Hacker News Reader

Comment by vouaobrasil | Hacker News Reader

In my opinion, the goal of increasing automation has gone from a useful relief of labor to an obsession that will lead nowhere. Globally, this seems to be reflected only by increasing resource use and increasing ignorance of the natural world by producing purely for the sake of making consumerism more efficient.

AI assistant in a browser...it's getting ridiculous.

View on HN

The browser will be even less under our control than before, surveillance will go to the next level, and your private data will land in even more obscure places than before.

Ridiculous maybe, but by no means pointless. “AI” will be the end of ad blockers and a whole host of measures to secure a remnant of privacy on the net.

> “AI” will be the end of ad blockers...

On the contrary, I feel like AI--even if it doesn't advance much further than the level we already have achieved--is putting us at the cusp of finally being able to have good end-game fully-working ad blockers: I'll just have an AI in my browser read / look at / watch / "experience" your content and then launder away any and all ads in its presentation to me... I can even have it make extreme edits if required, rebuilding the audio and video, to remove subtle bias your content wants to infect me with due to sponsorship deals.

What is fighting to prevent ad blockers is not AI... it is pervasive DRM: if we live in a world where the tech platforms continue to side with Big Content and prevent us from being able to capture and alter media as we see it on devices they insist we "buy" and yet never "own"--and where the business models of the most powerful companies rely on proving that an end-user human affirmatively was on the other side of an interaction (and thereby might have been infected with paid propaganda)--we are stuck inside of a dystopia :(.

That's definitely a valid way of looking at it. I do hope your first paragraph comes to pass.

Your "end-game" AI tool sounds good, and would work I think, but I think the opposing endgame strategy for advertisers is fully verified DRM chain all the way from incoming network packets to the display's pixels.

Network -> Web Environment Integrity -> HDCP. No room in there for an AI middleman. :(

Indeed, that's the endgame. End-to-end DRM, and there's enough money behind it that it will happen. I've been advocating heavily against DRM for many years now and nobody cares. Occassionally when somebody has a need in the moment where they are annoyed, and I explain "you can't share <thing> with your spouse/sibling/friend because DRM" they will say something like, "oh yeah, that's annoying" but just continue paying for DRM-ed media.

Unless perhaps people boycott, but that sure ain't gonna happen (in anywhere near large enough numbers to make a difference), Apple, Google, and Microsoft will build it into their platforms, and anyone not on the big three will be unable to consume more and more as time goes on, including most web browsing. It won't happen overnight. It will be an iterative/progressive encroachment where just a little changes each day. Boiling frogs and all that.

Who knows, maybe the climax of the advertising/drm war will result in an ideological split of the Internet:

On the one side: Big tech. E2E DRM, Chromium is the only browser engine sites will respond to, unavoidable and unblockable ads, siloed content and whitelisted domains only. Verified real identities. Linux clients and VPN IPs completely banned or pushed to captcha hell.

The rebels: Fediverse, the small web, fringe projects, experimentation, perhaps some "indie" ads. (And the problems that come with it: Spam, scraping, etc.)

I know which part I'd like to participate in.

While I clearly generally do agree with you--this is why I have that second paragraph about pervasive DRM being the enemy here--and have spent much of my life fighting DRM in various ways, I will at least try to offer you a ray of hope: building functional DRM gets much harder as the analog loophole gets stronger, and now that we are nearly at (not quite, but so damned close to) being able to strap a "retina" screen over our eyes and implement real-time AR on our perceptions, to build DRM that functions is going to end up requiring AI detectives that suss out whether we are actually there or not... it will be ridiculous if we get to the point where instead of merely putting a piece of tape over our webcam we find ourselves having to hang a tiny screen in front of it designed to pretend to the computer that we are awake / present / not wearing a headset.

> requiring AI detectives that suss out whether we are actually there or not

Please drink verification can to continue

What do you mean that the analog loophole gets stronger? Are you saying that the AR technology is getting good enough that ad-blocking can be done there instead? That's true, I guess.

The analog loophole is the idea that I can always just use a camera and/or a microphone to record any content I want to pirate, rather than rip the original exact digital bits. The reason why this resulted in low quality results in the past is that cameras and microphones lose fidelity of the original source... but the technology behind cameras and microphones just gets better over time, and there is only so good that the original digital content with the original hardware playing to the human perception system was going to be in the first place, so if you point a good enough camera at a video and re-render it on a high-quality enough display you don't care that it goes through that analog step anymore, making DRM implemented inside of display protocols meaningless (and yes: opening the door to using the analog loophole to pirate and then modify/remix everything you see and experience in real time using AR).

>: I'll just have an AI in my browser read / look at / watch / "experience" your content and then launder away any and all ads in its presentation to me... I can even have it make extreme edits if required, rebuilding the audio and video, to remove subtle bias your content wants to infect me with due to sponsorship deals.

Ehh sure, but then websites have no incentive whatsoever to publish content. The future you're describing is one where:

- Only the big players operate (more centralized)

- Only walled gardens, because who would publish content publicly for everyone to ingest and then modify, with no gain whatsoever?

- More arms race to create DRM to prevent ad blockers, or deals with browsers not to tamper with specific sites (so, more centralization)

What you're proposing makes no sense, and will only deteriorate things further in the long run.

I not only do understand that DRM is what fights against ad blockers and would inherently be expanded (and that that would suck), that was my point: remember that I was responding to someone who was anti-AI as they felt AI marks the end of ad blockers, and yet I think the opposite is true. The consequences of these wars playing out in either direction and whether you like the result is a different scope, one where I don't even disagree with much of what you are saying...

