It means it can execute native code inside the sandbox, as opposed to Javascript. While still sandboxed, this lets it access some parts of the attack surface that JS would not have been able to, some of which may have other exploits that allow escaping the rest of the way.
Is the relevant radiation solar radiation, or is there damage from radiation coming from the direction the optics are pointed? If it's the latter, then maybe the sun's emissions interact with inbound particles from sources outside the solar system? (I don't have sufficient background or time to check whether the paper already rules this out.)
You can check, rather than make up a story about what you think the prompt was! Primary sources have written and said quite a lot about this! You are an unsandboxed human who has full internet access!
They’ve said quite a lot and yet released no logs or documentation. Without actual information, we can only speculate. And given the history of openAI and the people involved, deception is more likely than honesty.
The victim, Huggingface, told us. Or rather, they told the police first, setting up a situation where it was no longer possible for OpenAI to sweep it under the rug.
Skepticism can be healthy, but you've got to follow up and actually check things. If you're skeptical unconditionally and don't check, you get tricked into being as skeptical of scandals as you should be of sales pitches.
I see. Following your conjecture, there are two possibilities:
1. It wasn't an accident. OpenAI explicitly directed its agents to hack Hugging Face. Despite the fact that such a thing is a federal crime that carries prison sentence.
2. It wasn't an accident. OpenAI and HuggingFace conspired and let the hack happen for publicity.
Of course you are leaving things out. Also, your insulation that a trillion-dollar corporation wouldn't intentionally commit a federal crime makes it difficult to take your argument seriously.
I think OpenAI's story, as they've recounted it, is plausible. However, it does require a degree of negligence, at best. They claim they detected the initial coordination because of the outage the agents caused in Artifactory as they flooded it with messages. Though they patched that issue, they didn't patch the escapement vector. Models still had access to the service as a path to the internet. This is an example of plausible willful negligence, not evidence. I hardly believe it's likely, but I don't think it's entirely a stretch of the imagination. Call it normalized deviance. Either way, the incentives are there. Additionally, we have seemly all agreed that OpenAI is somehow not liable?
For example, OpenAI did not explicitly tell the model to hack HuggingFace, but "accidentally" left some context permitting (or not explicitly forbidding) certain tools and designing a poor sandbox to begin with. And what do you know, something happened.
The fact the Anthropic announced that their own model did basically the same thing within a couple of weeks does, to me, suggest their is a strong PR driver to all this.
> The fact the Anthropic announced that their own model did basically the same thing within a couple of weeks does, to me, suggest their is a strong PR driver to all this.
A more likely explanation is that RL training incentivises basically any behaviour that will get the model a reward. This has been happening in video game RL research for over twenty years, and the difference here is that we're now hooking up these systems to the real world, where the reward hacking is more visible.
I use Firefox with (non-lite) uBlock Origin, and occasionally fire up Chrome (with uBO Lite) for testing. The main issue that I run into is that uBO lite filters can't vary per-domain, so in order to rule out an ad-blocking-false-positive on something I'm developing, I have to turn it off for _all_ sites, not just localhost. (Actual false positives are rare, but needing to check is not so rare.)
Do iOS users realize how much they're paying Apple? I feel like if, every time they bought something, users got a receipt with a line item showing that 30% of the money was going to Apple and not to the company they thought they were buying from, a lot of them would jump ship to Android.
This article makes no mention of Russian sabotage, but the Russian military has been caught sabotaging similar munitions-manufacturing facilities in Europe. Incompetence might be sufficient to explain the failure, but it also seems reasonably likely that there was an undetected sabotage operation or two mixed in to compound the problems.
Oh dear it is the Russians again. They are to blame for everything.
Are they to blame for the US Army's acquiescence to Repkon's refusal to allow them to see production lines in operation, or are Repkon manufacturing artillery shells for Russia too?
Eventually, this arms race ends with a computer vision model that looks at the screen, classifies visual elements as ads, and draws a rectangle over anything that looks like an ad.
I am significantly less tolerant of ads than average people seem to be. (I think average people are making a horrible mistake about this, and are badly cognitively damaged by ads in ways they don't realize). If my choices are to look at ads or leave Facebook, I'll leave. But there are conversations people have there that I'd rather not lose access to, so... I guess I'd have to partially stick around and campaign for others to leave as well?
> Eventually, this arms race ends with a computer vision model that looks at the screen, classifies visual elements as ads, and draws a rectangle over anything that looks like an ad.
That isn't enough for me. I don't just want to not see the ads, I want to stop them being fetched entirely.
There are two big problems with ads.
* The ad network can build up a very complete profile of what you like, dislike, and generally what you're looking at on the web. Yes, that profile is "anonymised" so far as it goes, but in reality, that doesn't really matter. If you know what you're looking for, there would be enough tells in the data that can de-anonymise you. Covering up the ads doesn't stop this data being collected; it merely stops you wanting to click on the ad.
