Related: I recently wrote a blog post about using docker to containerize my ML research dev environments -- I've never found another way to make properly reproducible dev environments.
http://vedder.io/misc/research_dev_env.html
It's stupid sour grapes comments like this that make me hate HN commenters.
Meta and Google have both contributed an enormous amount to open source (creating PyTorch, TensorFlow, Chromium, contributing to clang/LLVM, Linux kernel, HTTP standards), but you're mad because they didn't sponsor your work and a bunch of people mindlessly upvoted it.
It's terrifying that we live in a world where a legislator saw the need to regulate this while I am not disagreeing with the need for this once brought to my attention.
Humans give explanations that other humans find convincing, but they can be totally wrong and non-causal. I think human explanations are often mechanistically wrong / totally acausal.
As a famous early example, this lady provided an unprompted explanation (using only the information available to her conscious part of her brain in her good eye) for some of her preferences despite the mechanism of action being subconscious observations out of her blind eye.
Insurance tends to have maximum profit percentages mandated, at least in the US. I'm not an expert in this, but I assume that this results in a bit of a perverse incentive where the only way to increase your profit is to increase your costs.
No, because their income as capped at the premiums collected. They want to spend 80% on expenses, but not more than that. But they also want projected expenses for next year to go up.
From what I'm seeing, I suspect insurance companies have multiple internal groups, at least one that want to reduce expenses and at least one that wants to increase revenues. Each is motivated independently of the other.
I've spent a significant amount of time playing with the variety of Diffusion models available and DALLE 2 tends to produce much better quality images. The other killer feature is DALLE 2 has support for in-fill.
>without government intervention you end up usually with a monopoly or oligopoly.
I would love to hear your explanation of how government intervention is the only thing preventing a monopoly or oligopoly of plumbers, electricians, pet groomers, or halal carts.
no explanation needed, just look at medieval history where guilds had de facto monopolies on certain services or manufacturing and if you wanted to be independent... tough luck.
> where guilds had de facto monopolies on certain services or manufacturing and if you wanted to be independent... tough luck
This conflates too many historical threads to support the hypothesis. The guild era was one where most countries required official consent to start a business of any kind. That's how the guilds enforced their monopoly.
I'm an adult who understands tracking tech and in particular Google's level of tracking quite well (I used to work on AdWords), and I'm one of these people -- tracking doesn't seem to negatively impact my quality of life at all but it keeps the many free services I value greatly running.
Visual or audio representations of the amount of data they're collecting are hardly an argument for why it's bad, as the logic goes "see, they're collecting a lot of data about your browsing... ...and that's bad" but it doesn't fill in the logical leap in the middle. You are going to be a lot more persuasive if you can fill that in instead of making fancy graphs.
Same here when it comes to how it affects me personally, but I can see how it affects society as a whole (or how it could affect kids) so easily.
Most people don't have an understanding about how their browsing is a part of a positive feedback loop that pushes them into echo chambers on Facebook, Youtube, TikTok, etc. It's a consequence of relying on advertising as a revenue generator for these services. Teaching your kids about how their data is being used for this purpose is important. Sending them raw DNS logs to do that is definitely not going to be effective though.
It's valid to point out to children that platforms have an incentive to suck them in and keep them on there, but this has nothing to do with ads.
Netflix is a paid subscription service that does not show you ads, yet they have just as strong of an incentive to keep you on their platform as much as possible so you don't go to a competitor. This is simply the nature of online services -- a better online service keeps you on there longer, and in a world of significant online competition, everyone is trying to suck you in.
The same is true of all media really, including books (e.g. the various tactics used by the authors of novels to keep the reader turning the next page). It's just that online media is more potent and compelling to most people so the effect is more noticeable.
> tracking doesn't seem to negatively impact my quality of life at all but it keeps the many free services I value greatly running.
Your data is not worthless, and generally I consider data harvesting harmful because of the nation-state subpoenas it enables.
If I attend a protest for abortion rights, and my geo data is bound to that event and then I seek an abortion in another state, ive now created a wonderful fingerprint that may soon lead to my incaerceration.
(For americans)
But the united states is not even the worst perpetrator. Imagine your least favorite government demanding google gives them a list of anyone that might be a political dissident based on audio they have collected.
Your complaint is with the government, not Google. The FBI could just as easily setup a StingRay [1] and capture all cell information. The only protection against that is robust digital civil rights legislation protecting Americans from governmental over reach, not lampooning big tech companies for having data. It's not like Google is willing to just shovel data at anyone who asks nicely, they will actively fight the government on nominally lawful requests it thinks are unreasonable/won't hold up to further legal scrutiny [2].
Well, the "before Google" was librarians, and they have a history, a long, long history of saying "no" to the feds when it comes to giving out reading and checkout lists of people.
*Of course* Google and friends happily give that information out "pre-warrant".
> Well, the "before Google" was librarians, and they have a history, a long, long history of saying "no" to the feds when it comes to giving out reading and checkout lists of people.
In the US they lost that fight and now reading lists are handed over to the state.
It's a valid point but, is pretty different in its intent from what GP comment is getting at.
This wasn't a result of "let's target this group we've deemed political enemies", it was "we're attempting to enforce this law (mind you, effectively everyone agrees with) via data collection".
I can buy how that's a pretty darn thin line, but, in practice I think it's also true that we haven't seen much evidence of the more shady interpretation that GP's post implies.
I think this example was even worse, because it wasn’t the government asking Google for data about a user, Google went out and proactively offered it (and erased this poor guy’s email and cancelled his phone).
There’s also been plenty of cases where people were charged with a crime on the basis of a warrant to find all phones in a given location at a given time.
The political-dissidents-from-audio-clips sounds a little far fetched, but certainly there have been cases where Alexa recordings were supoenaed and used against their owners. GP was certainly right that companies collecting data about me is a liability for me.
This a bit orthogonal so I put it in a separate comment, but let me be even more aggressive and do a bit of offense:
I've read a lot of the "pro privacy/anti tracking" arguments over the years, and (outside of a few exceptions) they almost always hinge upon some aesthetic dislike for companies having this information, rather than having any material justification for why it's harmful.
But attacking ads and ad tracking, the revenue model of the free web, is materially harmful. Imagine how much worse the world would be if Google search weren't free -- even a nominal fee of a few dollars a month would preclude most of the third world from accessing the best index of humanity's collective knowledge (not to mention how much worse the user experience would be to quickly Google something). Humanity collectively would be very non trivially worse off.
Seriously, what other business model besides tracked ads can generate enough revenue to keep the lights on, let alone power growth, at companies providing free services like Google Search, Maps, etc, that don't have such grossly negative externalities like severely curtailing human development and don't offend the aesthetic sensibilities of people who are prima facie annoyed by a company using the interactions with their free service to make money?
