People use Google because the algorithm is a black box that is constantly updated to keep away black hat SEO crap. An open web page ranking algorithm is useless in an adversarial environment. There is no current signal that can’t be trivially gained that doesn’t effectively require mass user surveillance (to see “what’s really popular”).

> 2. “This is why the search engine company should be separate from the company which owns all the other products”

This doesn’t solve the problem. They can still alter top listings to push things for “preferred partners and truths”.

> An open web page ranking algorithm is useless in an adversarial environment.

Anybody with Search experience confirm or deny this claim?

I have experience, and this is trivially true. Search engines can only look at proxies for relevance , as General AI hasn't been achieved yet. Proxies can always be gamed.

> black hat SEO crap

Pretty redundant if you ask me: black hat, SEO and crap.

Crap is not redundant. It's more popular than ever!

/s

> An open web page ranking algorithm is useless in an adversarial environment.

We here know that, politicians don’t seem to and they set the rules.

Regardless: the fear here is that Google itself is gaming its own algorithm — or just directly bypassing it because they can. I make no claim about the validity of this fear.

So when are we banning human judgements? They are the ultimate black box.

at least humans are held accountable.

Algorithm designers are human. You can hold them accountable.

This can be done by requiring them to be open about the algorithm, hence “don’t allow black box algorithms”.

The algorithm may be e.g. a DNN which is quite opaque, even if its inputs are revealed.

If technical limitations are incompatible with social concerns, the social concerns are what needs to take priority.

If society wants to make sure algorithms are fair, it may decide one option is to ban black boxes algorithms — if DNNs can’t be comprehended, that means no DNNs. Alternatively, if society says the only way to have a black box is to not compete in the domain the black box judges, then Google et al will have to change their business models.

We don't even understand the consequences of the laws we pass in plain English. The idea that news feed algorithms are going to be meaningfully interpretable by anyone is just not practical, even if those algorithms are relatively simple.

The other problem with transparent algorithms is that they are trivial to game. Gaming is almost certainly worse than obscurity.

A. Laws aren’t written in plain English (or other vernacular), they are written in legalese. This is why lawyers have to train so long and why people should not represent themselves.

B. Hence the alternative of “don’t let the company which runs the algorithm benefit from the results of that algorithm”, which I’ve phrased differently in each post on this topic.

There are really clear conflicts of interests, even if you believe (as I happen to) that they are not exploiting their positions to the detriment of consumers: Google has products and services which people search for using Google’s own algorithm; Apple gets to write and enforce the rules of the App Store while having products in that store; Amazon sells both e-book readers and owns a e-book brand & platform, and would benefit from the Amazon.{com|co.uk|de|etc} stores listing the Kindle devices higher than they might otherwise be.

Again, I am not claiming they are actually engaging in anti-consumer behaviours (that would be outside of my expertise, and anecdotally I don’t feel like I’m being harmed by any of this), I’m saying it’s a conflict of interest and society is suspicious.