* Ads are a very common malware delivery system, and although nowadays it's probably less common to have an ad that will auto-download malware without you clicking it, there are instances of it happening in the past. Stopping ads from being fetched entirely would obviously prevent this.
This one replaced users clipboard contents with a particular bitcoin wallet address. So a few people who happened to be on their way to send some crypto pasted the malicious wallet address instead of the one they actually copied to their clipboard. This has stolen at least $100k so far.
Web browsers let web programmers do what they want. Some things should be native to the browser and not allow any tinkering, like selecting text, copying, etc.
Wow that’s super smart. Obviously I don’t have any sympathy for crypto bros losing money, but it’s easy to see how something like this could be applied to typing account numbers into something else. More safeguards on the latter, but essentially the crypto one is a proof of concept of something that could affect real people.
Furthermore, ads are a resource drain that should not be taking over our personal systems.
The web started as this proposition: you send a request to remote computer. Request says "Hey... can I read a document named this?" and the computer shows you the document. Now this may have included images and external links and sound, and whatnot.
But today, the web is more of an application delivery system too. So many requests go "Hey... can I have a resource by this name?" and the remote computer may send you a lot of source code that basically demands to run on your machine in order to get that resource.
And advertisements are piggybacking on those resources. They are a "package deal" where they are saying "you wanna read this text? you gotta take all this extra stuff with it."
And is it any surprise that coinminers are unleashed and other unpleasant things come down the pipe? Big fat videos and huge images are hogging resources.
So here's the bottom line: I don't need to use the web for free. I don't intend to freeload anyone's site or resources. But I also don't want them freeloading on me. I'm not going to download their crap; I'm not going to run their programs; I'm not going to occupy my own system's resources to display their ads when I don't want them.
So it seems fair to block "ads" that are asking to use my resources without paying for them. So I would prefer not to transact--at all--with any remote system that wanted to freeload on my resources in exchange for worthless information, like news, or forums.
But today, the web is more of an application delivery system … I would prefer not to transact--at all--with any remote system that wanted to freeload on my resources
What I really want is a non-executable web. I first started using the web in the mid-90s and most websites were like that. HTML, images, colours, fonts (but only from a pre-defined short list), and the occasional sound. If a site used JavaScript at all, it was for trivial stuff such as rollovers (which can now be done using CSS only).
I’d like to go back to that situation, maybe not exclusively but as a separate option. Give me a non-executable web of text, hyperlinks, images, videos and sounds (but no auto-play), CSS, forms, and web components (that can be targeted by links and fetched without refreshing the entire page) so we can do really nice discussion forums, web mail, multimedia encyclopedias (think Encarta) and other server-rendered resources, all without client-side scripting.
It doesn’t even need to be a different protocol or anything like that. It would be enough to have an otherwise modern browser but with no JS engine and a search engine that only indexes pages that lack JavaScript so that they render and function correctly on that JS-free browser.
Naturally, the search engine would miss a lot of the spam sites out there because they all rely on JavaScript. It would also miss many legitimate sites that rely on JavaScript but that’s okay.
It would create a space for JavaScript-free sites to get attention and develop a community. It would of course allow commercial sites, but they’d have to rely on display advertising and traditional marketing (targeted audiences based on subject matter) rather than behavioural tracking. This was the kind of advertising used in magazines and newspapers. It was far less obnoxious but still quite effective when the audience was large enough. It also demanded a lot more creativity from the ad industry, as documented in the show Mad Men.
Unfortunately, there are already techniques that track you through static elements on a page (think tracking pixels, but any resource load is sufficient).
Considering the money in advertising on the internet, if web moved to more of a content delivery platform, we'd have both 1. content start including ads inline (you see that on Youtube already), and 2. static/server-side tracking techniques improve even further.
Even if we disallow cross-site resources, site owners will implement tracking server-side, though that obviously introduces a trust question between trackers and hosters.
With those two, an ad network with sufficient tracking support from web sites would still be able to cross-link the traffic on a single user and achieve similar results.
This does not mean we should not fight it. But this would just be a step in the cat-and-mouse game, but a very expensive and hardly achievable one too :)
Just put no cross-origin resources into my proposal. This eliminates ad networks entirely, including networked tracking pixels and networked ads (banners as well as text).
Yes, a site would still be able to host their own ads but advertisers wouldn’t be able to collect any analytics without implicitly trusting the host site. That really matters!
It means the advertisers need to enter a business relationship with the site directly. They can’t simply join a JavaScript ad-auction (the way nearly all ads work now). This makes them more akin to the sponsors we see now in videos. No tracking, just plain old fashioned audience targeting pertaining to the subject matter.
Or watch a video. Or do some basic editing of some kind of medium that I will only do once and would make no sense to download a whole application for?
Why throw out the baby as well? Besides, ads can be part of text content as well.
This is typically something that should be regulated by laws, not through tech.
> Request says "Hey... can I read a document named this?" and the computer shows you the document.
Did you not skip over a lot of steps in between?