So let me bring up some of the exceptions I can think of.
Recently, a woman got an abortion and part of what the prosecution used against her was data collected from facebook.
With google's tracking, is it not unlikely that a government body could subpoena them for search histories? (which, they do already) to look for questions like "has this individual asked about abortions recently?"
Perhaps you view that as moral, but what about other things a totalitarian government might want to control? What about China trying to find the identities of people protesting them? Is it not unlikely that they might have search history data betraying that they are a protestor or likely protestor?
But then there is the question of why they even need that data in the first place. You argue "how else would they make money?" and I'd argue "well, ads?". The reason, the real reason, google collects this data isn't because they have to, it's because it's more profitable if ads are better targeted and more customized per individual. Google would certainly take a profit hit if they stopped the data tracking, but that doesn't mean they wouldn't still be one of the biggest and best places to advertise on the internet. They'd simply be less profitable.
At an object level, "not having XYZ actions occur from the government" is a problem with the government, not any tech company. You almost certainly live in a western style democracy, and so you possess mechanisms to modify the government behavior (if you don't, I suggest moving); it's absurd to complain about tech companies when you're actually upset with the government. This does doubly so when the primary threat to the ads tracking business model in the West is government legislation produced by elected representatives.
At a meta level, this is just a list of grievances. You've not suggested a workable alternative business model that, on balance, provides more net good. It's possible that all your complaints about tracking ads are legitimate and it's still the best business model for providing the sort of massively positive societal value Google search provides.
> is a problem with the government, not any tech company. You almost certainly live in a western style democracy, and so you possess mechanisms to modify the government behavior (if you don't, I suggest moving)
Yeah, I'll get right on voting for the candidate that has "limit the tracking of data by internet companies" as one of their issues. I mean, come on. This is a niche issue that has little public awareness. To suggest "Oh, you just need to vote better" is simply laughable. I (and I'm guessing most people) don't live in a government where you can simply vote in new laws or regulations, we vote for candidates and whether or not those candidates care about privacy rights is a crap shoot. Most people don't have a cursory understanding of how data tracking works.
This is a google problem. They don't have regulations pushing them to track their users in a way that makes their data easily accessible and consumable by government agencies. There aren't regulations out there forcing it. This is made evident by Apple's run ins with the FBI because their encryption wasn't crackable.
And frankly, while this is an issue, there are so many other issues I care about politically that a candidate running on the "limit google tracking" platform wouldn't have enough to win my vote. It's important, but so are so many other issues of the day.
> You've not suggested a workable alternative business model that, on balance, provides more net good.
More net good? Or more profit for google? These are not the same things.
I did, in fact, suggest an alternative business model, ads without tracking. One that was particularly popular throughout the internet right up until google took things over with doubleclick.
Again, this is not a business model that will be as profitable, but you are conflating "good" with "profit".
And as a counter example, duck duck go appears to be doing fine even though they aren't tracking users like crazy.
>Yeah, I'll get right on voting for the candidate that has "limit the tracking of data by internet companies"
Again, your list of complaints revolved about the government getting their hands on all this data and using it for purposes you don't like. If you're upset about that, there's plenty of people interested in civil rights.
But if digital privacy is such a niche issue, how come Europe and the United States have GDPR?
>This is a Google issue... ...the FBI...
Again, you're upset with the behavior of the FBI, not Google. Vote for people who restrict the FBI's ability to request this data instead of getting mad at the existence of tracking ad data.
>I did, in fact, suggest an alternative business model, ads without tracking. One that was particularly popular throughout the internet right up until google took things over with doubleclick.
And the fact that Google captures information about what sites you've been to in order to make ad auctions more efficient is so odious to you that you want to move everyone back in time to a demonstratively less efficient business model which hurts ad buyers (other businesses) at least as much as it does Google? You must know your proposed business model isn't nearly as good, but you've yet to articulate what exactly we're getting for it that makes the cost worth it.
>duck duck go appears to be doing fine even though they aren't tracking users like crazy.
DuckDuckGo uses Bing search under the hood, and they're not making nearly enough money to do R&D on things like making the next BeRT. If only DuckDuckGo existed, search algorithm quality would stagnate and then get worse (which is bad because SEO is by definition adversarially optimizing against the ranking algorithm in order to get a page to rank higher than it naturally would, thereby imputing noise into the natural relevance signal. Google is aggressively innovating on this front and still struggling.)
> DuckDuckGo uses Bing search under the hood, and they're not making nearly enough money to do R&D on things like making the next BeRT. If only DuckDuckGo existed, search algorithm quality would stagnate and then get worse (which is bad because SEO is by definition adversarially optimizing against the ranking algorithm in order to get a page to rank higher than it naturally would, thereby imputing noise into the natural relevance signal. Google is aggressively innovating on this front and still struggling.)
If you couldn't make money from ads like now, there'd be way less SEO. Most SEO just drags people onto ad spam sites with vapid auto content.
Your argument about DDG is equivalent to saying "look at this natural cyclist, he's not making any prize money compared to his doped rivals". Yeah, duh. You have to remove the doping to see how he really fares against the others.
>If you couldn't make money from ads like now, there'd be way less SEO
This is just nonsense. If there were no ads, hyperoptimized SEO would only be more important because it's the only way for you as a web admin to surface your content in people's searches.
Okay, and that's going to benefit blogspam and auto generated content how exactly? That's the SEO that people complain about. Not being able to search for reviews for products because the results are all spammy "BEST VACUUM IN 2022" type results. Googling tech support questions leads you to auto generated mirror sites that only exist to scam you into viewing ads. etc. etc.
Anyway, you conveniently ignored the rest of the post.
> If only DuckDuckGo existed, search algorithm quality would stagnate and then get worse ...
You appear to be one of the people who surprisingly still get good results from google.
There are dozens if not hundreds of discussions here in HN where people complain about exactly the opposite.
Google search results have become worse and worse in recent years, mostly except for search queries that involve finding buyable products anyway.
Many other searches where the goal is to simply find a text containing the search terms in a common context (simple retrieval of information) have become consistently worse, is my impression.
The results for these searches are typically littered with non sense sites that contain ads for related products and random collections of key words.
> You appear to be one of the people who surprisingly still get good results from google
Add me to that list. Google results are invariably superior every time I've tried out alternatives. I wish it weren't so, as there's still a lot of room for improvement with Google but they have such a monopoly they're unlikely to invest heavily in further advances if there's any risk it reduces their ad revenue.
Do you have some specific examples where an alternative web search returned better results than Google?
My point wasn't about google results possibly being superior to other search engine results. But about them having become worse and diluted by SEOed sites that sell something over time.
