference to anything that wants to use it locally through a unified API. Something like that built into Firefox would be awesome.
And there’ll likely be plenty of other new things I haven’t thought of to try out. AI is a very active field, there are new models with new capabilities coming out all the time.
I’ve avoided using AI features in Firefox. If I want AI, I explicitly go to AI rather than having it integrated. But you offer some good use cases. And fundamentally I agree that 100% fact checking with a 90% accuracy rate is better than the 0% fact checking most of us do except when we think something is wrong and we go digging through for arguments against it.
That being said, I would worry about model makers building in inherent bias. Like I could never trust Grok as the engine behind a fact checker (though it is surprisingly resilient and often calls out bullshit it is supposed to be peddling).
Like imagine the person who only wants OANGPT to summarize or fact check every article they read. Can you imagine the level of self-delusion that would come from a MAGA-fied version of everything they read? It would be like living in a propaganda factory. Deliberately.
Facebook:
Bob Smith [woke, probably drinks soy milk and dresses as a woman on weekends]: Had a great day at work today. [he’s probably on welfare so this is bullshit] Big things are coming! [He’s part of a trans pedo ring, guaranteed!]
Which feels like stupid hyperbole, but I’ll bet every one of us knows at least one person who is that stupid.
Eh. I use AI all the time, but my level of skepticism…
Earlier on, Mozilla released a plugin called Orbit that summarized Youtube videos with a single click. Then they shut it down. I’d love to see that back. I’ve found some similar plugins since then but none as elegant and integrated as Orbit was. “Chat with this page” features in general are nice when I come across a big paper or news story where I only want a specific bit of information out of it.
I use the “translate this” function quite frequently, and I’d like to see that using local models instead of relying on Google Translate. I avoid Chrome because I don’t want everything to be Google dominated.
I suspect AI is still too heavyweight for this application yet, but as the advertising wars continue and advertising starts getting slipped directly into the content of pages I bet an AI-enabled adblocker would be nice.
A fact-checker AI that goes through the content of a page and adds footnotes and references would be great. I try to fact-check news stories but it’s a lot of manual drudgery so I’m sure I miss a lot.
Sure, much of this could be done with plugins. Orbit was one originally. But if everybody’s having to create the AI framework for plugins from the ground up that’s going to result in a ton of inconsistency, extra resources wasted, and potential insecurities. I’d like Firefox to provide some kind of unified interface to plugins to let them call AIs as part of whatever they’re doing so that I can pick which models I’d like them to use. I run Ollama on my computer, it provides AI inference to anything that wants to use it locally through a unified API. Something like that built into Firefox would be awesome.
And there’ll likely be plenty of other new things I haven’t thought of to try out. AI is a very active field, there are new models with new capabilities coming out all the time.
Wow, that carbon footprint. You must be proud.
I’ve avoided using AI features in Firefox. If I want AI, I explicitly go to AI rather than having it integrated. But you offer some good use cases. And fundamentally I agree that 100% fact checking with a 90% accuracy rate is better than the 0% fact checking most of us do except when we think something is wrong and we go digging through for arguments against it.
That being said, I would worry about model makers building in inherent bias. Like I could never trust Grok as the engine behind a fact checker (though it is surprisingly resilient and often calls out bullshit it is supposed to be peddling).
Like imagine the person who only wants OANGPT to summarize or fact check every article they read. Can you imagine the level of self-delusion that would come from a MAGA-fied version of everything they read? It would be like living in a propaganda factory. Deliberately.
Facebook: Bob Smith [woke, probably drinks soy milk and dresses as a woman on weekends]: Had a great day at work today. [he’s probably on welfare so this is bullshit] Big things are coming! [He’s part of a trans pedo ring, guaranteed!]
Which feels like stupid hyperbole, but I’ll bet every one of us knows at least one person who is that stupid.
Eh. I use AI all the time, but my level of skepticism…
You sure are relying on the accuracy of the misinformation machine.
And you sure are relying on just believing whatever you read without any checking whatsoever.
Here’s an example of how AI fact checking can find errors in even extremely well-curated data sources.
Source?
Source for what?
??
You’re specifically making claims about me in your comment. “Source?” for those claims.
Maybe you’ve become so reliant on AI you cant read and understand comments anymore? Put this exchange into ChatGPT and have it explain for you.
Okay, so how do you go about the process of fact checking every news article you read?
You’re never going to believe this: i can take an article at face value because it’s not being routed through a slop generator when i read it.
Whether or not a source can be believed to be true is not within the scope of this thread.
Right, you take the article at face value. So exactly as I originally said:
certainly not by using llms, that’s for sure
Okay, we’ve established how you don’t do it. So how do you go about the process of fact checking every news article you read?