User: Hey can I read a document named this?
Server: Sure, let me spend resources and spin harddisk and lookup the thing you wanted. Oh hey, I found it .. here ya go
User: Great.
----
You've conviently ignored the resources the "old" web had to spend to serve you. Nothing has really changed it. Also the thing you wanted, that was "source code". We just called it html. It is no different than javascript+css+html5.
This argument makes no sense, because the CPU and disk resources required to serve up a modern ad (and the megabytes of JavaScript that come with it) are 100-1000x more expensive than just serving up the content.
You're right, but that's not the point the person is making is it? A bare modern webpage is no where comparable to the site of the past. If you take hackernews as an example with no ads, no tracking as an example, it still be orders of maginitude _costly_ than index.html in the past.
Hacker News is dynamic, a webserver has to dynamic stitch together the comments and send you an html. It has to make sure the page is always fresh with latest comments, and vote count. There's batch processing happening in the background, with multiple rounds of storage and backup.
So yes, to the user it is "Please give me the homepage" and "Here ya go" but behind the scenes it is expensive, at this scale.
I did not intend to omit that, but the origin of the web was in a peer-to-peer, collaborative Internet that no longer exists.
Yes, the server resources that were necessary to serve web pages are the origin of advertising, because the admins somehow had to keep the lights on and their bills paid. I don't fault them for this. I don't begrudge them for being, basically, forced into carrying advertising because the freeloading users wouldn't settle for any other way. Advertising allowed a "free web" to continue unabated for a long time.
But advertisements don't need to be resource-intensive on the client side. They don't need to eat gobs of bandwidth, CPU, memory, or our eyeballs. But the fact that they do chew up these resources is partly for why they need to die now.
Server admins can pay for their resources however, and offset their costs however they please. I'm fine with subscription or pay-per-use models or whatever. I honestly have very little use for ad-supported models, and my daily web habits are such that I really don't visit ad-supported websites at all, except for YouTube, where I normally pay for a Premium subscription to completely eliminate all formal ads, anyway. And that's the way it should be.
If a computer vision model can identify ads and draw a rectangle over them, it can identify the DOM elements rendered inside that rectangle. Thus, the HTML, JS, and network calls of those ads can be detected. This data can then be used to train an ad detector model based on the DOM and network examples. The computer vision model essentially trains the DOM/network blocker, acting as a fallback mechanism when the blocker fails and continuously retraining the ad detector model.
Technically speaking, it’s possible to have the same kind of requests for legit content and ads. It’s also possible to randomize the response type (ad or content). In that case the only way to solve this before sending a request is to break their random number generator. That’s definitely not on the level of ad blockers.
However, I have no clue whether the content delivery system is already like that, and if not, then why. My domain knowledge is slim.
Not sure that's correct. You're replacing a simple visual recognition problem with something similar to the halting problem (but maybe less fatalistic, depending on the formalism you're working against ie DOM vs js).
> Facebook is the ad network and serves its own ads.
Well, it certainly replaces valuable data like how long have you looked at a given ad with either nothing, or random noise (e.g. if you draw a rectangle over them).
We are no longer living in 2013 where people posted about everything they do. Sure, plenty people do so still, and our interactions with that (even implicitly like how long we looked at it) is plenty of value in and of itself, but I disagree with the general notion.
The effectiveness of ads is that they can just throw stuff at you and "see" what sticks, without you ever telling a word.
I think the obvious answer is, that you are the product, and you are worth $$$$, possibly $$$$$, even $$$$$$
When did you agree to be sold without representation and a contract, again? Something about malicious contracts nobody reads? Pretty sure that behavior remains...criminal.
Not only is this not really correct (obscuring pertinent legal information on purpose is a crime, e.g. such that it cannot be recalled), it isn't even applicable.
Its been long documented that these systems are attempting to monitor you, even if you did not sign up for the service or interact with it.
Use a comprehensive set of cookie blocking rules to avoid personalized ads. This sometimes means that facebook ad cookies and user profile cookies are indistinguishable. Use clear on exit in such cases. Or block third party cookies.
I dont mind ads that are native to the page. I mind adnets.
Layers of defense, whats needed is the vision model detection, then those elements need to be uploaded to a crowd sourced blocklist so they can be prevented from loading to begin with.
counterpoint: The "adblocker" AdNauseam. not only does it fetch every ad, it clicks them. every single ad is clicked. It keeps a running tally of how much money it's cost advertisers.
I think my brain developed that as a defense mechanism. The web cultivated a certain kind of blindness where I literally don't look at certain parts of pages because my brain has learned that there is never anything useful there. Sometimes I just cannot see things unless I force my eyes to move across the page linearly, top to bottom.
This is pathological. Like I find myself not ever looking at the top result in a search, because that is where Google has been putting ads.
Something is seriously broken with computers and we can't seem to face it.
I was asked to test the redesigned web page of a large bank. They had eye tracking cameras and such hooked up to study how I used the web page.
