9 ms·

I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itse

I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. The quality of such a system should be comparable or better to server-side technologies with the right amount of tuning, and much less privacy-invading than existing approaches. Personally I'd love to support more websites through ads if two conditions could be met: - A way to ensure that ads don't try to harm me by e.g. leading me to websites serving malware or abusing my computer's resources (e.g. miners) - A way to keep my privacy and control what data is collected about me (and who has access to that data) Currently I simply can't turn off the ad-blocker even if I wanted as most sites become completely unusable and outright obnoxious by showing large, blinking or content-hiding ads, videos, popups or fake overlays. That's why most people use ad-blockers (IMHO). If ads are decent, relevant and non-obtrusive I personally would be happy to see them. Also, go to any large website these days (without ad-blocker enabled) and check how many third-party trackers they load. There are many sites that send my data to more than 50 (!) different ad networks and partners, which is just insane.

ketzu 8y ago
> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. When I was at Middleware2017 I saw a poster/demo about this, MoCA+: https://koreauniv.pure.elsevier.com/en/publications/demo-moca-incorporating-user-modeling-into-mobile-contextual-adve https://koreauniv.pure.elsevier.com/en/publications/demo-moc... I guess the problem is the same as with privacy techniques in general. If you ask companies to restrict their access to data, they just tell you no(1), as it might be worth a lot of money or open up new business oportunities they haven't thought about yet. (1) This is anecdotal from projects of my colleagues in privacy research
- Nasrudith 8y ago
  I think another issue is the first rule of network security - don't trust the client.
  - BrendanEich 8y ago
    That cuts both ways, so we trust the client (users have to, even if remote sites do not) and do not trust servers to take individual user data in the clear and use it wisely & fairly. This applies to Brave servers too. Flip the model.
apinstein 8y ago
This is the big idea I always pitched at hackathons (like startup weekend). I think it’s a great idea. 100% opt-in ads. My tagline was “a sufficiently relevant ad is indistinguishable from content.” It’s meant to be aspirational. The idea is that if you are looking to buy something, awareness of choices or information about that choice should be so valuable that it’s considered to be a benefit by the user. It’s basicallt an inversion of control approach to ads. The cpc’s and cpm’s would be amazing. :)
- Hnrobert42 8y ago
  Unless I specifically seek them out, I have no interest in learning about new products or services ever. Thus, an ad would never be relevant enough to be content for me.
  - Y_Y 8y ago
    But you read HN don't you? Can you tell articles from ads here? What about Paul Graham's submarines?
    - freeone3000 8y ago
      Being tricked by false headlines and articles published as news when secretly advertisement doesn't lead me to opt in to more ads.
      - snaky 8y ago
        It depends on how useful the advertising is. It's not the tracking or privacy the regular user is most concerned about, it's the everyday experience that tells him advertising is still stupid and annoying. Despite all the billion dollars ML systems with gazillion of factors analyzed on petabytes of data, all we've got is 'hey, you googled for a 10uA accuracy bench multimeter yesterday! now for two weeks we will show you all the $15 multimeters ever existed!'. Look at this (don't worry, it's short one-page) thread for example of how it might be instead - https://www.eevblog.com/forum/testgear/benchtop-dmm-advise-needed-%281-ua-to-10-a-2w4w-ui-dual-display-labview%29/ https://www.eevblog.com/forum/testgear/benchtop-dmm-advise-n... - and what if the message from Tek engineer that closed a deal was actually advertising? He just posted a link to relevant appnote, that's it. What if automatical advertising system showed the same link to the same appnote as an advertising to the same user?
wbl 8y ago
I worked on this a while ago at Mozilla. There was a presentation at ACM CCS '16 on what we did. Unfortunately the effort was discontinued, but we definitely solved a lot of the technical issues.
- ThePhysicist 8y ago
  That sounds really cool, is there any documentation (papers, presentations, code) available of that effort?
  - wbl 8y ago
    http://delivery.acm.org/10.1145/2980000/2978407/p1591-green.pdf?ip=213.205.242.208&id=2978407&acc=OPENTOC&key=4D4702B0C3E38B35%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35%2E9F04A3A78F7D3B8D&__acm__=1537197475_671361c4f5d60f9a080c8fbb143edea8 http://delivery.acm.org/10.1145/2980000/2978407/p1591-green.... and https://www.youtube.com/watch?v=G62JD9fdxNM https://www.youtube.com/watch?v=G62JD9fdxNM