I'm a trillionaire who can hire maybe 10,001 minimum wage workers to follow you and the top 10,000 earners in the US around anywhere you go and record every single thing you do from a safe distance in public spaces (including stores, malls, church etc, looking in through your windows from the street). I'm not doing anything illegal, but you probably think it's creepy. How can I explain to you that it's not?
Earlier you said, tracking pays for free services you use but doesn't have a net negative effect on your life. How can you tell? And if you can tell, what's the problem with being informed of how often you're being tracked?
I've never seen such a concise and well-articulated statement of this position. Thank you for presenting it! I instinctively disagree with it, but I can't actually provide a coherent and convincing counter-argument, which is probably a sign that I need to think more on it. Thank you for that prompt.
The best counter-argument I can think of is:
> the claim that only Google (or, "only organizations large enough to require ads as revenue support") can provide these services is false - Open Source solutions (like OpenStreetMap, etc.) provide "good-enough" value. That is, the drop in value from "premium shiny BigTech solution" to "less-fully-featured Open Source privacy-respecting solution" is smaller than the gain in value from the associated privacy-aesthetics.
That's a subjective statement that applies on a case-by-case basis (is Google Maps more socially valuable than Google Drive?), and won't be true for everyone - perhaps, not many people.
A lot of AV companies bootstrapped their maps off of OpenStreetMaps -- it's great anything is free, but they kind of suck. The "cost" of a company using info about me to run more efficient ad auctions is totally worth the benefit of having highly accurate maps with up to date business info with stuff like hours and direct links to their website/contact info.
> The "cost" of a company using info about me to run more efficient ad auctions is totally worth the benefit of having highly accurate maps with up to date business info with stuff like hours and direct links to their website/contact info.
This is a perfect example of the "subjective statement that applies on a case-by-case basis" that I was referring to.
There's a free market of platforms without subpoena power, and people time and time again choose the free platforms powered by ads. I'm in the overwhelming majority with my statement when you look at people's revealed preferences.
The problem with your argument is that the same mechanisms that are used by government for broad-reaching surveillance are also used for legitimate government purposes. These are things like subpoenas and warrants, which are impossible to eliminate completely while keeping a functional court system.
You could try to argue that we should make sure that these are narrow in scope (that's already the rule despite the current practice), but that doesn't eliminate the problem completely: someone can already subpoena Amazon for "recordings taken in kylevedder's house between 4/22/2022 and 6/22/2022" related to an investigation, and they will get them. Even if you are not the target of the investigation, but they have reason to believe that the target may have visited you during that time period. Such a subpoena has not been tried in civil court yet, but it will likely be allowed. In an alternative world where tech companies don't track you, these subpoenas wouldn't work.
Google, Amazon, and other data collectors have a principal-agent problem with respect to data about you: while you would likely fight a subpoena for your location data tooth and nail, they don't care so much. They will often give up the data, and they won't even tell you they did.
The alternative business models are ad support without tracking and subscriptions. Google makes most of its money on ads that don't really need aggressive tracking, like ads for toasters when you search "toaster" or ads for other car brands when you search "Ford SUV." Arguably, the tracking might hurt their system since they try to produce fully personalized results for you. WolframAlpha is a search engine on a subscription model. Micro-subscriptions are already a thing (albeit invented/normalized after Google was invented), and it would be very likely that Comcast would bundle a Google subscription into your cable plan the way they do for entertainment products.
As far as this being bad for people, allowing companies to track you invasively is a little like not buying insurance. You won't care most of the time, but you may care a lot. Lots of people don't buy insurance, even though they should, because they severely underestimate tail risk. This kind of cognitive distortion is traditionally addressed with laws: social security, healthcare mandates (or single-payer healthcare), and car insurance mandates in some states all operate on this rationale. So should privacy legislation.
I don't see why you can't add guardrails to things like subpoenas -- it's not like they are allowed to engage in aggressive subpoenaing of non digital info without a due process procedure, just make the digital one more rigorous. The issue lies in the flawed execution of the current legal system, so fix that instead of taking a hatchet to big tech.
>Google makes most of its money on ads that don't really need aggressive tracking, like ads for toasters when you search "toaster" or ads for other car brands when you search "Ford SUV." Arguably, the tracking might hurt their system since they try to produce fully personalized results for you.
A very non trivial subset of their revenue is personalized ads, and I promise you they work pretty well; if you don't believe me I recommend taking a job on the ads team. This is a huge revenue hit, for a company that's consistently putting that money into product innovations as well as societally important moonshot projects, all in order to side step the issue of your concerns with the American justice system's subpoena process.
>it would be very likely that Comcast would bundle a Google subscription into your cable plan the way they do for entertainment products.
If you're a SWE you probably have coworkers raised in India. Ask them how likely they would have been able to afford such a subscription system and by extension how much worse their life would be if they didn't have access to Google. I would bet a substantial fraction of them would still be in India and not software engineers.
>This kind of cognitive distortion is traditionally addressed with laws: social security, healthcare mandates
We have healthcare mandates and social security because if a person has no money and is out in the street or gets health care and then can't pay, it's society's problem. If you get nailed by the government for doing something illegal due to invasive subpoenas, it's not only not society's problem, depending on your opinions on the role of the state this is a feature not a bug.
There are already guardrails in place on subpoenas. They are about as strong as they can be. That doesn't stop overreach, it doesn't stop corruption, and it doesn't stop people from being people when trying to get them and when granting them. The best way to stop abuses of a system is to remove the potential for abuse. Do you want to remove subpoena power from courts around the world? I don't - you kind of need it to prevent people from destroying and hiding evidence. The only thing that is left is helping people understand the harm that comes to their lives from giving intimate data up, and regulating its collection.
Also, privacy legislation does not need to "take a hatchet to big tech." It will just reduce profits for a few months while the ads team figures out how to work around it. I promise you they will do a really great job figuring it out. Some of the smartest people I know work on those ad models, and they will still be able to use things like location, web browser, and search query to find relevant ads.
Most of the people who can't afford a subscription fee (who, by the way, are not your Indian colleagues, who largely have come from rich or middle class families and can afford a subscription fee) are not worth serving ads to anyway. They don't have money to spend on stuff that you are advertising. I see no problem with offering a free tier with limited search capabilities, and a version with a subscription fee. Lots of SaaS does this. This is basically what Google does now with the ads: the high-value users (young rich people like the HN crowd) subsidize everyone else by having high ad impression prices.
By the way, I did work at G. I'm not convinced that personalization actually helps the bottom line at all for Amazon and Google search when you balance against the reduction in CTR due to latency (personalization prevents caching), and the added compute and engineering cost of personalization (since personalization prevents caching, you need a ton of additional servers and a lot of extra brains thinking about how to make them work). The idea of a search engine that produces personalized and highly specific search results feels like a local maximum to me in light of the costs and the research around CTR and web UX.