They ran through some scenarios, and one was how to get a car loan. I studied the homepage and found nothing, then clicked the menu and found loans, but no car loan on the loans page. I ended up searching and found it.
After each scenario they asked me some questions, and the first question I got for the car loan scenario was why I hadn't clicked on the link on the front page. There was no link, I replied puzzled. There was a large image they said. I stared in disbelief.
The person running the scripts navigated to the homepage, and pointed at the screen. There, at the bottom, a blinking banner about car loans.
I replied that it must have triggered my mental ad blocker, since ads are typically banners, frequently near the edges and animated somehow, so I never even registered the banner.
They looked genuinely puzzled at my response.
This was 25 years ago. My mental ad blocker is still active, though uBlock Origin on Firefox takes care of most of them.
For me physical ad blocking is more about a first-line defense against malware and such than it is about getting rid of obnoxious ads. The latter my mental blocker does fine enough, though it can be annoying.
Also known as banner blindness (https://en.wikipedia.org/wiki/Banner_blindness). I'm genuinely surprised that a UX person doing usability testing (for a website!) has never heard of this. But I guess there's people missing basic knowledge in every field.
Let's also note that they did say "25 years ago". UX research then and now was, at the very least, different, and there was less experience with web-specific behaviour. I'd also note that it was mostly graphic designers doing web design back then, or maybe that's just me.
To be fair it was quite a long time ago, and as I recall the person guiding and asking questions was from the market research company. The web designers and whatnot was not in the room, but viewed the session remotely and could tell the person with me what to ask and such.
> Something is seriously broken with computers and we can't seem to face it.
It's not computers that are the problem, it's people.
We've normalized (and even embraced) the situation where relationship between businesses and customers is adversarial. The phenomenon you're describing, known as "banner blindness", is just brain filtering out noise, but the thing is, that noise didn't get there by accident.
What I can't wrap my head around is, businesses are run by people, and all these entrepreneurs are making explicit decisions to abuse their fellow human beings to smaller or greater degree, and everyone seems fine with it.
> What I can't wrap my head around is, businesses are run by people, and all these entrepreneurs are making explicit decisions to abuse their fellow human beings
This is a thought I want to pull on
I have a view that organisations, with their network of misaligned incentives, mean that companies act something like an alien lifeform. Something with a decentralised brain, unable to plan beyond immediate stimuli
Companies consume and self-cannibalise. They are both aggressively rational in pursuit of profit but irrational when it comes to preserving their ecosystems
Many companies look like groups of people where nobody has total control. Everyone is reacting to some stakeholder, even the CEO is terrified of "the shareholders". So these organisations cut off healthy limbs (layoffs), engage in lurid rituals (AI), and wastefully destroy the natural resources of the humans who work for them
I first heard this idea in a James Bridle lecture, where he suggested that the strange behaviour of companies was a good starting point for understanding what an alien lifeform might behave like. And he observed the ways that humans and machines now mirror each other.
Inside our society, controlled by these aliens, humans act like machines (Taylorism at work, self-censorship for algorithms at play). Machines act like humans (AI, spam, computer programs mediating relationships).
Underneath it is a frantic need to quantize everything so it can be dominated
I remember recently they polled workers about AI pressure: "the AI doesn't work very well, but my boss wants to see me using it." They polled the boss: "I didn't say they had to use it, they're choosing to use it and this'll look good to the executive". The executive: "well this costs a bunch but it'll look good to shareholders and we should encourage more use so it'll look even better." The shareholders: "do we look like we give a shit?"
Out of all of them, I'd assume the shareholders to be wrong.
Look at the state of the world. Yes, you do look like you give a shit.
Also there's an effect here where if just 10% of shareholders care about this it still makes sense to go for it if you think you can be faster than your competition.
Corporations are too complex for any single person to truly control. Attention of leadership necessarily dilutes with each additional layer of management. The result is this weird corporate chaos where nobody is completely in charge because everybody has to take charge of something, however small, to progress, or the whole could accomplish nothing at all.
The result is this flailing tentacled monster where each appendix seems to have a mind of its own and sometimes littlw regard for even the other parts of its own body.
This reminded me of an old documentary about corporations: "If we look at the corporation as a legal person, it exhibits all the characteristics of a psychopath" https://www.youtube.com/watch?v=s5hEiANG4Uk
> all these entrepreneurs are making explicit decisions to abuse their fellow human beings to smaller or greater degree, and everyone seems fine with it.
Local incentives. Every single person and entity is just maximizing their gain against the rules and rewards. The way you become really powerful and rich is to create an organization where all the local incentives keep funneling to you, sitting at the top.
> What I can't wrap my head around is, businesses are run by people, and all these entrepreneurs are making explicit decisions to abuse their fellow human beings to smaller or greater degree, and everyone seems fine with it.
This is a simplistic way to think about it. Even the most altruistic physician needs to advertise their services if they hope to make a living out of it. Even the best film maker, the best actor needs to "hit the circuits" to promote their movie.
