Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google Fights Back

stratechery.com

Post reply on HN
Print view View original

451–460 of 475 posts

Re: Google Fights Back

#451

Post by martinlaz »

icebraining wrote: post #370
martinlaz wrote: post #248

Earlier quoted context omitted.

> Google collects so much absurd amount of data - that all the Proton-mail, DuckduckGo, Wire/Signal, Firefox (loaded with adblocking and tracking plugins) apps in the world can't keep you totally from it as long as you're on Android. You disable things, you opt out of stuff and it just keeps on collecting anyways. There must be some kind of way outta here. How about carving out a Google-free chunk of the web, and the…

Not crazy, but you would be excluding 85% of the top 100k sites, including stuff like Github. I'd rather just try to block the Google domains themselves.

Well, that's a good start. It even already includes cool stuff like Hacker News and Wikipedia. From there on, site owners need incentives to enter and remain in the "cool" zone. No idea ATM how/if that can be realized but there seem to be enough bright folks out there wanting it to happen.

> I'd rather just try to block the Google domains themselves.

This works to some extent but falls short of really penalizing the site owner for installing the trackers (of course, they probably lose some revenue but that's not enough).

If sites had a rating reflecting how "tracky" they are, then, given a choice, people who care would tend to pick the less tracky one. Benevolent browsers, search engines, and other tools can enable them to make this choice.

Re: Google Fights Back

#452

Post by gnode »

kerkeslager wrote: post #450
gnode wrote: post #49

Earlier quoted context omitted.

It's not really about it being cloud based. It's about having a central trusted party with access to private data which has public utility. As an example, Tesla collects lots of private data about their users' driving, in order to train their self-driving tech. Doing this in a free software solution (if you consider the training to be part of the software) would be less practical, as it would require training journey…

> Doing this in a free software solution (if you consider the training to be part of the software) would be less practical, as it would require training journeys to be publicly available. I think most free software folks would agree that the training data doesn't have to be public. I do agree that there's some problem to be solve here, though. Part of the point of free software is to be able to trust that the code yo…

> I think most free software folks would agree that the training data doesn't have to be public.

That seems antithetical to me. The trained model is the program, and the training data is in-essence the source code.

> manipulating what an AI does by manipulating the training data

You don't even have to do this. You can manipulate the model directly. Writing neural networks manually isn't difficult. You can't do more with that than conventional programming, but its sufficient to add malicious behaviour.

Re: Google Fights Back

#453

Post by neop1x »

loudtieblahblah wrote: post #70
areoform wrote: post #7

Google owns my soul. Their boxes know when I sleep, when I wake, how much I exercise, what I listen to, my innermost thoughts, my chats with loved ones, what I watched on Netflix last night, what my company does, the flu I have at the moment, what the hypochondriac in me looks up in the middle of the night, what I buy, whom I call, what I spend money on, where I spend it... I nominally pay for these services, but I s…

Google is already not Google. The "Don't Be Evil" corporation that valued open source rather than open-washing, that valued openstandards over "oops, we didn't mean to break that for you!" isn't here anymore. The company that bends over backwards, much farther than the law requires, to enable the surveillance state. I despise Apple. Especially on mobile. No SDCards, no headphone jack, walled garden app stores. Ugh ug…

There are things I would miss on iOS (e.g. widgets, dual sim, sdcard, headphone, more RAM) but for example Backup is is so much better on iOS! I fear that my android breaks at some point and I lose some data because there is no proper way to do full system backup. I have to use as much cloud services as possible and it has always been a big pain for me to upgrade to a newer android device. And I remember how straightforward it was and probably is on iOS.

Re: Google Fights Back

#454

Post by TeMPOraL »

pdimitar wrote: post #445
TeMPOraL wrote: post #379

Earlier quoted context omitted.

Also, for those believing in free markets: freedom to tinker for you is also the freedom to ask your friend or pay a local professional to tinker on your behalf. It doesn't mean everyone has to be a tinkerer, only that everyone can be. The opposite of that is requiring to go through the official vendor for every little thing you need to tweak.

I don't believe in free markets. I am too much of a realist for that. :) I ask you what can we realistically do though? It's very apparent that budget Androids that lag like hell and break down often aren't to the general populace's taste if they have a choice -- in my "poor" country (Bulgaria) people get loans so they can buy the Galaxy S10 or Huawei P30, en masse. Most people I've known in my life buying budget And…

Given that all Androids offer you the "freedom to tinker", you can always shell out for the Galaxy instead of iPhone if that's what you care about. I understand either choice on individual level. Myself, I'm on Galaxy S7 now, previously on S4. Before that, I bought a cheap Android phone and learned the hard way that what you save in money, you'll repay back with interest in mental health. The "death from a thousand cuts" isn't worth it, and I recommend everyone around me too to save up for a better phone instead of taking the cheapest one.