- forapurpose 8y ago
  Wasn't that how ads were actually implemented on Firefox's homepage several years ago? IIRC, Mozilla developed it in cooperation with IAB as a solution to privacy (the ad-funded web being practically unavoidable), deployed it (maybe only in alpha or beta), and then people went ballistic over advertising in Firefox - without realizing it was a huge privacy gain - and Mozilla pulled it. Here's one article I found quickly; note that ads are based on local browser history. https://www.cnet.com/news/mozilla-officially-kicks-off-ads-in-firefox/ https://www.cnet.com/news/mozilla-officially-kicks-off-ads-i... And it looks like Mozilla is trying it again, though this is from a few months ago. What happened to it? https://blog.mozilla.org/futurereleases/2018/04/30/a-privacy-conscious-approach-to-sponsored-content/ https://blog.mozilla.org/futurereleases/2018/04/30/a-privacy...
Kalium 8y ago
This is an intriguing idea! Thank you for bringing it forth. Out of curiosity, how familiar are you with how ad systems work? Generally there's a huge inventory of ads, each with different criteria attached to them and metadata about bids. When a request comes in, ads whose criteria are met are selected out and an instant auction takes place. The resulting ads are then shown. I cannot think of a way that allows for both honoring the ad criteria and keeping personal interests and preferences solely client-side. You can't even take advantage of any form encryption here, as the results of the filtering would allow the server to infer the private data. This means you'd almost certainly have to get all the data to the client to allow querying locally. OK, fine, that's workable. Then you just track impressions and so on to figure out when advertisers get charged. A potential drawback is that it's very possible that with impression and click data it could be possible to reconstruct most or all of the data a user might wish to protect. And there's no way to get away from this, either - pretty much all online advertising models rely on tracking one of impressions, clicks, and actions. As for your two conditions: > - A way to ensure that ads don't try to harm me by e.g. leading me to websites serving malware or abusing my computer's resources (e.g. miners) Policing the contents of ads can be quite the task. Ensuring the contents of arbitrary external websites is next to impossible. There are no good ways to do this in a fully automated system at scale when someone else controls the other server and can change what content it serves at their discretion. The best way I can think of to ensure this is to limit access to this hypothetical advertising platform to entities with the expertise and resources to protect themselves and anyone who comes into contact with their servers. Works for me, but being shut out of access to the biggest and best advertising systems might be a problem for many groups. > - A way to keep my privacy and control what data is collected about me (and who has access to that data) You know what? I think I know exactly what you want. You want the newspaper model. Collects no data from you, preserving your privacy. Only accepts advertising from partners that can be trusted, ensuring your safety. Doesn't need to closely track impressions, views, or actions.
- alkonaut 8y ago
  > Generally there's a huge inventory of ads, each with different criteria attached to them I think what he is describing is a future where the ONLY criteria ads are selected on is the standard user data set. Basically I open my browser settings, go to its "relevant ads"-section, enter some basic ad targeting info such as my age group, gender, and 2 hobbies. Then because I entered "fishing" as an interest, I'll see a lot of fishing gear ads. Great! If I'm NOT willing to enter any targeting info into my browser, or if I enter bogus info (or install a plugin that randomizes the info) then sites will not show me relevant ads. I don't mind seeing ads for fishing gear if I told the site I'm interested in fishing. That's completely fine. I do mind seeing ads for hotels in San Francisco just hours after I searched a different sites for cheap flights there, or ads for that exact shoe I made an incomplete checkout of in a webhop store last week etc.
amaccuish 8y ago
> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side I'm pretty sure iAd works like this no? Also it wouldn't be massive advantage, you'd be serving less relevant ads (so losing out to competitors) and you can still track which ads a device requests.
- ThePhysicist 8y ago
  Do you have more information on this?
Ajedi32 8y ago
> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. Correct me if I'm wrong, but isn't that the idea behind Brave? Blocks ads and (assuming you opt-in) replaces them with less obnoxious ads chosen by the client?
SmellyGeekBoy 8y ago
Or just show ads relevant to the content? What better indicator of what the user is interested in right now?