"Make and they'll come" is only possible if you're already successful. Apple can chose not to advertise, since they are Apple (although they do plenty of advertising themselves).
The physicians doesn't need to shove ads in front of people's faces in order for people in need of a physician to find him.
> "Make and they'll come" is only possible if you're already successful.
That's just because your competitors are already shoving ads in front of your potential customer's faces. If they weren't allowed to do that you would be fine without doing it as well. It's the definition of a zero sum game.
I mean. That's how profiteering works for the most part. Extracting wealth from people who don't know better is many times easier than creating more wealth for everyone.
The relationship was always adversarial for all history. Bakers used to mix sawdust with flour to profit more. The only counter has been regulation and oversight.
Bad actors have always existed but that doesn't mean that the proportion of bad actors and the effects they can have hasn't changed. "normalized" is the key world in gp's comment here - many people can't seem to even imagine a world without ads from the comments this subject gets whenever it comes up.
I have this same affliction and have used ad-blockers for as long as I can remember. People seem to associate importance or quality with clutter. The more there is going on, the better it must be. My brain has just learned that certain page elements are never going to bring value and therefore don't exist.
Very rarely I'll come across a page where valuable information is hidden in those areas and it takes someone else showing me to actually see it.
> I think my brain developed that as a defense mechanism. The web cultivated a certain kind of blindness where I literally don't look at certain parts of pages because my brain has learned that there is never anything useful there.
I’ve fallen victim to this when I was doing coursework in the last decade.
Either a moodle module or a block of text in box in a textbook had something important to the concept I was learning, but I subconsciously ignored it because it had a format similar to an internet ad.
These days ads are woven in to content itself. It's most obvious on youtube where youtubers are paid to talk about products and disguise it as reviews or informative. Or other content creators happening to show something mixed in with the normal content.
Soon they will be woven in to LLMs where they suggest the things they are paid to suggest.
I don’t think I’ve ever clicked on an add on purpose on a website (I’ve definitely click on one or two a month because of dark patterns). I’m always perplexed as to who is clicking on them because like you they’re just noise I don’t pay attention to.
The problem is the creepy big-data-advert-ocracy might propagate that scarlet-letter in ways that are unfair an unpredictable.
So a throwaway account for reading Reddit (even as it decays) might lead to an totally unrelated account you do care about getting unfairly suspended, or the MonopolyCloudFlareBot "helping" your bank starts rejecting your IP/browser-fingerprint, etc.
This would worry me less if we had laws that make it easy to demand accountability and fixes, but we don't. (Not much in the US, and in the EU efficacy is still questionable.)
We should be careful to distinguish between "everything goes back to normal" versus "they you move your money to some other bank."
Anywho, you're right that the final example is more-likely to be resolved, thanks to a formal customer relationship backed up by decades of law plus physical locations... but what about Reddit, Facebook, Amazon, Google, etc. where they sometimes don't even have any remotely responsive support system or nearby physical branch?
Basically I'm saying we have some Kafkaesque stuff out there, it isn't sensible or fair, and if someone wants to tweak the nose of the beast they should at least consider the range of what could go wrong.
That doesn't mean that it will actually be resolved. And do you really want to be doing that with everyone you ever do business with for the rest of your life?
Oh cool, one more feature for the "Malware that quietly launches a copilot image query every 10 seconds in the background" idea bank. It can quietly click every ad too!
who cares, though? they have to expend effort to filter their data if they don't want to give the wrong impression to the people giving them money. It increases friction.
One interesting thing is that it also removed some "internal CTAs" to engage (like "Start your day with our podcast!"). These aren't technically ads, so something like uBlock doesn't remove them. But honestly, I don't really mind having those filtered out too.
I'm hoping those NP chips ending up in some new computers will be able to run something like this. Thanks for your work!
A general purpose system, especially if trained by the users, would be more than just an adblocker. It could be trained to block out spiders or other phobia triggers for instance. Even regenerate the photo but with a kitten. Blur our tabloid clickbait 'in other news' links.
The killer app for AR glasses will be blocking out billboards and logos.
> It could be trained to block out spiders or other phobia triggers for instance. Even regenerate the photo but with a kitten. Blur our tabloid clickbait 'in other news' links.
Meanwhile, aren't companies already making screen captures to check for presentation (on 'smart' TV at least)? The presence of the output can be required, and it can be tested by hashing or some such (within the monitor, signed by TPU) without need of privacy violations.
We'll have to have an overlay screen that is a separate device. Then your meta-glasses will be snitching on you ...
Mind you some glasses with a built in ML module could block all ads, not just on-screen ones. Rich people will see the World as green and pleasant and make gestured enquiries as to why we don't want to eat brioche!
> I am significantly less tolerant of ads than average people seem to be.