Unfortunately, that local maximum is currently so heavily entrenched in the market that it can strangle any competition. Google got its market share initially by having really fast search results that had the words in your query, and having no ads (and then later, a few clearly marked ads). This was in competition to web search providers that tried to offer personalized search results and highly relevant ads. Since then, they have become the thing that they proved was an inferior product, and they keep out upstart competitors through market power.
Putting this in a sibling comment because it's tangential to my more-relevant response about an alternative business model.
---
The claim that citizens of western-style democracies have mechanisms to modify government behaviour is technically true, but meaningless. The average Western citizen has more chance of becoming an Astronaut or solving a Millenium problem than they do of effecting real govenmental change without access to extraordinary social capital or Super-PAC-level donations. Especially given that most western-style democracies' mechanisms consist of "pick which of these two broad bundles of choices you want to support" (as cogman10 pointed out, "the candidate who is opposed to data-tracking" does not really exist as a viable option; and even if they did, they would need to have a broad platform of popular positions, not just that one) - so if the option you want is not offered, you not only need to change voting behaviour to support it, you first need to create that option out of whole cloth.
I disagree vehemently as someone involved in transit and housing activism. The cause did, does, and probably will continue to feel hopeless. There are days I walk out of talking with others about issues or days where a Mayor walks back a commitment that punch you in the gut. But through organization and by looking back over the last 5 years of activism in my community/area, I can see the real work that's been done. (It helps that our work results in infrastructure changes, of course.)
As far as a candidate, trust me there are hungry junior political candidates all the time that are happy to take campaign donations and meet their constituents demands. The quickest way to become a popular, well-funded candidate is to represent a wealthy, grassroots cause with no representation. If you're concerned about this, contribute to the EFF, and maybe even consider joining them! Don't put your anger into online zealotry if you live in a democratic country, use that energy to petition the government for change.
The EFF is hardly a "large organization". But yes, some amount of organization is necessary for political change. Fundamentally, an individual in a democracy pays into a system and cannot change the entire system at their behest.
And you think that complaining about tech companies, companies owned by share holders who have no particular obligation to satiate the complaints of non-shareholders, is a more rational strategy to achieve the outcomes you want?
Let's be intellectually honest here; the only actual threat to tracking ads as a business is governmental legislation, something which we've already seen occur with the implementation of GDPR in the EU and the US. It's pretty obvious privacy advocates have enough teeth to impact the legislature if they're able to ram those bills though, so I wholesale reject the argument that privacy is a niche complaint and you're powerless to create change in the government.
> And you think that complaining about tech companies[...] is a more rational strategy to achieve the outcomes you want?
No, I don't. I was pointing out that your claim that individuals can effect political change is broadly incorrect. That doesn't imply that the alternative (direct petitioning of companies) is effective.
Attempting political change probably _is_ a more rational strategy than petitioning tech companies. As an individual, neither seems powerful.
> Let's be intellectually honest here; the only actual threat to tracking ads as a business is governmental legislation
No disagreement here!
> I wholesale reject the argument that privacy is a niche complaint and you're powerless to create change in the government.
To be clear, these are two separate claims. Privacy might or might not be a niche complaint - I feel that it is (given how little my non-technical friends care about it), but I'm purely working off anecdata, and I concede that GDPR (and the slew of other Privacy regulations following in its wake) are datapoints that contradict me. The other claim - that individuals are powerless to effect political change - remains undisproven to me. In particular, I reject the position that _individual_ privacy advocates were responsible for GDPR. It's large groups like the EFF (as a sibling commenter pointed out) that cause governments to take notice, not individuals.
> Recently, a woman got an abortion and part of what the prosecution used against her was data collected from Facebook
Two crucial details you omitted:
1. Facebook didn't just hand over the data. It was a court order. At no point had the data just randomly got collected in some dragnet. They got a targeted precise warrant from the court to obtain specific messages of a specific person. Unless you got a court warrant against you, I don't see a reason to worry about here. None of this is related to "data collection".
2. It was unencrypted messages where the defendant was conspiring with another person to perform an abortion past 23 weeks (aka 5mo+, which is illegal almost everywhere in the world), and then conceal the crime by burning (in a literal fire) the physical evidence. Mind you, abortions are totally legal up to 22 weeks in Nebraska. Meanwhile, EU has only 2 countries with limits up to 24 weeks, with the rest being 20 weeks or less[0].
Tldr: if you send an unencrypted message to your crime co-conspirator (where you are discussing specific details of your actual plan to commit a crime), and then get caught via other methods, don't be surprised when the court sends a legal order to obtain your unencrypted messages as evidence.
It is a lot easier to get a court order for something than you probably think. Dragnet versions of court orders also exist, such as geofence warrants. Also, that data is often sold packaged up and sold to third parties via data brokers. The FBI has been known to buy data from data brokers instead of getting it by court order.
Sure, those things might happen. None of those things happened in that Nebraska case though, so using it to as a specific example to support your claim feels a bit off.
Especially considering the fact that the original comment omitted some pretty crucial details that end up painting the overall situation in a very different light.
What happened in the Nebraska case was more subtly bad: if the government had to subpoena HER for her location data on that day, there would likely be no criminal conviction at all. The 5th amendment would be a valid excuse to avoid giving out the data. However, they only had to subpoena Facebook, whose data is not covered by 5th amendment protections, and who would like to maintain a friendly relationship with governments.
EDIT: This is generally called the principal-agent problem, and there are lots of rules around agent behavior when it traditionally applies (eg attorney-client privilege and doctor-patient confidentiality).
Also, you should note that there are lots of crimes, and you probably commit a lot of them. If you have ever used someone else's WiFi without permission, that's one. If you carried a screwdriver or a sharpie into a store, that's a separate crime. The list goes on and on.
The fifth amendment provides a right not to testify against yourself. It has little to do with documents, whether they are in your possession or someone else's. The fourth amendment is the one that applies there. If her phone were recording location data locally (or if she were keeping a paper diary of her location, for that matter), the government could have obtained that information with a search warrant, and it would have been perfectly admissible.
What you're saying is true only if the data is unencrypted or not password-protected. If the data is encrypted, the government would have to crack it (which is not usually possible with decent encryption and anything short of the NSA breathing down your neck). They could not just ask you for the password.
The case law on compelled decryption is less clear-cut than you're describing. (Possible exceptions include: biometric, as opposed to password/passcode, encryption; and the government being able to prove that the defendant knows the password.) Moreover, even when the government can't obtain the decryption key from the defendant, cracking the encryption is not the only possibility. They could also seize the device while it is running, with the decryption key and/or decrypted data already in memory. That was done in the arrest of Ross Ulbricht, for example.