If the only Android phones where the cheap, shitty ones, I'd probably be on an iPhone now.

Re: Google Fights Back

#455

Post by BeetleB »

nullc wrote: post #420
BeetleB wrote: post #320

Earlier quoted context omitted.

>You think everybody has time to read the dozens of pages of legalese in Google's privacy policy, which anyway doesn't actually specify in which ways they abuse one's information? No, but I do think that everyone who gets a service for free should by default assume they don't have much privacy. That is not a huge burden to assume.

It is increasingly difficult to not use one of these "free" email services because of their aggressive blocking of any autonomously operated email servers.

Has Fastmail ever been blocked?

I've used a paid email provider since before Gmail existed. They've not been blocked.

I know there are the occasional posts on HN about Gmail blocking their servers, but I doubt it's anywhere close to the norm.

Re: Google Fights Back

#456

Post by groovybits »

blub wrote: post #423
groovybits wrote: post #148

Earlier quoted context omitted.

FWIW, I agree with this sentiment. Google did not need to give this stance at I/O. It appears that they're being as transparent as they can about the fact that our data is their business model. I get it: our online privacy is important. It is becoming increasingly more difficult to remain truly anonymous on the Internet, and this is mostly thanks to Google. However, the converse could be Google using our data for mal…

But they did need to do this, because it's great for them. Now there's two new apologies that can be breathlessly recited to excuse surveillance capitalism: * Google are brave for admitting they have your data! * Sure it's bad they have so much data, but they can be so much more helpful . Your reply's funny in a way, because if you jump from the beginning to the end it reads like "I get it: our online privacy is impo…

> if you jump from the beginning to the end it reads like "I get it: our online privacy is important" and then you give N reasons why it's not.

I don't believe I was listing reasons why privacy is not important. I was simply noting a benefit that has come from Google's data collection: increased heuristics to combat malicious online activity. I believe this is a valid thing to consider.

Regarding your list of evil deeds Google has done, lets refer to the definition of 'evil'. According to Merriam-Webster (https://www.merriam-webster.com/dictionary/evil), I believe we're focusing on the 'morally reprehensible' aspect of Google's actions.

> they lock people out of their online lives

I have seen reports of folks being locked out of their Google accounts because Google marked them as 'suspicious', which ultimately locks them out of logging into many other sites that they choose to link their Google account with. I agree that this is a valid concern. If this is an intentional act by Google, I would say this is morally reprehensible. Is it intentional though? And how quickly does Google typically fix these lock-outs? (I genuinely don't know).

> out victims to their stalkers

> manipulate people into handing over their data with dark patterns

I'm not sure what context you're referring to with these two.

> subvert or take over web standards

Google has contributed to many web standards. However I have not seen them 'subvert' or 'take over' any existing standard. Is there an example of this?

> conspire to keep employee salaries low

I don't have salary reports for Google employees in front of me. I will not pretend to know the business logic for their payroll dept. I think the only folks who can accurately answer this are actual Google employees.

> sometimes serve malware through their ads

Through Google-owned ad companies? I have not seen an example of this. Malvertising is a real concern, but I have not seen this in the wild on a Google/Alphabet-owned website or on a site that serves ads directly from AdSense.

> crush competition by illegally promoting their properties on search

Yes, I have seen what you are talking about here. Is this illegal though? And is it crushing competition? If I search for 'duckduckgo' or 'bing' on Google.com, I get results for those websites.

> illegally prevent Android suppliers to use Android as they wish

Again, would love examples. AFAIK, phone manufacturers are free to alter AOSP however they wish. Again, what is the illegal part?

> They've been found guilty in a court of law multiple times on both sides of the ocean.

Guilty of what, and are those things morally reprehensible?

--

To make my point clear, I completely agree that Google has a lot of power in their data collection, and it is something we should be concerned about. The reason why I'm replying is because I am genuinely curious about a lot of these points. Lets not romanticize the problem. Provide hard facts. That is the only way things change.

Re: Google Fights Back

#457

Post by kerkeslager »

gnode wrote: post #452
kerkeslager wrote: post #450

Earlier quoted context omitted.