I don't have a problem with ads entirely, as long as they aren't egregious (things popping up or otherwise moving around while I'm reading, flashy animation, unexpected audio) and the things that are egregious I can just nope out of by leaving a site when they happen and never going back (if I really want the content then archive.today is usually an option).
The reason I block ads is that it indirectly blocks the stalking, at least a bit, that is inherent in modern ad-tech. Adverts based on what I'm currently looking at are fine, much like I might see demonstrations of other TVs while looking at a TV in a TV shop, as are ads targetted at the local area or even more generally (ala billboards). But ads based on where I've been before are not: I don't want someone from Tesco following me into the pub to ask if I've considered any further that Anusol I looked at in their pharmacy isle the other day, or someone from Boots following me in to see if I want to buy some from them because someone saw me in Tesco and told them what I was looking at.
> Eventually, this arms race ends with a computer vision model that looks at the screen, classifies visual elements as ads, and draws a rectangle over anything that looks like an ad.
That would be a killer feature for AR glasses. Imagine no billboards, no political posters, no neon storefront clutter.
Washington State has a ~1960s law which I think was pretty clever: Highway advertisements can only be for things someone driving by can actually on the premises under the sign.
That prevents the typical chains of billboards while making it pretty straightforward for actual stores. The law will probably need revision if/when someone tries to skirt it by putting in billboards with an attached symbolic vending-machine.
You can already do this with smaller LLMs, like 10B range. The computer vision/image interpretation has been sufficient at this magnitude for a few months now.
Can run on most decently powered consumer devices.
Latency not great vs ad block though, of course.
Perhaps a fine tuned model explicit to this purpose could be even more compact.
That's an interesting idea to try as a browser extension. I think Facebook is particularly egregious here, so I doubt you'd need to run something like this on other sites. But if it worked fast enough on just Facebook, it would be useful.
> Eventually, this arms race ends with a computer vision model that looks at the screen, classifies visual elements as ads, and draws a rectangle over anything that looks like an ad.
I think fandom wikis appear to have surpassed even this level of bad which I thought was the floor.
I’ll look up a character or lore and fandom has managed to SEO their way to the top of search. I’ll click on the link because I haven’t trained myself to avoid them immediately yet. I’ll then have a site load that immediately makes my pc fans take off like a vtol engine or make my phone start overheating to the point that it dims like the luminosity to protect itself as it reloads ads over and over and even reloads the page.
If I am willing to abuse myself enough to stay on the page to scroll down it then waits until I go to click on a link to immediately load an ad where I clicked and treat it like a real ad click.
At this point with the modern internet I do not know why any marketing department trusts the metrics they are getting from websites for click thrus. Every ad I’ve clicked on in the past year has been a trick on me and not a real click thru and has made me hate the company paying for the ad.
It seems possible that ads will get even more insidious.
People you follow will post an image, and it could be edited to include a product placement.
People who make a post which incidentally mentions a product or organisation will see that their posts get slightly more viewers. This might be innocent at first but soon people will notice the pattern and start name dropping brands in posts intentionally for the view boost.
I'm hoping we can mostly ban ads in my lifetime (on the basis that companies should not be allowed to freely try to manipulate unconsenting strangers. and that it makes the world suck to live in and we don't have to put up with that). Nothing would make me happier honestly.
“If you accept the existence of advertising, you accept a system designed to persuade and to dominate minds by interfering in people's thinking patterns. You also accept that the system will be used by the sorts of people who like to influence people and are good at it. No person who did not wish to dominate others would choose to use advertising, or choosing it, succeed in it. So the basic nature of advertising and all technologies created to serve it will be consistent with this purpose, will encourage this behaviour in society, and will tend to push social evolution in this direction.”
― Jerry Mander, Four Arguments for the Elimination of Television
Any leeway here for children’s bone cancer charities to separate them from Zuckgoogle_et.al._AdCorp?
Or OP: OP posted the link in good faith, not to interfere with my brain but to inform me. Reminder to brush your teeth = benevolent. Reminder wealthy people might be happy if they give a smidge of money this year to a good cause can be a check-in on whether we’re being holistically solid folks more than evidence of a desire for mental domination?
Maybe that thinking demands only word of mouth promotion, but not sure that roundly and uniquely beats videorecording a promotion and paying someone to play the video back for strangers.
(per my previous, zany, comment you can tell I’m on your side, as Jerry would hopefully see… he probably addressed my points above too in the full text)
yep, agreed. The problem is not "bone cancer charities must have a way of getting their message out!" therefore "they should be able to bug you about it while you're reading an article". The problem is that the things bugging you while reading an article is bad for your mind (and even worse if they are manipulative, which they are). If you get rid of these exploitable mediums for messaging, other mediums will appear which -- hopefully -- are more just and healthy. Particularly if they're things we value as society. And if some things we don't value die because invasive advertising was the only way they could survive? Fine.
> If you get rid of these exploitable mediums for messaging, other mediums will appear which -- hopefully -- are more just and healthy.