Also, this is entirely different from your original claim, because it has nothing to do with where the encrypted data is stored. If you store encrypted data in the cloud, then the fifth amendment provides you with the same degree of protection from disclosing the decryption key as it would for locally-stored encrypted data.
The attitude that if the government is using the data collected by tech companies to bypass constitutional limitation, then it's a problem with government and not big tech, is rather reductionist.
This is a tangled system and you cannot completely separate the issues as you so simply claim. Big tech colluded with government, and got nice advantages for their participation with the surveillance system. The citizen/consumer is massively under represented and underpowered to make changes to the system once it's in place.
Not to mention the completely immoral act of gathering very private information without consent. I bet that most ppl twenty years ago, if asked if they would consent to giving an exact timeline of their activities to third parties in exchange to free entertainment would probably not readily agree.
Big tech lied to consumers until it normalized the new surveillance status que, that's immoral and wrong, even if it's extremely profitable.
Your individualistic perspective, where you dismiss any harms if they do not relate to you personally, is very short sighted. Authoritarianism can creep up on you, first it targets the weekest, and before you know it, it could be you. Surveillance tech is a tool for authoritarianism, apart from it's inherent immorality because of how it was hidden from the consumer, it is corrupting democracies and gives them a very authoritarian tint.
Could it be that you can not understand the harms because you actively took part in creating this reality?
I know it's condescending, but it's also human nature. Nobody likes to acknowledge he is a part of something harmful, and to deal with this cognitive dissonance we can put on very powerful blinders.
1) Big tech colluded with government, and got nice advantages for their participation with the surveillance system -- they're getting raked over the coals at dog and pony show anti trust hearings.
2) completely immoral act of gathering very private information without consent -- you literally gave these websites this information, and there's legal privacy policies that no one reads describing how it's used. If you don't want them to have your info, simply don't give it to them.
3) Authoritarianism can creep up on you -- ok and? The fact that Google's ad auctions are more efficient isn't going to cause authoritarianism.
4) Could it be that you can not understand the harms because you actively took part in creating this reality? -- lmao no
1) The antitrust hearings are literally dog and pony shows. Nothing has come of them, except slightly increased lobbying spending by Google and Amazon.
2) Other people give it to them, not just you. I tried for years to keep my face off Facebook, and finally made a page to take control of my "shadow profile" because people kept putting it there. Companies often give it to them too, through data brokers.
3) Google's ad auction technology will make authoritarian government much more efficient, the same way IBM's census technology did in the 1930's and 1940's.
Upton Sinclair would be proud. Try using a smartphone today, which is becoming more and more required, without giving up PII.
- Our school district requires our kids to do business with Google, there is no way around around it without significant pain. They don't have another option.
- Last week my attempt at buying concert tickets was refused because I would not disclose my phone number and Ticketmaster no longer allows other options. A small thing but they add up.
- More and more restaurants won't accept cash.
- DMV is selling our information, ADP is selling our paycheck information.
The idea that we have "every choice in the matter" is already silly and getting falser by the day.
A significant quantity of the ad space is pernicious: The ads seeking to be displayed are explicitly intending to manipulate the viewer to change their cognitive orientation to the benefit of the buyer, with anywhere from callous disregard to outright contempt for the viewer's outcomes.
Big tech has significantly aided in the delivery of these harmful messages, and seeks to use personalized information to find the channels, times, and context that are _most_ effective at undermining the will and agency of the viewers.
Are there models where an information / recommendation broker could justifiably collect what would otherwise be concerning amounts of personal information? Potentially, if it were using that information to compute and facilitate the maximal individually beneficial outcomes of each individual user, but that's not what's happening here. Right now, we're building a system that would happily have people die painfully and preventably at 40 after living miserable lives because the economic returns of that model maximize the returns on advertisers / delivery networks.
> rather than having any material justification for why it's harmful
Material - During the bombings in Austin, Google provided a blanket location/data dump to the police of everyone with Google accs within x-miles of the event. That data, regardless of linked to the bombing in the end, is kept by Austin PD and likely the FBI, forever. You have given the police a warrantless log of citizens basically based on “well they use Google and live in Austin.”
If you don’t see the glaring material issues, that explains a lot about why Google is Google.
The police asked for stuff and the company complied and now you're upset that the government has it, but somehow it's actually Google's fault? Pass laws so your government can't do things you don't like.
>If you don’t see the glaring material issues, that explains a lot about why Google is Google.
I don't work at Google anymore. Feel free to click on my profile to see what I'm up to these days.
Putting aside the business model side of it (I can't speak for that), I agree with you about pro-privacy arguments tend to be annoyingly insufficient on the material justification ... when I can think of plenty.
Put simply, it enables massive power inequalities that threaten to be permanent, if they aren't already. Several living examples:
* Black advocates have already raised attention to pre-emptive tracking mixed with AI to "anticipate" crimes committed by "at risk" individuals (read, racial profiling) when this goes against everything that the Constitution stands for re: presuming innocence.
* The recent controversies with Roe that another sibling comment is a textbook example about how information mined from menstrual apps or location services can be used to criminalize women seeking an abortion.
* On a similar tangent, there was a case early on in adtech's history when a young woman who was pregnant and hiding it from the people she lived with got found out because ads about baby supplies were shown to them (after she had searched up the items). People's living situations can be volatile and this could have been fatal for her if they were sufficiently abusive enough.
* Health/Medical apps tend to be especially obscene with tracking - collecting data on the most medically vulnerable (eg disabled people) and letting that data be resold often times means locking them to a sub-standard life because jobs and insurance can and will discriminate in obvious and subtle ways if they know about certain medical conditions, if they can get away with it.
> even a nominal fee of a few dollars a month would preclude most of the third world from accessing the best index of humanity's collective knowledge
So you seem to be referring to Google (annual revenue ~$250 billion.)
But Wikipedia (annual revenue $160 million) also qualifies.
Also The Library of Congress (annual budget $800 million.)
Another interesting data point is the USPS $77 billion annual revenue.
So there are definitely proven business models that provide services basically at-cost and without all the negative externalities. And Wikipedia's budget is so small that even relatively poor countries could fund it with less than 1/5th of their education budget. (e.g. Uganda's education budget in 2020 was roughly $700 million.) And that would let them pay globally competitive salaries to hire software engineers to build indexes that are actually optimized first and foremost for knowledge that's focused on their culture rather than the US.
Lack of privacy harms journalism and activism, making the government too powerful and not accountable. If only activists and journalists will try to have the privacy, it will be much easier to target them. Everyone should have privacy to protect them. It’s sort of like freedom of speech is necessary not just for journalists, but for everyone, even if you have nothing to say.
