Firefox is expanding its AI-powered features, all designed to keep your data private. We believe technology should serve you, not monitor you. Our team und
all these AI features seem like the most useless waste of GPU I have ever seen implemented! Why the hell do I need recommendations for the names of my tab groups? I freaking love that the tech industries’ idea of progress in the 2020’s is wasting resources on nothing.
There are other practical things to use a browser-based AI for, like accessibility improvements for sites that didn’t provide proper alt text or other accessibility features.
But I think an important one few people are talking about is an AI “babysitter” for my grandma so she doesn’t fall for phishing scams. Ad blocking does a lot to protect people there but some smarter detection would be good for a significant chunk of society. Not that this exists yet.
Yeah, I think accessibility is one of the few potentially actually valuable use cases for generative AI.
If you can generate alt text completely locally in a private open source way that’d be pretty cool. It’d be nice if it funtioned as a extension that only does anything when you intentionally call it, at least by default. Maybe mapped to a keyboard shortcut, but I don’t know enough about what visually impaired users need to have a meaningful perspective on how the user experience should be implemented.
But I’ve yet to see any companies talk about how that’s what they’re gonna use AI for. To me, AI has hypothetical usefulness specifically for tasks that are really important but that in practice no one actually puts resources into, or has any resource to put into.
I also kinda wonder if local-only AI moderation tools could also make it a lot easier for fediverse mods to cover a lot of ground, or could abstract really disturbing content to reduce the mental load of moderating out the worst content folks post on the internet.
But no one is interested in AI tech so they can build useful things, theyre interested in it because they can steal peoples intellectual labor and build products without having to pay the people who would produce that intellectual labor.
You can make software accessible without Ai, they just don’t want to spend the money on it. It’s a business decision that every company makes to save the money on accessibility. It’s a calculated cost to them
Education is the answer and not creating a nanny state with fucking Ai babysitting and spying on everyone.
This is some big brother lvl of bullshit just because oh no old people might be stupid enough to fuck themselves over. If they are that technically incapable, maybe it’s time to set up a device without any extra options.
You might make software accessible without AI but I sure as heck don’t trust everyone else to do it right. Unless you control all software (spoiler, you don’t) then your point is merely theoretical.
Automated browser level accessibility to cover for other people’s errors is a perfectly reasonable solution to the problem that most developers are fucking lazy.
Plus, there is no “big brother bullshit” when it comes to locally run AI. Your data never leaves your device, and the model weights are freely auditable. You seem to fundamentally misunderstand how this works.
And you are fucking gullible to think giving Ai Agents root level access for the sake of accessibility is reasonable or that the big companies are even going to allow local models in the future.
Don’t put words in my mouth. Root level access is obviously stupid. AI doesn’t have to be owned by corporations. If you think all AI is the same, you are the gullible one - open your eyes and actually look at the options out there. Apache licensed models exist out there, and no corporation can control that.
You could say the same thing about all software - when it’s run by corporations it’s almost certainly there to harvest your data. But there is a ton of ethical software too!
I believe their point was that site developers should just include alt text, without ai ever being needed.
But in practice, sometimes they don’t, and that’s been an issue for ages. And in practice, a lot of images on the internet are user generated/submitted, and most sites don’t have a culture of users writing alt text for their images. Mastodon does, but even lemmy doesn’t really.
this education thing has been said and said again since what, the 90s? and guess what, people still fall for scams! for the same reasons that an antivirus is a crucial security component for the non-tech enthusiasts crowd, an AI scam detector that runs locally will likely be a crucial tool in the next few years.
TBF, I love my tab groups, and honestly having to come up with and type out a name for each of them is the biggest hurdle. It’s a small thing, but I can imagine myself using tab groups even more because of it.
I, on the other hand, don’t keep many tabs open normally, and the groups I use are of temporary convenience if I’m doing a lot of things at the same time.
I can easily spend a few seconds typing the group name in because I already know what the group is used for. I already have a name in mind, I don’t need the AI to slowly second-guess me.
…and if you don’t know what the group is used for, maybe allowing groups to remain initially unnamed would be a good design choice?
Oh speed is definitely going to be the make-it-or-break it here. I do usually know the name without thinking, but it’s having to type it (and preferably without typos) that’s the effort that would be nice to save.
It feels really stupid, because it’s really not that much effort. But I know myself - I also really like contactless payments, even though it really wasn’t that much effort to insert my card and type in my PIN. Yet it’s a small convenience for something that I do with some regularity. I imagine the same for this feature.
Honestly if it prefilled something like that that would already be an improvement, at least to create temporary ones (though in practice, they’re probably never temporary).
all these AI features seem like the most useless waste of GPU I have ever seen implemented! Why the hell do I need recommendations for the names of my tab groups? I freaking love that the tech industries’ idea of progress in the 2020’s is wasting resources on nothing.