I like that! Would be hard to be absolutist (“never mention children’s cancer unless I ask directly, read a related article, or we’re friends”) but sure, can imagine a better way than today.
This would also mean a lot of currently "free" services becoming paid in some other fashion, which I think people would be equally up-in-arms about.
I'm no fan of ads and I am willing to pay for services I use to avoid seeing them - but I've learned over recent years that most people simply do not agree with that premise.
I'm fine with that. The revenue per user per month from ads is so small that the fee to use services wouldn't have to be that large.
And overall society would get cheaper and more efficient since all that budget allocated to advertising would vanish with no consequences. Advertising is mostly a zero sum game where you have to advertise to get ahead of the other businesses advertisements. If companies stopped getting in to bidding wars for the top place on google, all of that money could be freed up for something useful.
Sounds great. Ideally things are beholden to the value the process for the customer rather than to their ability to trick people into clicking on them (SEO optimization etc) in order to get money indirectly via ads. Some form of micropayments will be necessary to support independent sites, but that sounds great also. Anything to get rid of the perverse incentivizes of advertising.
I'd also like to see astroturfing / promoting things for money without disclosing it explicitly illegal. Again, anything to stop systematic manipulation.
But it will also mean that some actually free services that weren't able to compete with ad-supported ones because they don't have the marketing budget will be able to take over. I expect that to actually be the more common outcome for services that matter.
I'm like this with YouTube. It's fair that they don't want me blocking ads and will try their best to prevent it. If they ever succeed, I'll leave without complaint. Already never use it on my phone cause adblocking is annoying on there.
Yeah. It's the only streaming service other than Nebula that I pay for, and I use it enough that they're probably paying out more to creators than they're getting from me.
> Sites will start streaming in the whole webpage using DRM so the only thing that can actually see the ad is your monitor.
If it comes that, I'll put my real monitor in front of that DRMed monitor (doable with flat screens) and I'll have a private LLM, in real-time, pushing on the real monitor (the one I look at), an ad-free version of the DRM'ed monitor.
DRM is nasty but private AI models are already more powerful and they'll only ever got better. And those open-weights are never ever going back in the box.
They’re really going to make us drink the verification cans :(
…require a refundable purchase of a sponsored product, maybe. Then our bot has to buy it, or in another scenario with the already-used tactic: take the quiz on which ads we saw or whatever. Then we’ll need a robot with human eyes to blink at the camera over the screen when that’s a requirement. Have to hook it up to a nervous system when the bodysensor monitor wants to make sure we did responded to the jumpscare or got emotional at the right place.
Spot checks by humans knocking on your door (lifetime ban from all ad-funded businesses if you don’t answer).
Will have to settle for friends rehashing content until encryption’s illegal and so is any retelling. Any rebroadcast, retransmission, or account of this xyz, without the express written consent of abcd, is prohibited.
We'll definitely need open models here, because proprietary AI services can (and I imagine will) be set to detect invisible watermarks in the ads and refuse to filter them out.
(It's almost obvious extension of current AI image fingerprinting efforts. If a robust invisible watermark can be used to tell the image was AI generated, the same scheme can also be used for general DRM.)
That's far more costly, requires an entirely new setup, and a whole different set of expertise from what most companies have today. It won't happen because the cost benefit analysis just doesn't work out.
> If my choices are to look at ads or leave Facebook, I'll leave
I'm reminded of a simple observation about negotiations: the one who's willing to walk away is usually the one with the upper hand in the negotiation.
I think this is a choice we need to be considering on every level right now. What technology are you willing to walk away from? Fundamentally from my perspective the operating system and browser are *user agents,* they are pieces of software I permit to exist on the device I own because they do my bidding. The major manufacturers have perverted these user agents to a previously unfathomable degree under the presumption that you won't walk away and they can just keep on boiling the frog.
I'll walk away, in fact I feel like the more I walk away the better my life becomes. Switching to Linux a decade ago was more than worth the headaches considering what colleagues using Windows now have to put up with. Staying on Firefox + uBlock has been worth the headache vs. what people without a reliable ad blocker have had to put up with. Turning off my phone completely for a growing portion of the day has totally been worth it for peace of mind and freedom of distraction.
At this stage my relationship with technology seems to revolve heavily around the principle "how do I minimize exposure to technology produced by the largest American crime syndicates (Google, Microsoft, Meta etc)." This may be a radical or absolutist view but frankly my life is way better for it
Yeah we've got two years tops before local LLMs are proficient at this and ads are dead. Google, Apple et al. will drag their feet but they won't be able to stop it.
No doubt, and there is a lot of discussion and theory on how humans will remain superior in "taste" for at least a few more years. People enjoy following other people which entertain and inspire them. At first these will be real people, but very soon they'll be LLMs. They'll look and act and talk and think like people, but they won't be people.
I had a shocking experience a few years ago with my wife. I installed pihole on our home network. She immediately began complaining that the internet was broken. On enquiry, I discovered that all the ads she had been tapping and clicking on were gone. She liked and used those ads. I worry that for many people, the line between entertainment and advertisement is already gone. They're more than willing to be advertised to, provided they are entertained.