You can have ads without tracking. I don't think anyone is against ads per se. People are against pervasive continuous tracking and wholesale sale of data to anyone.
You know Google doesn't do this, right? Besides being against their own terms of service (and outright illegal), they're economically incentivized not to; they use the data to make their ad auctions more efficient, an edge they would lose if they sold the raw data.
“Free services” are not free. They are ad-subsidized, and thus anticompetitive. They are affecting your life by messing up the competitive market economy. The fact that you value them greatly means you should pay for them, which you are - with your limited attention span. Its an insidious price to pay.
You may accept this bargain, but most of us hate it, and are not being offered a competitive market alternative to this version of reality.
They are using their ad business (wherein they have an effective monopoly on search engine advertising) to subsidize their other businesses. That is a fairly standard definition for predatory pricing in the US.
Holly.. you made me understand why vertical anti-trust is a thing. (I'm slow..).
A single company, google, gives a photo-service for free, while making enormous bread through ads. On its own the free service would be dead on arrival but now it's sustained by the ads branch of the company, making it unbeatable. Then pure-photo-service competitors die and you get a monopoly.
You're asserting that's not how they work, but you're also asserting they're illegally subsidizing their search and maps business with... the revenue they generate from selling ads space on their search and maps businesses.
I'll leave it as an exercise to the lurkers to decide who is probably right.
> you're also asserting they're illegally subsidizing their search and maps business with... the revenue they generate from selling ads space on their search and maps businesses.
Yes, that is how these things tend to work! Seriously, read some books on antitrust, you have no idea what you are talking about. Tim Wu is a good starting point.
A majority of Googles revenue comes from running ads on their own sites. Their third party ad network is a small part of their revenue, so the correct direction is to say that Google is subsidising its third party ad network via their first party ad business.
The effect of antitrust would then be that Google separates the ad service for its own services and third party services, which would force their third party ad service to become more expensive since it can no longer rely so heavily on the draw of all the Google properties.
> they're illegally subsidizing their search and maps business with... [their ads business]
I suspect this is what what @arrosenberg is getting at. If the Search (and maps) business is considered separate from the ads business (Search runs at a massive loss) and it's provided for free because they can subsidize it with the revenue from their ad business, then it's an anti-competetive advantage they have from a monopolistic position that their advertising business has.
Anyway, personally I have no clue how that "works" in practice, but that seems to be the perspective.
2) the service and the advertising are tied together, with the amount of money being earned by said advertising is private.
The second point can be addressed without necessarily banning advertising, but having the advertising separate from the services - you have some kind of ad platform/etc that by itself is neutral and you can browse it (and look/click at ads) to earn money, which you can then choose to spend on anything, including other services. This means services still have to compete on price, even though that price remains "free" to the end-user as they're using the advertising platform to earn money to then spend it on services.
Yes you can, just make yours also free and put up ads like DuckDuckGo.
>the amount of money being earned by said advertising is private.
Google is a publicly traded company with a legal obligation to disclose financial data to their shareholders as a matter of public record. You can go read their quarterly earnings reports stretching back years.
> Visual or audio representations of the amount of data they're collecting are hardly an argument for why it's bad, as the logic goes "see, they're collecting a lot of data about your browsing... ...and that's bad" but it doesn't fill in the logical leap in the middle. You are going to be a lot more persuasive if you can fill that in instead of making fancy graphs.
But that just means the novel representations are a step forward, just not the whole solution.
Making the invisible visible creates a situation where people can start asking questions about that thing, instead of ignoring or being ignorant of it.
>But that just means the novel representations are a step forward, just not the whole solution.
No. The part where people argue Google is collecting a lot of data is sound. The part that's missing is why is this bad.
No chart, no numbers, no sounds, or anything else will in their own argue why this data collection is bad. This argument as to why is what's missing, and until nominal privacy advocates start putting together coherent arguments as to why it's bad (and, importantly, why these costs do not merit the benefits), they're not making forward progress.
This form of data collection and use has broken a status quo that has existed as long as life itself. Shouldn't it more conventionally be the ones in favour of this change to provide evidence of its benefits rather than the other way around?
They already did. People vote with their feet and attention, and people continue to flock to free services powered by targeted ads, systematically out competing services that had other models. If you don't like these services, don't use them; they don't have subpoena power. However, you might find the other options to be lacking, because they have inferior business models that do not allow them to compete on product quality.
Not all cards were on the table back then when. The general public and perhaps even Google and such themselves didn't really know what they were getting into or where this was headed - at least not today's scale, ubiquitousness, or invasiveness. Hence I would argue that this contract or trade of sorts was not a conscious choice.
By the way, for what its worth, I really do appreciate all the time and effort you are spending on this discussion and the defence of your stance in it.
People continue to flock to new, free, targeted ads powered platforms. Look at the enormous growth of TikTok in the US; it has exploded well after the discussion on targeted advertising entered the American zeitgeist (Cambridge Analytica happened in 2018). In light of this, I don't think you can argue that Americans would have chosen something different if they knew more about targeted ads -- they are still, to this day, choosing to join free targeted ad platforms because they like using apps with that business model even when the app is effectively controlled by a foreign power.
I appreciate your kind words. I think people tend to be pretty myopic about targeted ads ("I don't like them/find them creepy") and fail to reason about all the positive things they are getting because of them, and as a consequence they fail to suggest alternatives that are actually better for the general population.
I think it's sad, because a huge amount of good for humanity has come out of Google search and other Google products being free to access while generating enough revenue to truly innovate for 20+ years. It didn't have to be this way, and one can easily imagine a world where Google was paywalled behind institution licenses, limiting its access to only powerful companies and wealthy universities. That world would be so much worse for all of us, and it was avoided because Google found the ads business model. People are now attacking that model but aren't suggesting truly viable alternatives (Google's biz dev is hard at work trying to crack this nut as well).
Google making ad auctions more efficient and "state surveillance" aren't the same thing. Did you legitimately not know this or are you being intellectually dishonest in your argumentation?
Google making ad auctions more efficient becomes state surveillance the instant a subpoena arrives at their door. You might believe you are safe from a subpoena because you haven't done anything wrong, but you would be wrong about that:
* Things like geofence warrants are used to surveil people who are merely within a given radius of a crime.
* Subpoenas don't have to be for criminal cases, they can also be civil or investigatory (eg congressional subpoenas).
* Things you don't think are wrong are often crimes. Abortion clinic visits in Texas are the most obvious example. Other examples are things like using someone else's WiFi without permission, using a fake name online, or carrying around things like screwdrivers (which are a burglary tool) or permanent markers (which are used for vandalism). Estimates suggest that the average American breaks 3-5 federal laws per day.
