Need to rethink the system that allows for (and encourages) this kind of plausible deniability. From "Oh we need this permission for [non essential feature] and you need to accept it if you want the app at all" -> to giving the user ultimate control over what happens on their personal device. Virtualize what the app can see and use fake data/identifiers/devices if necessary to get it to do what its supposed to. If the App isn't going to act in good faith why should the user? Fine grained permissions don't really work in practice because the app can keep annoying the user until they give in and hit Allow.
Also perhaps AI agents are now capable enough to run these apps the way the user would and recognize these dark patterns. Flag those and feed it back to a warning at the point of sale that users can upvote there to signal their disapproval and a threshold score that risks removal of the app from the store. Because bad behaviour continues to make business sense if the rules allow it. Moreover, it penalizes and puts pressure on the good actors as a "missed business opportunity".
Maybe just have Claude rebuild the app in a non-scummy way. Analyze the site for the relevant data APIs, while avoiding the tracking stuff. Maybe someone maintains an MCP for these popular websites so people can use agents on them.
What you want is basically how it works. On both phone platforms and PWAs, all permissions are visible to the user explicitly. All of them can be revoked at any time. Apps are disallowed from requesting an already-denied permission.
Obviously apps can tell if they haven't been granted a permission (even if you tried to fake this, they aren't dummies and will know if it's not working), and obviously third party software isn't under any obligation to work without them.
But the platforms have done what the platforms can do, at the architecture side, really. The next stage is human-audited enforcement of malware, which this AliExpress nonsense might hopefully run afoul of.
> Obviously apps can tell if they haven't been granted a permission (even if you tried to fake this, they aren't dummies and will know if it's not working),
How can they tell? For the permissions I can think of: location, filesystem, etc. it should be easy to lie/spoof.
That does nothing but start an arms race. Fine, audio "works" but do you get noise? Can you read back the sounds you play? No, right? It doesn't work, QED. Now the platform needs to fake the noise.
Likewise for any other hardware access you want, and most of them are harder. How do you fake local storage without storing anything? How do you fake Bluetooth access without virtualizing an entire device? Do you fake the screen dimensions to look like something else? Input latency? Where does it stop?
No, "does this work" is among the easiest questions to answer in technology. We aren't going to win this war.
No, but you also can’t do that if the user approves your app’s permission to play audio, but has the volume permanently turned down to zero.
> local storage
Give the app a temporary, space-constrained store that isn’t shared with other apps.
> bluetooth
Don’t let apps access raw bluetooth signals, but instead ask the platform to display a pairing interface (possibly with a custom stylesheet), ask the platform to call you back when a specific device is in range, etc.
> screen dimensions
Not personal data.
> input latency
Fudge the numbers even when the user grants permission.
This isn’t a war, it’s just engineering for the benefit and autonomy of the user, and not the platforms’ commercial interests.
Fingerprinting and other quasi-malware techniques being used by desirable third party software are absolutely a war, otherwise we wouldn't have this permission structure in place in the first place. And as much as you believe the contrary, Alibaba et. al. are 100% smarter than you about how to exploit these tricks and will win.
I don’t appreciate the personal attack. You don’t know me.
Permissions have nothing to do with preventing fingerprinting, they have to do with user control. Every permission you add that the app can inspect is one extra bit of fingerprint, so it’s actually worse.
In TFA’s case, if the app was simply not hooked up to an actual audio sink, they would still be able to use this mechanism for fingerprinting, since it’s the processing pipeline itself that leaves the fingerprint.
Note that the article isn’t calling out for fingerprinting specifically. I’m sure AliExpress does more fingerprinting than this. The problem is specifically that they’re interfering with multipoint Bluetooth audio.
The best you can do on the modern web is reduce your fingerprint footprint, though it comes at a cost of websites breaking from JS disabling, or local time zone anonymisation.
Soundd like a nightmare to build legitimate apps if you for example are building an app that uses the camera but you can’t in anyway tell that using the camera fails (because user had denied the permission 6 months ago and has no memory of it) and instead of being able to give a helpful error you are just ending up with I am guessing fake images (maybe just a black screen).
Just because it's hard doesn't mean you shouldn't try. What about a popup saying "xyz has tried to access the camera a lot lately, continue blocking?"? Or apps can apply for set of 'core permissions' that can't be denied but are much more tightly controlled? Apple and Google are some of the biggest companies in the world, they don't deserve as much leeway as you're offering.