- nine_k 8y ago
  An ad which is relevant to 1% viewers still sells, likely enough to keep it published. But it's irrelevant to 99% of the viewers, and it's impossible to tell them from the 1%: among other things, because the viewers are not keen to share too much data about themselves. I think we should see ads as the (annoying) cost of commercially-produced free content. I wish a micropayment solution would take off to allow for easy and reliable paid opt-out on multiple sites I might visit. (I already have Youtube Red, and it's great.)
rock_hard 8y ago
Star Wars of all things provides a interesting vision here. Did you ever notice that for the most part the machines/robots are not networked? That AI seems to exist but is intentionally kept dumb (C3PO). I think this is a interesting model. To as society come to the conclusion that more connection of devices, smarter AI is not the right solution. That what works best is the right amount of connection and AI
- nine_k 8y ago
  See also: Butlerian Jihad.
  - rock_hard 8y ago
    Oh interesting https://en.wikipedia.org/wiki/Butlerian_Jihad https://en.wikipedia.org/wiki/Butlerian_Jihad
- wlesieutre 8y ago
  Except that they helpfully leave unsecured network ports all over literally every ship and building so that R2 units can plug in and do whatever they want. We can ignore that part for a “right amount of connection” example.
chimeracoder 8y ago
> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. Mozilla did this for Firefox, except a step better: the client pulled the same set of ads from the server regardless of the user preferences, but decided (client-side) what to display, so the server could never even infer behavior about the client from the requests. Unfortunately, people didn't care, and they complained that this was an intrusion of privacy regardless, so they dropped it. It turns out, there just really isn't market for privacy-focused advertising. People who care about privacy generally dislike advertising in all forms and block it, without regards to whether the advertising actually is an invasion of privacy or not.
- hnaccy 8y ago
  >so the server could never even infer behavior about the client from the requests. It reports click through so this isn't true.
- forapurpose 8y ago
  > people didn't care > there just really isn't market for privacy-focused advertising My impression was that people just didn't understand it and Mozilla's communication about it was poor. That doesn't mean that there's not a market.
madrox 8y ago
This idea keeps coming up. There’s a few reasons it hasn’t happened: 1. To really leverage the personal data, you need to run it through a model and correlate it with ad inventory. Those are two things you don’t tend to want to deliver to an insecure client. Then you need to collect data on how those ads performed to update the model. 2. For all that, it’s still more convenient to manage GDPR opt-in. Don’t underestimate the convenience of centralized management.
tzs 8y ago
In the late '90s there was a place that did something like that. That was a time of slow internet connections, with a majority still on dial-up. They had a product for Windows computers that provided a caching, prefetching web proxy. The deal with it was that if you let it show you adds occasionally (I think it was whenever you launched your browser to your homepage or once a day if you visits to your homepage were less frequent) you could download all of the company's other Windows programs and use them free as long as your browser was configured to go through the proxy. These were programs that normally were sold on floppy or CD-ROM for $20-40 each. The way the ad delivery worked is that the proxy would download ads, which consisted of an HTML page and some additional data. When you went to your homepage the proxy would serve up one of the ads instead, with a link inserted to take you to your real homepage. There was a Forth-like language built into the proxy. The additional data for an add could include code written in this language, which would be run when the proxy was choosing an ad to show. The code had access to various things the proxy knew about the local system and user. It did not have any access to the internet. The code would decide how strongly the ad would like to be shown at this time. The only thing that went back to the internet in regard to ads, as far as I remember, was counts of how many times each ad was shown. (Not that it would have mattered much if more went back. As far as I know the proxy didn't really know much about you other than the physical characteristics of your computer and your internet connection. I don't think they had gotten to the point of trying to infer interests from browsing habits). This particular approach would probably not be feasible today. They only had a handful of ads available at any one time, with the inventory changing slowly by today's standards. It did not take much resources, even for modem users, to download the entire ad inventory and keep it up to date. Can you imagine trying to put Google's entire ad inventory on every PC, and keep it up to date, so that the client can choose the ad entirely locally?