Copy and paste from elsewhere because it's the same talking point over and over: big tech is not the government, and your complaint is with the government. The FBI could just as easily setup a StingRay network [1] and capture all cell information. The only protection against that is robust digital civil rights legislation protecting Americans from governmental over reach, not lampooning big tech companies for having data. It's not like Google is willing to just shovel data at anyone who asks nicely, they will actively fight the government on nominally lawful requests it thinks are unreasonable/won't hold up to further legal scrutiny [2].
If you're going to go after someone, at least go after the right people.
Stingrays and other technical things that people could do are a lot more expensive than just asking Google or Verizon for their records. Also, that kind of indiscriminate tracking is a fourth amendment violation, so the government actually can't do it legally (and the evidence gathered by such a technique can get thrown out in court). Both of those things make a big difference.
The right people to go after are the ones who are not bound by the fourth amendment and who are creating a huge pool of data that is available for the government to grab. Technical solutions that collate data have been very useful tools for government overreach over the last century. IBM's census tracking system in the 1930's is a very well-known example of this sort of technology: a well-meaning innovation that efficiently organized Germany's demographic data and made it accessible... to the Nazis who wanted to find Jews.
If you were able to fight a subpoena for your own records (for example, because you owned the data instead of Google), I promise you that percentage would be a lot lower, solely because the incentives are different. Google doesn't particularly care about protecting you. Google cares a lot more than that about protecting its status with politicians. What this means is that they will do the minimum possible review on a subpoena for data on your account to not get sued by you.
There is a massive corpus of research on this that I think you’re unaware of. Start with the Snowden leaks and work your way back.
Since I can’t reply to below, the general approach of “I don’t know about this so it must not exist and I won’t research” is very backward if you’re in a PhD program.
I get what you’re getting at. Google just builds the technology, the rest is up to the users.
That said, re: Snowden, IC piggybacked off Google cookies and direct taps into Google infra (PRISM). Email, chat, videos, photos, stored data, file transfers, logins, all tracked.
The classic approach from tech until recently is what I said above - we just build it! That didn’t work well for the cigarette companies.
In other words, it is difficult to convince people that they are doing wrong/causing wrong when their research and pay is dependent on not believing they are.
The Snowden leaks were about NSA data collection. Last I checked the NSA was not a tracking ads powered big tech company.
Also I hope you realize pointing me to an entire metaverse of memes is not an argument. If this corpus is so massive it should be easy to concretely point to several very specific examples, not hand wave at the entirety of one of the largest national security leaks in American history.
This reminds me of the Slashdot days where commenters would have a hard time distinguishing between the RIAA, the MPAA, and the Government (TM). They were roughly convinced that the 3 just formed this nebulous malicious entity.
I’ve worked for the govt and no they aren’t a nebulous blob of power with tech.
What they are though is an entity that has a set of directives supported by access to the data tech provided in conjunction with either not understanding tech or really getting it.
The tech company approach of staying willfully ignorant or positions like ITT with “we just build it!” has caused a lot of damage in the hands of the govt users it’s delivered to.
I don’t think it’s possible to in good faith ignore this dynamic or pretend it doesn’t exist and tech products’ role in it. This is actually the first time I’ve run into it like this where both the dynamic and the impact is just totally denied.
> I’ve worked for the govt and no they aren’t a nebulous blob of power with tech.
Yes and I've worked for tech companies that use tracking in their products, though not (my products) for ads, so I know what a tech company can and cannot do.
> The tech company approach of staying willfully ignorant or positions like ITT with “we just build it!” has caused a lot of damage in the hands of the govt users it’s delivered to.
This isn't what anyone in this thread is saying at all, this is a strawman. There are material concerns about collecting too much data. Everything from leaked credentials allowing attackers to access personal data to accidents in targeting systems allowing individual level targeting.
But unless you have proof that there are internal systems in tech companies that build up surveillance platforms and that the government is given access to these surveillance platforms, both of which need proof, then this nebulous concern of "all data gathering is EVIL" is just FUD. The fact of the matter is, individual surveillance isn't useful for most tech companies. Cohort or product level information is required to make decisions. Moreover, storing surveillance data is extremely expensive. The amount of drive space and the systems necessary to facilitate this kind of per-human level surveillance requires large engineering teams and systems. Have you ever created a data warehouse? Imagine that but for this supposed surveillance machine.
> I don’t think it’s possible to in good faith ignore this dynamic or pretend it doesn’t exist and tech products’ role in it. This is actually the first time I’ve run into it like this where both the dynamic and the impact is just totally denied.
Yes everyone on HN and Twitter are convinced that data gathered by tech companies is all evil and that evil billionaires are twirling their mustaches as they enslave entire nations. This is clear. And there are clear risks to data gathering, cogent concerns about social media addiction, and definite monopoly concerns over the networks and moats that social media creates. But to have a coherent conversation on this, we need to move away from "all data gathering is EVIL" to understanding what kinds of data gathering is necessary and useful, and offering individuals rights over their data. TikTok was able to create its own network from scratch which is at least a small counterpoint to the idea that the incumbents are impossible to unseat due to their network advantages.
GDPR is a great step in the right direction and I'm hopeful for even more potent legislation to come out. But the sort of fearmongering that happens in these online spaces over data gathering makes it impossible to have a reasoned conversation.
And if you think tech companies are terrible entities abusing your data, then what are the telecoms doing? Telecoms in every country control the flow of information in every direction. The metadata available at a telecom down to an individual subscriber far dwarfs what is available at most tech companies. Moreover the PRISM scandal actually happened because of telecom companies. So why the ire at tech companies?
You should read up on cybernetics for what I mean regarding mass collection of data leading to undemocratic outcomes, my guess is you haven’t ran into it. But, large area of research.
For the obtuse, NSA's primary sources are big-tech, among others. There's an information-superhighway from one to the other. But you must have known that already, right?
Information is power. Lot of examples from history, say WWII, holocaust and Japanese internment from IBM census data. Stasi, Red scare, blacklisting, Alan Turing coming and going.
More recently, IRS targeting the tea-party, facial recognition, DEA parallel construction, data brokers, data breaches, Chinese persecution, social credit.
Yes, Snowden outlined in his writing what was specifically immoral and undemocratic, such as x-key-score sourced largely from google and friends. Educate yourself.
It extends its tendrils into government services where corporate authoritarianism subverts democratic ideals. The ill-conceived attempt to force face id for IRS services via a third party is but one example.
This feels more like "anti-libertarian" rather than "anti-democratic." The state having fairly tremendous power through warrants and other policing systems is largely orthogonal to the idea of government by the will of the people.
With financing through ads comes a dependency of the content provider aka publisher. And this leads to a lower quality. Its about getting attention, not about quality content. It does affect you, you just dont see it, I dare to say.