Yeah, translation is the only thing here I have any desire at all to use
There are other practical things to use a browser-based AI for, like accessibility improvements for sites that didn’t provide proper alt text or other accessibility features.
But I think an important one few people are talking about is an AI “babysitter” for my grandma so she doesn’t fall for phishing scams. Ad blocking does a lot to protect people there but some smarter detection would be good for a significant chunk of society. Not that this exists yet.
Yeah, I think accessibility is one of the few potentially actually valuable use cases for generative AI.
If you can generate alt text completely locally in a private open source way that’d be pretty cool. It’d be nice if it funtioned as a extension that only does anything when you intentionally call it, at least by default. Maybe mapped to a keyboard shortcut, but I don’t know enough about what visually impaired users need to have a meaningful perspective on how the user experience should be implemented.
But I’ve yet to see any companies talk about how that’s what they’re gonna use AI for. To me, AI has hypothetical usefulness specifically for tasks that are really important but that in practice no one actually puts resources into, or has any resource to put into.
I also kinda wonder if local-only AI moderation tools could also make it a lot easier for fediverse mods to cover a lot of ground, or could abstract really disturbing content to reduce the mental load of moderating out the worst content folks post on the internet.
But no one is interested in AI tech so they can build useful things, theyre interested in it because they can steal peoples intellectual labor and build products without having to pay the people who would produce that intellectual labor.
You can make software accessible without Ai, they just don’t want to spend the money on it. It’s a business decision that every company makes to save the money on accessibility. It’s a calculated cost to them
Education is the answer and not creating a nanny state with fucking Ai babysitting and spying on everyone.
This is some big brother lvl of bullshit just because oh no old people might be stupid enough to fuck themselves over. If they are that technically incapable, maybe it’s time to set up a device without any extra options.
You might make software accessible without AI but I sure as heck don’t trust everyone else to do it right. Unless you control all software (spoiler, you don’t) then your point is merely theoretical.
Automated browser level accessibility to cover for other people’s errors is a perfectly reasonable solution to the problem that most developers are fucking lazy.
Plus, there is no “big brother bullshit” when it comes to locally run AI. Your data never leaves your device, and the model weights are freely auditable. You seem to fundamentally misunderstand how this works.
And you are fucking gullible to think giving Ai Agents root level access for the sake of accessibility is reasonable or that the big companies are even going to allow local models in the future.
Ai apologists get blocked
Don’t put words in my mouth. Root level access is obviously stupid. AI doesn’t have to be owned by corporations. If you think all AI is the same, you are the gullible one - open your eyes and actually look at the options out there. Apache licensed models exist out there, and no corporation can control that.
You could say the same thing about all software - when it’s run by corporations it’s almost certainly there to harvest your data. But there is a ton of ethical software too!
How do you propose generating alt text for images without some form of machine learning?
I believe their point was that site developers should just include alt text, without ai ever being needed.
But in practice, sometimes they don’t, and that’s been an issue for ages. And in practice, a lot of images on the internet are user generated/submitted, and most sites don’t have a culture of users writing alt text for their images. Mastodon does, but even lemmy doesn’t really.
this education thing has been said and said again since what, the 90s? and guess what, people still fall for scams! for the same reasons that an antivirus is a crucial security component for the non-tech enthusiasts crowd, an AI scam detector that runs locally will likely be a crucial tool in the next few years.
TBF, I love my tab groups, and honestly having to come up with and type out a name for each of them is the biggest hurdle. It’s a small thing, but I can imagine myself using tab groups even more because of it.
I, on the other hand, don’t keep many tabs open normally, and the groups I use are of temporary convenience if I’m doing a lot of things at the same time.
I can easily spend a few seconds typing the group name in because I already know what the group is used for. I already have a name in mind, I don’t need the AI to slowly second-guess me.
…and if you don’t know what the group is used for, maybe allowing groups to remain initially unnamed would be a good design choice?
Oh speed is definitely going to be the make-it-or-break it here. I do usually know the name without thinking, but it’s having to type it (and preferably without typos) that’s the effort that would be nice to save.
It feels really stupid, because it’s really not that much effort. But I know myself - I also really like contactless payments, even though it really wasn’t that much effort to insert my card and type in my PIN. Yet it’s a small convenience for something that I do with some regularity. I imagine the same for this feature.
This feels like satire (but I worry it isn’t).
Tab group 1
tab group 2
…
Honestly if it prefilled something like that that would already be an improvement, at least to create temporary ones (though in practice, they’re probably never temporary).