Hacker News new | ask | show | jobs
by ForHackernews 2075 days ago
Honestly, how would anyone ever know if Google does what they say they're doing? Is there any trusted third party auditing Google's ad auctions? If Google says $X was the price at auction, how would I ever dispute that?
4 comments

How would anyone ever know that Facebook hasn't designed it's algorithms for spreading misinformation and disinformation? How would anyone ever know that Apple doesn't use your photos to train an AI? How would anyone ever know that the Amazon doesn't peek into your RDS database?

https://www.nytimes.com/interactive/2019/06/08/technology/yo...

>> Several current and former YouTube employees, who would speak only on the condition of anonymity because they had signed confidentiality agreements, said company leaders were obsessed with increasing engagement during those years.

>> One problem, according to several of the current and former YouTube employees, is that the A.I. tended to pigeonhole users into specific niches, recommending videos that were similar to ones they had already watched.

There are real human beings working on those systems. That's how everyone would know if Google does as it says

If something fishy indeed goes on in a large company, it must be on a need-to-know basis.

But the higher access you get, higher the penalties of sharing company secrets go.

Snowdens come into the picture every once in a while, but most average programmer Joes wouldn’t risk being internationally manhunted for whistleblowing questionable business practices.

So I do not think your argument is valid.

While it’s possible for companies to keep secrets, most companies lack the secrecy and paranoia required to really keep big secrets for long periods of time. These companies tend to leak, both because people move around a lot, and because unlike governments they lack the ability to throw leakers in jail.

If google was actually futzing about with the auctions like that, we’d probably find out eventually.

Snowden is running from the government, not from a corporation.
Facebook has designed its algorithms for spreading misinformation and disinformation, because emotion drives engagement.
>> Facebook has designed its algorithms for spreading misinformation and disinformation, because emotion drives engagement.

If emotion drives engagement, then this should apply to positive emotions as well. All emotional posts, positive or negative, will be treated equally by the algorithm according to your logic. Hence it follows, that facebook hasn't explicitly designed it's algorithms to spread only misinformation and disinformation, acc to your logic.

I would like to see some proof that would back up your statement.

Two points:

1. It doesn’t matter what the valence of the emotions are - positive or negative valence feedback loops will spread misinformation and disinformation. Positive emotion is not correlated with truth.

2. If you haven’t heard about this idea before, here is a starting point: https://www.nytimes.com/2019/12/05/opinion/digital-technolog...

>> Positive emotion is not correlated with truth.

I assumed from your previous comment(below) that you are correlating negative emotion with misinformation and disinformation. If that's not what you meant, then I guess I misunderstood what you meant by your previous comment

>> Facebook has designed its algorithms for spreading misinformation and disinformation, because emotion drives engagement

The easiest negative emotion to cause with a post is outrage, which goes nicely with fake news.

The easiest positive emotion to cause are at the result of puppies and wholesome/faith-in-humanity-restored posts, neither are generally related to actual truthful news.

I don’t see how that assumption is implied by anything I said in what you quoted. There is no reference to the valence of the emotion.
If emotion drives engagement and you optimize for engagement, then you will optimize for emotionally-charged posts over emotionally-neutral posts. If you then assume that emotionally-charged posts are more likely to be mis/dis-information, then you have a case.

I don't think this assumption is necessarily true for random misinformation (think common myths), but propaganda is usually designed to be emotionally charged, from PR to state propaganda.

The whole debate isn't even wrong in the first place as the questions are beyond "have you stopped beating your wife yet?" and into adding absurdities as qualifiers like "have you stopped beating your wife with a prized family heirloom yet?".

So bad it arguably qualifies as misinformation in itself. If an algorithim spreads any information and cannot classify it (if you have a general purpose algorithim which can know all truth apriori what are you doing here instead of creating a singularity!). Then /design/ which implies intent and effectiveness. Otherwise if I burn a Trump Voodoo doll I will have designed something with the "intent" to kill the President. Even if Fox News would be very offended by it nobody sane would take it as a serious assassination attempt.