- bobthepanda 8y ago
  This type of thing is also generally considered malware today. As a person who used to do IT support for a university, one of the most common malware issues would be Chrome extensions or whatnot switching your homepage to point to a ads-infested search engine instead of Google, and redirecting all search engine websites to the ad-infested one.
- 8y ago
mahmoudimus 8y ago
We're working on something similar to this @ my company. Feel free to ping me directly, I'm happy to elaborate.
BrendanEich 8y ago
Brave, with the Basic Attention Token (BAT), is building exactly the client-side anonymous contribution + ad-matching system you describe. We will take BAT to other apps after proving the model in Brave. BAT in Brave is opt-in -- each user consents before anything local happens with data or zero-knowledge/blind-token attestations -- and users can get _gratis_ BAT grants right now using the stable desktop browser (this is coming to mobile in about a month). The anonymous contribution system is the basis for the also-opt-in Brave Ads system, which uses local data only, local machine learning agent, and no cookies or user tracking by any server (even ours). Ads match against a catalog fixed daily or less frequently for a large set of users in a region who speak the same language. Attribution and confirmation use Chaumian blind tokens. Users get 70% of revenue for opt-in, user-private (in tab), high quality ads at user-configurable frequency. We are working with publishers to provide user-opt-in ads for sites too, 70% revenue to the publisher, 15% to the user. User ad trial is under way right now, ping me if you want to be included. System should be available in Brave 1.0 in a couple of months.
- user812 8y ago
  I'm a big fan of that idea. Imho the beauty of your BAT system (the way it is envisioned) is it's independency from the current model of monetizing the web, which is ads, gradually evolving towards direct transmission between publisher and consumer. Ads in the way they work on the web are just a very inefficient system of transferring this value, and they don't serve the original function of marketing anymore. It turned into a big game of psychological warfare. The system is so inefficient that it finances almost the complete operation of Alphabet/Google. As a user I don't know how the system works in the background, and when I read the recent news about Brave attacking Google for GDPR violation it was the first time I read about RTB and the technical aspects in the media. People need to know, so they can decide if they want to feed such a system! Funding Choices seems to be a part of Google's answer to the growing problem of ad-blockers, but it can also bee seen as Google's answer to competition like Brave/BAT. I read somewhere that under the umbrella of Funding Choices Google is also experimenting with subscriptions like BAT, but without the token. I don't know how successful Google is with this, but this might be a tough competition for Brave, they will fight tooth and nails, and they control the Android ecosystem. BAT is attractive for power users as you ride on the wave of privacy-friendliness which Google can't, but I think the real challenge will be the average user that wants a standardized, seamless cross-platform solution that can be used as the main payment gateway for accessing content, which is increasingly via gated Apps. With Google controlling so much of the market with Android and Chrome, I wonder how they will react to BAT if it ever becomes successful, as they could theoretically quickly scale any competitive project. I think the biggest advantage of BAT would be if big players could acknowledge it as a de-facto standard for decentralized transfer of micro-payments and privacy friendly ad networks. For this to happen it would be necessary to be somewhat "Open-Source", i.e. not strictly tied to a singly company controlling much of the tokens. I am thinking in line of an open consortium with different players holding a significant part of the BAT tokens each.
eleniS2 8y ago
We are also building such a product. The core idea is: Your browser knows you best. We leverage this to display offers tailored to your browsing behavior that include a valuable user benefit without any of the information ever leaving your browser. You can read more of how we do this in https://myoffrz.com/en/fuer-nutzer/ https://myoffrz.com/en/fuer-nutzer/. Plus the code is open sourced here: https://github.com/cliqz-oss/browser-core/tree/master/modules/offers-v2 https://github.com/cliqz-oss/browser-core/tree/master/module.... The product name is MyOffrz and is integrated in the Cliqz Browser/Extension as "Cliqz Offers" and in the Ghostery extension as "Ghostery Rewards Beta" for German users.