That said, the biggest tangible negative impact of tracking is that the profiles companies build on you are overly simplistic caricatures that reinforce societal biases. For example, for many complex reasons, women do not apply to executive roles as much as men. Google is thus less likely to show me an ad for an executive role knowing that I’m a woman [0], since statistically, I’m less likely to click it. This further reinforces the societal trend of women applying to executive roles at a lower rate than men.
> tracking doesn't seem to negatively impact my quality of life
It’s nice you’re willing to make this trade, but obviously some of us are not. I value my privacy, and I do not want Google to know what I’m interested in, how old I am, or my gender. I find the fact that Google tries to infer these things about me deeply, deeply creepy.
Google also goes out of their way to collect information you don’t want to share. Turn off “Location history” on your phone, and Google Maps will no longer remember where your “Home” is (or, actually, it will claim it won’t remember, but it will still randomly tell you how long it will take to get home). This is of course nonsense. I can want to tell someone where I live and also not want that same someone to track everywhere I go in minute detail. Google’s position here is kind of like an abusive partner saying if you give them your phone number they obviously have the right to see everything in your phone. All the while claiming “You’re in control!” Soooo creepy!
The more insanely creepy things Google does with my data, the less I trust them. Again, it’s fine if you trust them. You’re free to trust anyone you want (even my abusive ex partner, although I’d caution against it). But some of us don’t trust Google, and we don’t want to move to the opt-out-village[0] so here we are.
You're only seeing things from the Google bubble where the data they collect is kept only for themselves and law enforcement/nation states. There's a whole ecosystem of unscrupulous trackers selling your personal details indiscriminately to anyone who wants it. It all goes into profiles that most definitely have an impact on you in real life. No amount of goodwill for supporting free services is enough to permit this activity to operate without your knowledge or consent.
> tracking doesn't seem to negatively impact my quality of life at all
Like a frog boiling in hot water, just because you don't perceive it, doesn't mean it doesn't affect you. The business model of ad-based services is that they sell the ability to manipulate real world behavior to the highest bidder. That manipulation ranges from innocuous ads for products you don't need, to political ideas and misinformation. There's a chance that you're self aware, self disciplined and very observant and manage to avoid this manipulation. But given the fact that it works extremely well at scale, it's more likely that you're just not noticing how you're being manipulated.
I once had an idea for a project where I would roughly attempt to translate personal information into a value amount. For example, how much is it worth to companies to know your gender, on average? You could probably figure this out with some rough ideas of CPC for various segments. Obviously the economics don't work quite the same as money, but there's something to it.
This is all to say I don't think it's "bad" in the way malware is bad or a security breach is bad, but there's a dollar amount attached to these things, and you may be giving a company more money than you're willing to pay.
If everyone started blocking their data even when used in aggregate we would have no more free services. $10/month for video hosting (YouTube,), $50/month for GPS navigation (GMaps), $5/month for IM (Messenger)
Do you know for sure that surveillance capitalism isn't pushing up the prices of your health insurance because you once googled a symptom, or your flight prices because your search history reveals your disposable income? Or even, taken to extremes, it could even preclude your from even getting health insurance because it can prove knowledge of a pre-existing condition.
Do you know for sure that surveillance capitalism isn't pushing up the prices of your health insurance because you once googled a symptom
Yes, I do, at least for the very large majority of Americans who get their health insurance through their employer. As problematic as that relationship may be, it does set up a wall such that (a) the cost to me, personally, is completely standardized; and (b) there's a large gap between my individual online identity and my identity as an employee within my organization.
Further, health insurance is just about the most heavily-regulated industry in the universe. Aside from the trivial assertion that it's absurd to refer to this industry as "capitalism" because it's so far from a free market, the fact is that the insurance companies are watched very closely, and I think it's improbable that they could have a setup like this without anyone having learned about it.
Finally, again for the majority who are covered through their employer, most of these policies are self-insured. That is, it's the employer themselves who pays the charges, and they're also paying the insurance company to administer the program, but not to shoulder any risks. The more the insurance company has to spend on that administration, the less likely they're going to be hired. To make such a trade-off there would need to be a pretty clear demonstration to the employer that more money is being saved in medical payments than is being spent on that additional administration - and again, that's consistent with my argument in the prior paragraph.
The visualization is useful for those who already understand how it harms them and society. If someone likes to be watched and controlled, this type of visualization is not of much value.
"if you know you know" isn't an argument. This is just empty rhetoric -- dogmatically asserting that tracking ads damage society and that I'm being controlled by it is a pretty strong claim, and a claim that requires evidence to support it.
If this is the typical quality of your privacy argumentation then it should be no surprise that your children don't buy your arguments.
You're not going to win this "fight" on HN, I've been here for years. People here are convinced that every piece of data being collected is a moral evil.
They'll vaguely wave at things like DNS logs or server IP logs, and always assume that everything is always feeding huge tracking machines at big companies. They'll mix up tracking used for logs with tracking used for ads with tracking used to improve the product. There's no real understanding of what cohort sizes or signals are needed to make tracking meaningful. It's a mess.
IMO it's good to be sensitive about the data we send and hold tech companies accountable for how much data they take. I've also worked on big tech teams that collect data. But the arguments here are always so black-and-white that they don't make any meaningful point.
The target audience for my comments are the hundreds of people reading and not commenting. If you're casually reading the comments, without a foil of reasonability, I think it's easy to get sucked into implicitly accepting the premises of the zealots frothing at the mouth in the comments. You're right that I don't imagine I'm going to change the mind of the zealots.
I, for one, appreciate the points you are making here, and the doggedness with which you are making them. I basically feel the same way, but I don't usually comment on these discussions because of the general shrillness of the discussion.
I think it is a combination of shell shock reaction and lack of data literacy. Some companies have abused their sysadmin privileges, and people now have much less trust for other big entities. I also work in the data business albeit currently in operational/industrial data. We can abuse robots all we want as they have no rights. Cyborgs on the other hand are much more complicated. I understand the scale and types of data impacts.
I wasn't making an argument about privacy with that statement. I said that if certain knowledge is valuable to you, then visualizations of that data would be valuable. Some people like to be watched and controlled similarly to how some people want to be ruled by rulers. I accept that as human nature. If you are watched by someone in a greater power position who wants power over you, do you think this gives them more control or power over you or not?
Probably the most interesting trick from the paper is using the head as a soft supervisor for earlier layers of the network, with the intuition being that if the earlier layers learn to imitate the higher capacity later layers, it frees up the capacity of the later layers to better learn the residual and provides more dense supervisory signal.
Yes, but to my surprise the "compound scaling" provides 3x more improvement in their ablation study. Also, I don't understand Table 8 in their ablation study for aux heads, specifically: why does it have different base benchmark values from Tables 6 and 7?