> Doing this in a free software solution (if you consider the training to be part of the software) would be less practical, as it would require training journeys to be publicly available. I think most free software folks would agree that the training data doesn't have to be public. I do agree that there's some problem to be solve here, though. Part of the point of free software is to be able to trust that the code yo…

> I think most free software folks would agree that the training data doesn't have to be public. That seems antithetical to me. The trained model is the program, and the training data is in-essence the source code. > manipulating what an AI does by manipulating the training data You don't even have to do this. You can manipulate the model directly. Writing neural networks manually isn't difficult. You can't do more w…

> That seems antithetical to me. The trained model is the program, and the training data is in-essence the source code.

No, I think that trained models and training data are entirely new concepts. I can see the similarities between trained models/programs a and training data/source code, but there are obvious privacy concerns with training data that don't apply to source code. Calling the training data the source code is a leaky abstraction.

You could, for example, equally argue that the trained model is the source code, and the program which operates on the trained model is an interpreter which runs that source code.

Even if you refuse to concede that there are differences between training data and source code, you'll note that I am talking about free (libre) software, not open source. There are lots of cases where free software doesn't mean opening all the source up. Free software is about freedom, not about open source. For example, while there's plenty of source code in settings files, I don't think there's anyone arguing that we need to check in our settings files on open source projects. The point is to put the power in the hands of users. Putting users' data out of their control is the antithesis of that.

> You don't even have to do this. You can manipulate the model directly. Writing neural networks manually isn't difficult. You can't do more with that than conventional programming, but its sufficient to add malicious behaviour.

That's an interesting point, and may provide a workaround making training data public, i.e. you can generate the model and then make a representation of the model public, rather than the training data. However, there's an assumption here that the training data can't be reconstructed from the trained model--if a study of this exists, I'm not aware of it.

Re: Google Fights Back

#458

Post by 8ytecoder »

nirvdrum wrote: post #375
8ytecoder wrote: post #209

Earlier quoted context omitted.

I use my DSLR to shoot pictures in RAW and still don't fill ~10GB of space in a full vacation. Granted different people have different needs - but anything over 32GB seems to work fine for me.

I have the OS taking up space, games that now take up 1+ GB, some offline music, and offline maps so I can use the GPS in rural areas. All of that eats up a sizable portion of that 32 GB. Add in photos and 1080p video and it doesn't take long to fill up. If 32 GB works great for you, awesome. But I didn't make my story up :-) Now my phone has 64 GB based storage and I'm still in the same situation because everything…

Not doubting you at all. Also, I recently bought a 128GB iPad and copy my pictures to it - from phone and camera (using the SDCard adapter for iPad). It works great for me. iPads are cheaper and last quite a bit.

Re: Google Fights Back

#459

Post by aerique »

djsumdog wrote: post #374
aerique wrote: post #304

Earlier quoted context omitted.

I don't understand why no one mentions SailfishOS. It's an alternative that is available right now . Sure, it's not up to the level of Android and iOS in most respects, but it takes effort to get something like this going. Also from the users. Like Linux 20 years ago. Developing for it is not a nightmare either. And it's running Linux so there's advantages as well. I used it as my daily driver years ago and I've been…

I've looked into Sailfish before. Isn't it closed though? (It might have a Linux kernel, but that doesn't mean the rest is open) and as far as I can tell, it's not (easily) available in US markets.

It's not as open as the Librem phone, I think. What do you mean with "open"?

What I meant to indicate is that it is a viable alternative to iOS and Android, available right now. (Except it's hard to get outside Europe, China, Russia and India apparently. I missed that.)

Re: Google Fights Back

#460

Post by gnode »

kerkeslager wrote: post #457
gnode wrote: post #452

Earlier quoted context omitted.

> I think most free software folks would agree that the training data doesn't have to be public. That seems antithetical to me. The trained model is the program, and the training data is in-essence the source code. > manipulating what an AI does by manipulating the training data You don't even have to do this. You can manipulate the model directly. Writing neural networks manually isn't difficult. You can't do more w…

> That seems antithetical to me. The trained model is the program, and the training data is in-essence the source code. No, I think that trained models and training data are entirely new concepts. I can see the similarities between trained models/programs a and training data/source code, but there are obvious privacy concerns with training data that don't apply to source code. Calling the training data the source cod…

I think the trained model is analogous to a program (which may be interpreted by a virtual machine; not necessarily a machine code program) because it is not intelligible by humans. I'll admit that there are tools for analysing neural networks, but these are like disassemblers.

Trained models carry all the hazards of binary blobs, and so can't just be trusted. Training data is no true analogue to source code, but the concept of reproducibility is still relevant. At the very least, the production of blobs should be auditable.

Post reply on HN