That only makes sense if your voodoo doll didn’t actually kill the president. Facebook does spread misinformation.

As to intent - at some point in the distant past before it became clear what it’s effects were, you could argue that Facebook wasn’t designed to spread misinformation.

Once you know what a thing does and how it works, if you keep operating it, it is by design.

Hang on, you think there are individuals who can reason about how complex systems work with perfect certainty?
Heck, Thinking there are individuals who actually understand such a large system are complex.

Even at organizations orders of magnitude smaller than Facebook, it's pretty easy to have a few systems that are large enough you can only keep so many in your head at a time.

You'd probably need a few dozen of the right people together to begin to understand the complexity of their systems.

If you're truly paranoid, you can certainly encrypt data in RDS with keys that you have and Amazon doesn't.

But yes, the rest they expect us to take on faith. Or trust some boilerplate in their ToS (written by their lawyers, to absolve them of liability).

Yes, and you could do anything else that you want with your app. However, if your application is hosted on their servers you are just a traditional user from their perspective. And as the old adage goes, if you own the server, you own the user.

There is an implicit assumption, that the code that you push is actually the same one that is being run on the VM, in your statement. This brings me back to the point I was trying to make in the parent comment.

The encrypted traffic will be the traffic Amazon will be most interested in, since you took the trouble; they won't peek into the DB but they will be able to infer lots about what's going on if they want. It's the Tor Paradox.
I understand the line of reasoning above and would normally follow it - but Facebook, Apple and Google have all shown that they will do really shady things if they can get away with it. The reaction to Project Dragonfly was pretty nice to see, but you really have to wonder how many things like that make it through without any public outcry - I know real human beings and most of them are pretty awesome, some of them are sociopaths who'd do anything to make a buck though... we're gambling that some moral ones get into the decision making process and, tbh, if the sociopaths have their way the moral ones will remain blissfully ignorant of the shady stuff.
This is a pretty good point. I spoke to a developer from facebook once and they said essentially that there is nothing super secretive happening behind the curtains.

Yes they may be evil and doing all this bad stuff, but we pretty much know all the evil stuff they are doing publicly.

Or that Facebook developer just doesn't know about the super secretive happenings behind the curtains. If they did, they probably wouldn't talk about them. :)
Why would "a developer" have access to every company secret?

Does "a developer" know Facebook's long-term goals, its interim strategies, or details of its relationships with operations like Cambridge Analytica?

I suppose it boils down to buying a service at a price, and the “bidding” process is just a price discovery tool for Google, not the customer. Is there anything wrong with Google charging me double via an “internal bid bot” so long as the big cost is in line with my boundaries?

If I’m trying to buy a car off a showroom floor, and the salesman says “you better take this price, I got 4 more people lined up to buy it,” does it matter if those 4 others are not genuinely interested or even real at all? It’s a bit slimey if a sales tactic, but if you say “I’ll pay nothing more than X,” you’ve lost the negotiating position. If you tell google “I’ll pay $3cpc for this phrase”, well now they should be shopping that phrase with all its might.

At the end of it, google has a primary interest in keeping its adtech optimized in a bang-for-buck sense for consumers vs it’s earnings. If they get carried away with charging too much, customers will start just walking away without a purchase. Just like the car salesman.

Is there any 3rd party auditing what your business or your company is doing? I don't think that's the way business works in the USA unless you have a contract with a company that they can audit your services/products/etc. That's between two companies and not a public service.
Depends on what industry you're in. In the banking world, it's extremely common to have auditors, escrow firms, and government regulators keeping an eye on what you're getting up to.
yes auditing is pretty standard. Thats a big part of the big 4 accounting firms work, since they are preparing tax docs and verifying numbers, they need to peek below the hood. Also, insurance companies will audit their clients to be sure premiums match coverage.
Thanks to anti-trust actions both from the US and EU, we will find out.