ketzu 8y ago

> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. When I was at Middleware2017 I saw a poster/demo about this, MoCA+: https://koreauniv.pure.elsevier.com/en/publications/demo-moca-incorporating-user-modeling-into-mobile-contextual-adve https://koreauniv.pure.elsevier.com/en/publications/demo-moc... I guess the problem is the same as with privacy techniques in general. If you ask companies to restrict their access to data, they just tell you no(1), as it might be worth a lot of money or open up new business oportunities they haven't thought about yet. (1) This is anecdotal from projects of my colleagues in privacy research

Nasrudith 8y ago
I think another issue is the first rule of network security - don't trust the client.
- BrendanEich 8y ago
  That cuts both ways, so we trust the client (users have to, even if remote sites do not) and do not trust servers to take individual user data in the clear and use it wisely & fairly. This applies to Brave servers too. Flip the model.

apinstein 8y ago

This is the big idea I always pitched at hackathons (like startup weekend). I think it’s a great idea. 100% opt-in ads. My tagline was “a sufficiently relevant ad is indistinguishable from content.” It’s meant to be aspirational. The idea is that if you are looking to buy something, awareness of choices or information about that choice should be so valuable that it’s considered to be a benefit by the user. It’s basicallt an inversion of control approach to ads. The cpc’s and cpm’s would be amazing. :)

Hnrobert42 8y ago
Unless I specifically seek them out, I have no interest in learning about new products or services ever. Thus, an ad would never be relevant enough to be content for me.
- Y_Y 8y ago
  But you read HN don't you? Can you tell articles from ads here? What about Paul Graham's submarines?
  - freeone3000 8y ago
    Being tricked by false headlines and articles published as news when secretly advertisement doesn't lead me to opt in to more ads.
    - snaky 8y ago
      It depends on how useful the advertising is. It's not the tracking or privacy the regular user is most concerned about, it's the everyday experience that tells him advertising is still stupid and annoying. Despite all the billion dollars ML systems with gazillion of factors analyzed on petabytes of data, all we've got is 'hey, you googled for a 10uA accuracy bench multimeter yesterday! now for two weeks we will show you all the $15 multimeters ever existed!'. Look at this (don't worry, it's short one-page) thread for example of how it might be instead - https://www.eevblog.com/forum/testgear/benchtop-dmm-advise-needed-%281-ua-to-10-a-2w4w-ui-dual-display-labview%29/ https://www.eevblog.com/forum/testgear/benchtop-dmm-advise-n... - and what if the message from Tek engineer that closed a deal was actually advertising? He just posted a link to relevant appnote, that's it. What if automatical advertising system showed the same link to the same appnote as an advertising to the same user?

wbl 8y ago

I worked on this a while ago at Mozilla. There was a presentation at ACM CCS '16 on what we did. Unfortunately the effort was discontinued, but we definitely solved a lot of the technical issues.