...with two (small) exceptions: you fail to analyze a world where we all end up having to just pay for content as we access it--which is what I want: I also pay for all of the electricity I use and somehow the world doesn't end in the hell people insist microtransactions would cause--and (as I detail in another comment in this thread) I think technological solutions to DRM actually will have a hard time winning vs. AR passthrough ad block technology (though I then think the DRM war continues using legal means like expansions of Section 1201 to make it further illegal to traffic in circumvention tech... which is also bad).

well once KOSA[1][2] passes and every website requires KYC stuff like uploading a copy of your ID just so you can get access to potentially dangerous/offensive ideas, we might need an AI agent to automate that deep level of privacy invasion for us :-(

[1]: https://www.eff.org/deeplinks/2024/02/dont-fall-latest-chang...

[2]: https://news.ycombinator.com/item?id=39446355

> for the sake of making consumerism more efficient

If only. This wave of AI integrations is primarily focused on pleasing (potential) investors.

You need customers who bring in money to please investors over the long-haul.

If only.

This wave of AI integrations is happening in a context/framework for normalizing naked avarice so that preposterous power can be concentrated in the hands of the rich and dubiously ethical.

there are several AI features I would absolutely love to see integrated into the browser, primarily, content filtering with my own custom criteria.

for example, I don't want to see yuppie shit on news.ycombinator.com. right now, I have a userscript that filters out links from the worst offenders (e.g. newyorker), but every day, there are many more that shit up the front page and waste my time and attention. this could be solved with a very low parameter LLM by asking it to evaluate whether the text is related to science and technology or not.

hell, even images could be filtered with vision models, including local ones. I'd fucking love to hide a few broad categories of images from my sight, e.g. clickbaity youtube thumbnails (DeArrow helps to a degree)

This is just an arms race. It expends more energy, and the "solution" from the other side will be to pump more garbage in.

you could save even more energy by not using computing devices at all.

> for example, I don't want to see yuppie shit on news.ycombinator.com. right now, I have a userscript that filters out links from the worst offenders (e.g. newyorker), but every day, there are many more that shit up the front page and waste my time and attention. this could be solved with a very low parameter LLM by asking it to evaluate whether the text is related to science and technology or not."

FWIW, I once used naive Bayes classification solely on HN headlines and it worked reasonably well for solving the problem you're describing.

what is ridiculous about it? I read the gist of the articles I am mildly curious about by summarizing them on the spot. If it sounds more interesting than I thought I read it in full. Saves me time. Sometimes I ask one or two clarifying questions as well. Arc browser has this built in and I find it very useful.

I summarize the transcripts of 20+ min. long youtube videos to get the gist, which is what I am interested in anyways. Saves me time. (this is something I built custom for my use, maybe there are other options for this)

It is not an obsession. I always wanted something, someone that could do this for me, and now we have it.

If the AI is summarizing web pages, why couldn't this be done with a browser extension?

If the AI is summarizing files the user downloads, why couldn't this be done with external tools?

In general though, why does the browser itself need to be "smart"? Its primary, and perhaps only, goal should be to render web pages... I feel like this push to integrate AI features into browsers is done so that companies can promote their own (completely unrelated) AI services, and pull people even more into their ecosystem. It's obnoxious.

It can be done with a browser extension, and external tools. I only care about the friction, speed and ease of use. I am indifferent to where the functionality comes from, really. I'd like to have a very scriptable barebones browser alas such a thing does not exist as far as I know, so if the browser has them handy and it provides good experience, it is a net positive for me in any case.

Sure, I can see how it would be useful, but wouldn't you prefer to have the choice of which AI to use? What if Brave's AI is worse than Chrome's, which is worse than ChatGPT? Or what if it's better, and you're now tied to that browser, even though you'd otherwise use something else?

Even if I had this functionality built-in, I would still prefer to use something else. Even if that's slightly more cumbersome, though there will probably be extensions for this sort of thing, if they don't exist already.

But, to each their own. :)

AI assistant in a browser makes sense if it's acting as an extension of your memory - for example you're browsing some items you're considering buying, and the assistant creates a summary spreadsheet with their comparison, that you can later look at, etc. Could be useful for things like real estate.

Most AI tools I've seen however can't seem to get away from the classic "prompt -> response" chat UI.

You just assert that it is ridiculous without any motivations

Some ai features make a lot of sense. Automatically organizing tabs is amazing for helping me not get lost in a sea of tabs. Longer form auto complete saves me time, at least on mobile. I used to rely on Reddit comments to summarize trashy low quality articles to give me the primary points. Now that I've been reddit sober for a while, it's been helpful to have my browser do that summarization for me, helping me stay away from Reddit.

I can't say I will use every ai feature, but so far they've been helpful. I don't feel more ignorant. What I think makes people ignorant is outsourcing labor tasks like a cleaner for the home, a gardener for the yard, or handyman for basic repairs. I don't think those are necessarily wrong to outsource though.

What’s wrong with AI in a browser? Each SaaS app you use in a browser has or will have AI. Is the AI built by each of those ridiculous? If not, then what is ridiculous about a browser striving to create a general AI than can apply to any site you visit within that browser?

It truly is just a modern day Clippy.

Hardly. Clippy was hated because he interrupted you with unhelpful suggestions. I played around with the Leo thing a little bit and it never interrupts you; actually I had to dig through the UI a bit to figure out how to activate it at all.

Anyway, now that this sort of technology exists, Microsoft is missing a great opportunity to bring back clippy and cash in on some nostalgia. I miss that lil guy.

Reply on news.ycombinator.com