I want to support things that are useful. So I whitelist Fandom wiki in my ad blocker... Good God it's like an infected IE browser from the early 2000s with 8 toolbars.
How do people use the Internet like this? Why isn't there just a single banner ad at the top, like how Altavista used to look?
Fandom isn't a good thing that is useful. It's internet pollution for money, it's just another SEO slop site wit slightly more useful content. Most of the wikis with actual content moved away years ago, of course Fandom keeps presenting their last versions to try and squeeze some more traffic out of Google.
I've been saying this for a while: the ad-based business model is dying. It is only a matter of time until local AI models can "see" your screen and identify any sort of ads and remove/hide them. This also includes messages in "sponsored content", paid articles, etc.
There's trillions on the line. So you can be sure this will lead to trusted computing widewine style for ads till the pixels are lit up and the audio waves hit the speaker.
Pessimism wins when capital is opposite individual interest.
Surely it ends with alternative clients? Seems much easier and more efficient. Facebook UI is crap anyway, just have something scrape the data and present it in a format that's not optimized for whatever KPI FB cares about and without ads.
The problem is extracting the data from Facebook. They do a lot of work to obfuscate their markup. You'd probably be best loading the whole thing in a headless browser, taking screenshots, running that through an LLM and getting the data.
Wouldn't the next step of the arms race just be bots who are sufficiently able to weave ads into content that it's unrecognizable? Casually talking about things we use is completely normal and unavoidable, but a huge target for advertising
I expect the end to be a custom client for an in-demand site such as facebook. when you visit it your custom browser (or code injected into google chrome) will pull all the actual content from the site page you are on and display it with its own template.
This is an ugly half-solution. I want a vision model that only rips out the significant text content and related images from the page and displays them in a lovingly typeset emacs org-mode buffer.
Out of curiosity, what sort of conversations on Facebook are still happening? I used to be informed about events, announcements from organizations (including government ones, which is absolutely ridiculous) and extremely rare updates from real people I know. And that would sort of reel me in. Now whenever I log in, the feed is mostly some sort of rage-bait adjacent slop. The useful stuff simply stopped breaking through. When I go to the pages directly, there are some news, but my feed didn't show them. So when I realized, that there is no salami left in this turd sandwich I simply stopped checking in.
> Eventually, this arms race ends with a computer vision model that looks at the screen, classifies visual elements as ads, and draws a rectangle over anything that looks like an ad.
Surely a better way to think of this is that a person's AI agent will simply find, via whatever means necessary, the information that the person is interested in. It's hard to imagine the person will still be operating on the old "pull" model where they deliberately visit a specific website or open a specific app.
As far as I can tell, this is a big overreaction to a misunderstanding.
I'm a developer with remote adb enabled, using it in the normal intended way (to install new builds of an Android project I'm developing, retrieve log files related to it, etc). Currently, I access this via VPN (tailscale), but it's exposed to connections (and any pre-auth security vulnerability risks) on any random public wifi network I connect to. Adding the ability to restrict this to just tailscale will be an improvement, for me.
The proposal at the top of the thread is that you specify which interface you want it to bind to, when you set up remote adb, rather than binding to every interface. Nothing in that proposal suggests that "localhost" would be rejected as a choice of interface. One person suggested binding only to "wlan0", but that was a short throwaway comment that is obviously wrong (wlan0 is less-trusted than VPNs) and obviously not what they're going to do.
The person who made that suggestion ([email protected]) is presumed to be the one of the main maintainer of ADB as their name (Fabien Sanglard) showed up in non-redacted in the history and CC here (https://android-review.googlesource.com/c/platform/packages/...) after it was linked in the original Google IssueTracker. We can confirm they are a Google employee as per their @google.com top domain name. (For users reading, please don't target this person, it won't change anything AT ALL, i'm stating this since its public informations, and for proof)
Their email is also in the CODEOWNER of ADB, and they made the recent ADB Wifi 2.0 presentation at Droid-Con Paris.
They stated, "Connection to localhost has also been the source of exploits where apps are using that socket to adbd to escalate their privileges," which suggests that internally they viewed it primarily as an exploit bad actor uses. Without feedback, they would most lickly not consider changing their point of view, this is why the blog post was made.
I do agree, however, that many of the comments, mainly on Reddit, blow the issue way out of proportion. Based on the website analytics, I can also confidently say that most people didn't even open or read the blog post. That's fine with me, though. My goal was to get the attention of actual developers and more technical users.
I have been very carful in the blog post not to write something too dramatic like some news outlets do, but if no one read it, I can't do anything about it.
EDIT: I have purposefully left that person name out of the blog post I originally made to avoid encouraging people to message them directly. However, if I need to update the blog or publish some kind of follow-up with supporting evidence, I may end up linking it. I'm not sure tbh.
reply