Between humans, I feel that what we like to call a "good communicator" as opposed to someone who just rattles off facts or prepared statements comes down to the "theory of mind" skill and how advanced that is. The presenter knows what they have to say but beyond that they maintain a real time internal representation of the state of mind of the listener and continuously update their delivery based on that. LLMs today seem to achieve this to some degree(?) but its interesting to think of how far you could advance that skill. I think great human communicators develop a sense of different ways that people think over time and quickly get a sense of someones signature thinking patterns when communicating with someone new for the first time
I'm in my mid, approaching latter 40's, and have had significant change to how I think and communicate in the last couple years. I always struggled to communicate in the moment unless I was recalling rote rehearsed things, which didn't get me super far. Outside of that, I could eventually communicate something on the fly, but it was real rough, meandering, and did not instill confidence. I mostly got things done thru async methods, and stumbled through things like meetings.
In the last couple years, my skills here have vastly improved. My speaking circuit as it were, is able to run somewhat on its own now and without the direct pipe into my thinking it always required before, so my thinking is free to do other stuff, like keep a mental model of the audience, think about where the conversation is going, what questions someone might have, etc.
Reflecting, I have a couple theories on what happened. First, I have spent the last 10 years reading aloud to my kids at bedtime. That started as simple Seuss like stuff, but of course now is full-on literature. When you're reading aloud, you really have to work that muscle of speaking "behind" where your eyes actually are on the page, because you're also processing "who is this, what is their emotion I have to inflect, what voice was I using for this character, how do I pronounce this crazy word" etc. I think if you do anything daily for 10 years, you're gonna gain some proficiency.
Second, when my kids started taking music lessons, I figured I needed to set an example for practice and commitment, so I started learning piano, and practicing daily. Similar deal, it requires the same type of buffer where your thinking can work ahead of where your other thread is playing. Reading the music, processing the shapes, dynamics, tempo, emotion, remembering how you handle that difficult thing. Again, doing it daily, it builds some ability that I bet contributes across domains.
Without realizing it at first, the same patterns started to come out in meetings, and presentations, I could work ahead of what I was saying, I had way more buffer to process and deal with things that are the "nice to haves" above and beyond just getting the basic words out.
I suppose my long-winded point, is that it's way more developable a thing than I ever presumed possible. I always expected I would remain an awkward communicator, and hitting a big skill-up way into my 40's was a real unexpected development. I'm sure there are many other tracks to get there, outside of reading aloud, and learning an instrument.
In thinking of directions where LLM's could develop from here, I cant help but think that a models ability to self introspect would immensely improve their utility. The R&D on how to achieve that is beyond me though. How do you train someone how to introspect? Also would it require a continuous learning architecture that doesn't separate training and inference?
I think it’s possible this might just require reworking the transformer architecture a bit to give it an inherent awareness of the contents of its attention. Kind of a working memory workspace of sorts.
I.e. Instead of limiting autoregression to the token level, you introduce a persistent compressed global workspace latent memory vector that is fed back into the self-attention mechanism at every layer or every token step, allowing the network to attend to its own prior attentional states before computing the next token. Obviously that’s going to involve some compression steps.
Trouble is… I think the architecture there is much simpler a tweak than figuring out how to train it.
…that’s likely to just destabilize training
for not much if any gain at first. You’re probably gonna have to resort to some really clever (and currently missing) tricks to figure out how to train the network to actually use that feature.
Well, the algorithms in social media that are rewarding "engagement" (but accidentally ripping society apart by fueling rage) have made the training data — online discussions — a very poor training medium for introspection.
So, if companies want training data on introspection, they would need to first fix what they broke (trust in each other and society) and have those discussions and realizations made public (by bringing back blogs where people might jot down their "maybe I've been an asshole" realizations) and rewarding other things than "engagement".
I get the impression this isn't really what is meant by introspection here. I think it means much more plainly that the model is aware of its own "state of mind" so to speak, not so much the sense of reflecting on one's actions. I don't think you'd necessarily need training data about humans reflecting on their behavior for the former to come about in a model, i think it'd have more to do with the architecture of the model (does the model allow for "awareness of the state of mind")
reply