ThePhysicist 8y ago
That sounds really cool, is there any documentation (papers, presentations, code) available of that effort?
- wbl 8y ago
  http://delivery.acm.org/10.1145/2980000/2978407/p1591-green.pdf?ip=213.205.242.208&id=2978407&acc=OPENTOC&key=4D4702B0C3E38B35%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35%2E9F04A3A78F7D3B8D&__acm__=1537197475_671361c4f5d60f9a080c8fbb143edea8 http://delivery.acm.org/10.1145/2980000/2978407/p1591-green.... and https://www.youtube.com/watch?v=G62JD9fdxNM https://www.youtube.com/watch?v=G62JD9fdxNM
forapurpose 8y ago
Wasn't that how ads were actually implemented on Firefox's homepage several years ago? IIRC, Mozilla developed it in cooperation with IAB as a solution to privacy (the ad-funded web being practically unavoidable), deployed it (maybe only in alpha or beta), and then people went ballistic over advertising in Firefox - without realizing it was a huge privacy gain - and Mozilla pulled it. Here's one article I found quickly; note that ads are based on local browser history. https://www.cnet.com/news/mozilla-officially-kicks-off-ads-in-firefox/ https://www.cnet.com/news/mozilla-officially-kicks-off-ads-i... And it looks like Mozilla is trying it again, though this is from a few months ago. What happened to it? https://blog.mozilla.org/futurereleases/2018/04/30/a-privacy-conscious-approach-to-sponsored-content/ https://blog.mozilla.org/futurereleases/2018/04/30/a-privacy...

Kalium 8y ago

This is an intriguing idea! Thank you for bringing it forth. Out of curiosity, how familiar are you with how ad systems work? Generally there's a huge inventory of ads, each with different criteria attached to them and metadata about bids. When a request comes in, ads whose criteria are met are selected out and an instant auction takes place. The resulting ads are then shown. I cannot think of a way that allows for both honoring the ad criteria and keeping personal interests and preferences solely client-side. You can't even take advantage of any form encryption here, as the results of the filtering would allow the server to infer the private data. This means you'd almost certainly have to get all the data to the client to allow querying locally. OK, fine, that's workable. Then you just track impressions and so on to figure out when advertisers get charged. A potential drawback is that it's very possible that with impression and click data it could be possible to reconstruct most or all of the data a user might wish to protect. And there's no way to get away from this, either - pretty much all online advertising models rely on tracking one of impressions, clicks, and actions. As for your two conditions: > - A way to ensure that ads don't try to harm me by e.g. leading me to websites serving malware or abusing my computer's resources (e.g. miners) Policing the contents of ads can be quite the task. Ensuring the contents of arbitrary external websites is next to impossible. There are no good ways to do this in a fully automated system at scale when someone else controls the other server and can change what content it serves at their discretion. The best way I can think of to ensure this is to limit access to this hypothetical advertising platform to entities with the expertise and resources to protect themselves and anyone who comes into contact with their servers. Works for me, but being shut out of access to the biggest and best advertising systems might be a problem for many groups. > - A way to keep my privacy and control what data is collected about me (and who has access to that data) You know what? I think I know exactly what you want. You want the newspaper model. Collects no data from you, preserving your privacy. Only accepts advertising from partners that can be trusted, ensuring your safety. Doesn't need to closely track impressions, views, or actions.

alkonaut 8y ago
> Generally there's a huge inventory of ads, each with different criteria attached to them I think what he is describing is a future where the ONLY criteria ads are selected on is the standard user data set. Basically I open my browser settings, go to its "relevant ads"-section, enter some basic ad targeting info such as my age group, gender, and 2 hobbies. Then because I entered "fishing" as an interest, I'll see a lot of fishing gear ads. Great! If I'm NOT willing to enter any targeting info into my browser, or if I enter bogus info (or install a plugin that randomizes the info) then sites will not show me relevant ads. I don't mind seeing ads for fishing gear if I told the site I'm interested in fishing. That's completely fine. I do mind seeing ads for hotels in San Francisco just hours after I searched a different sites for cheap flights there, or ads for that exact shoe I made an incomplete checkout of in a webhop store last week etc.

amaccuish 8y ago

> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side I'm pretty sure iAd works like this no? Also it wouldn't be massive advantage, you'd be serving less relevant ads (so losing out to competitors) and you can still track which ads a device requests.

ThePhysicist 8y ago
Do you have more information on this?

Ajedi32 8y ago

> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. Correct me if I'm wrong, but isn't that the idea behind Brave? Blocks ads and (assuming you opt-in) replaces them with less obnoxious ads chosen by the client?

SmellyGeekBoy 8y ago

Or just show ads relevant to the content? What better indicator of what the user is interested in right now?

nine_k 8y ago
An ad which is relevant to 1% viewers still sells, likely enough to keep it published. But it's irrelevant to 99% of the viewers, and it's impossible to tell them from the 1%: among other things, because the viewers are not keen to share too much data about themselves. I think we should see ads as the (annoying) cost of commercially-produced free content. I wish a micropayment solution would take off to allow for easy and reliable paid opt-out on multiple sites I might visit. (I already have Youtube Red, and it's great.)

rock_hard 8y ago

Star Wars of all things provides a interesting vision here. Did you ever notice that for the most part the machines/robots are not networked? That AI seems to exist but is intentionally kept dumb (C3PO). I think this is a interesting model. To as society come to the conclusion that more connection of devices, smarter AI is not the right solution. That what works best is the right amount of connection and AI

nine_k 8y ago
See also: Butlerian Jihad.
- rock_hard 8y ago
  Oh interesting https://en.wikipedia.org/wiki/Butlerian_Jihad https://en.wikipedia.org/wiki/Butlerian_Jihad
wlesieutre 8y ago
Except that they helpfully leave unsecured network ports all over literally every ship and building so that R2 units can plug in and do whatever they want. We can ignore that part for a “right amount of connection” example.

chimeracoder 8y ago

> I'm still surprised no one has yet figured out a way to store sensitive data about personal interests and preferences on the client-side and let the client itself pull appropriate ads for the user to see. Mozilla did this for Firefox, except a step better: the client pulled the same set of ads from the server regardless of the user preferences, but decided (client-side) what to display, so the server could never even infer behavior about the client from the requests. Unfortunately, people didn't care, and they complained that this was an intrusion of privacy regardless, so they dropped it. It turns out, there just really isn't market for privacy-focused advertising. People who care about privacy generally dislike advertising in all forms and block it, without regards to whether the advertising actually is an invasion of privacy or not.

hnaccy 8y ago
>so the server could never even infer behavior about the client from the requests. It reports click through so this isn't true.
forapurpose 8y ago
> people didn't care > there just really isn't market for privacy-focused advertising My impression was that people just didn't understand it and Mozilla's communication about it was poor. That doesn't mean that there's not a market.

madrox 8y ago

This idea keeps coming up. There’s a few reasons it hasn’t happened: 1. To really leverage the personal data, you need to run it through a model and correlate it with ad inventory. Those are two things you don’t tend to want to deliver to an insecure client. Then you need to collect data on how those ads performed to update the model. 2. For all that, it’s still more convenient to manage GDPR opt-in. Don’t underestimate the convenience of centralized management.

tzs 8y ago

In the late '90s there was a place that did something like that. That was a time of slow internet connections, with a majority still on dial-up. They had a product for Windows computers that provided a caching, prefetching web proxy. The deal with it was that if you let it show you adds occasionally (I think it was whenever you launched your browser to your homepage or once a day if you visits to your homepage were less frequent) you could download all of the company's other Windows programs and use them free as long as your browser was configured to go through the proxy. These were programs that normally were sold on floppy or CD-ROM for $20-40 each. The way the ad delivery worked is that the proxy would download ads, which consisted of an HTML page and some additional data. When you went to your homepage the proxy would serve up one of the ads instead, with a link inserted to take you to your real homepage. There was a Forth-like language built into the proxy. The additional data for an add could include code written in this language, which would be run when the proxy was choosing an ad to show. The code had access to various things the proxy knew about the local system and user. It did not have any access to the internet. The code would decide how strongly the ad would like to be shown at this time. The only thing that went back to the internet in regard to ads, as far as I remember, was counts of how many times each ad was shown. (Not that it would have mattered much if more went back. As far as I know the proxy didn't really know much about you other than the physical characteristics of your computer and your internet connection. I don't think they had gotten to the point of trying to infer interests from browsing habits). This particular approach would probably not be feasible today. They only had a handful of ads available at any one time, with the inventory changing slowly by today's standards. It did not take much resources, even for modem users, to download the entire ad inventory and keep it up to date. Can you imagine trying to put Google's entire ad inventory on every PC, and keep it up to date, so that the client can choose the ad entirely locally?

bobthepanda 8y ago
This type of thing is also generally considered malware today. As a person who used to do IT support for a university, one of the most common malware issues would be Chrome extensions or whatnot switching your homepage to point to a ads-infested search engine instead of Google, and redirecting all search engine websites to the ad-infested one.
8y ago

mahmoudimus 8y ago

We're working on something similar to this @ my company. Feel free to ping me directly, I'm happy to elaborate.

BrendanEich 8y ago

Brave, with the Basic Attention Token (BAT), is building exactly the client-side anonymous contribution + ad-matching system you describe. We will take BAT to other apps after proving the model in Brave. BAT in Brave is opt-in -- each user consents before anything local happens with data or zero-knowledge/blind-token attestations -- and users can get _gratis_ BAT grants right now using the stable desktop browser (this is coming to mobile in about a month). The anonymous contribution system is the basis for the also-opt-in Brave Ads system, which uses local data only, local machine learning agent, and no cookies or user tracking by any server (even ours). Ads match against a catalog fixed daily or less frequently for a large set of users in a region who speak the same language. Attribution and confirmation use Chaumian blind tokens. Users get 70% of revenue for opt-in, user-private (in tab), high quality ads at user-configurable frequency. We are working with publishers to provide user-opt-in ads for sites too, 70% revenue to the publisher, 15% to the user. User ad trial is under way right now, ping me if you want to be included. System should be available in Brave 1.0 in a couple of months.

user812 8y ago
I'm a big fan of that idea. Imho the beauty of your BAT system (the way it is envisioned) is it's independency from the current model of monetizing the web, which is ads, gradually evolving towards direct transmission between publisher and consumer. Ads in the way they work on the web are just a very inefficient system of transferring this value, and they don't serve the original function of marketing anymore. It turned into a big game of psychological warfare. The system is so inefficient that it finances almost the complete operation of Alphabet/Google. As a user I don't know how the system works in the background, and when I read the recent news about Brave attacking Google for GDPR violation it was the first time I read about RTB and the technical aspects in the media. People need to know, so they can decide if they want to feed such a system! Funding Choices seems to be a part of Google's answer to the growing problem of ad-blockers, but it can also bee seen as Google's answer to competition like Brave/BAT. I read somewhere that under the umbrella of Funding Choices Google is also experimenting with subscriptions like BAT, but without the token. I don't know how successful Google is with this, but this might be a tough competition for Brave, they will fight tooth and nails, and they control the Android ecosystem. BAT is attractive for power users as you ride on the wave of privacy-friendliness which Google can't, but I think the real challenge will be the average user that wants a standardized, seamless cross-platform solution that can be used as the main payment gateway for accessing content, which is increasingly via gated Apps. With Google controlling so much of the market with Android and Chrome, I wonder how they will react to BAT if it ever becomes successful, as they could theoretically quickly scale any competitive project. I think the biggest advantage of BAT would be if big players could acknowledge it as a de-facto standard for decentralized transfer of micro-payments and privacy friendly ad networks. For this to happen it would be necessary to be somewhat "Open-Source", i.e. not strictly tied to a singly company controlling much of the tokens. I am thinking in line of an open consortium with different players holding a significant part of the BAT tokens each.

eleniS2 8y ago

We are also building such a product. The core idea is: Your browser knows you best. We leverage this to display offers tailored to your browsing behavior that include a valuable user benefit without any of the information ever leaving your browser. You can read more of how we do this in https://myoffrz.com/en/fuer-nutzer/ https://myoffrz.com/en/fuer-nutzer/. Plus the code is open sourced here: https://github.com/cliqz-oss/browser-core/tree/master/modules/offers-v2 https://github.com/cliqz-oss/browser-core/tree/master/module.... The product name is MyOffrz and is integrated in the Cliqz Browser/Extension as "Cliqz Offers" and in the Ghostery extension as "Ghostery Rewards Beta" for German users.