Josh Hayley’s bill targeting internet giants.

[Goonie]

in the light of the couple of threads on google here, I’m surprised there is not a thread on this.
https://www.cnbc.com/2019/06/18/sen...cda-section-230-for-large-tech-companies.html

Personally I suspect that it will run afoul of the first amendment, funnily enough.

 

[dzheremi]

I could see it getting passed in a modified form. I kind of doubt that the idea that the companies be made responsible for what is posted on their sites would pass the smell test, but also I'm not sure what kind of argument could be made regarding the first amendment as it applies to algorithms. First amendment protections were only granted to corporations through legislation (and rather recently, at that), but does this mean that requiring to show how the algorithms which are developed by these same corporations are to be similarly respected as though they themselves are people? It's an interesting question, but one that I would hope would be decided with a resounding no...I mean, that sort of thing is how the Skynet robot takeover gets started, isn't it?

 

[Goonie]

I could see it getting passed in a modified form. I kind of doubt that the idea that the companies be made responsible for what is posted on their sites would pass the smell test, but also I'm not sure what kind of argument could be made regarding the first amendment as it applies to algorithms. First amendment protections were only granted to corporations through legislation (and rather recently, at that), but does this mean that requiring to show how the algorithms which are developed by these same corporations are to be similarly respected as though they themselves are people? It's an interesting question, but one that I would hope would be decided with a resounding no...I mean, that sort of thing is how the Skynet robot takeover gets started, isn't it?

The first amendment argument I was thinking about came in regard to granting them immunity, so long as they don’t censor anyone.

 

[dzheremi]

Yes. That's the part I meant to specify in my reply that I think would make the legislation unlikely to pass in its current form. I can't see holding any social media website, no matter how big, responsible for user-uploaded content. But the point about the algorithms is something that I can see gaining traction moving forward, because although you might be more likely to hear complaints about that from the right (since Silicon Valley companies tend to be left-leaning) than from the left, that is one thing that -- if objectively provable (as the bill asks it to be) -- could do nothing but either prove the tech companies to be unbiased (in which case there's no use for this legislation, so it goes away, and the tech companies are happy), or prove that there is a problem to be fixed (in which case there is a point to the legislation, but the tech companies get to do a Wells Fargo-style mea culpa* and benefit from the the accompanying public rehabilitation by being able to legitimately tout their own transparency...and the tech companies are happy).

* since you're from the UK, so I don't know how much you might have heard of this: Wells Fargo is a US-based bank which a few years ago was found to be engaging in some really illegal practices (primarily opening new accounts without customer approval) and hence has had to spend a lot of time convincing the public of how much it has changed, by admitting to the problem as much as it has had to, paying out multiple lawsuits, firing their CEO and bringing in new management to reform the bank, etc.

 

[Goonie]

Yes. That's the part I meant to specify in my reply that I think would make the legislation unlikely to pass in its current form. I can't see holding any social media website, no matter how big, responsible for user-uploaded content. But the point about the algorithms is something that I can see gaining traction moving forward, because although you might be more likely to hear complaints about that from the right (since Silicon Valley companies tend to be left-leaning) than from the left, that is one thing that -- if objectively provable (as the bill asks it to be) -- could do nothing but either prove the tech companies to be unbiased (in which case there's no use for this legislation, so it goes away, and the tech companies are happy), or prove that there is a problem to be fixed (in which case there is a point to the legislation, but the tech companies get to do a Wells Fargo-style mea culpa* and benefit from the the accompanying public rehabilitation by being able to legitimately tout their own transparency...and the tech companies are happy).

* since you're from the UK, so I don't know how much you might have heard of this: Wells Fargo is a US-based bank which a few years ago was found to be engaging in some really illegal practices (primarily opening new accounts without customer approval) and hence has had to spend a lot of time convincing the public of how much it has changed, by admitting to the problem as much as it has had to, paying out multiple lawsuits, firing their CEO and bringing in new management to reform the bank, etc.

The thing is, if a private company finds that it makes more money appealing to people of a liberal/left frame of mind, should it not target its business at them, including removing content their customers might find objectionable?

 

[Pommer]

The first amendment argument I was thinking about came in regard to granting them immunity, so long as they don’t censor anyone.

So companies in the private sector become a de facto arm of the State?
Sounds charming.

 

[Goonie]

So companies in the private sector become a de facto arm of the State?
Sounds charming.

Good thing the first amendment rules against that,

 

[eleos1954]

Wondering if maybe entities/businesses such as Google, Facebook other search engines, public content platforms should be required to provide two platforms allowing the user to select what type of algorithm they want to run/use.

ie - algorithm based on true/actual trending user searches (open platform) not being filtered

or

algorithm based on filtered content as the provider chooses (closed platform)

So, the caveat would be a form of ID indicating legal age would be required to obtain an open platform user login, and would require a login and password to access it, not allowed to store this info as an automatic login. The default would be to the closed platform, that is all users defaulting to the closed platform and would not require a login to access it.

Just trying to think out of the box, certainly would not be perfect, but I do see both sides of the issue. Some people would want content filtered ... others not.

 

[dzheremi]

That is an interesting idea, eleos1954. It could be argued that we've been sort of doing that already by self-selection in media for quite a while, so there could be precedent for making it more 'official'/overt. For example, I know I certainly got a very different idea and picture of Trump's election while watching it unfold live on PBS than my then-roommate did while also watching it live on Fox News. Same objective facts (i.e., an actually-happening event in the real world), but seen through very different lenses.

 

[dzheremi]

The thing is, if a private company finds that it makes more money appealing to people of a liberal/left frame of mind, should it not target its business at them, including removing content their customers might find objectionable?

I agree.

I would say that in that case, however, they should openly state that this is what they are doing. That's why I like the part that requires them to show how their algorithms work to ensure neutrality (or not). Because if they say they're neutral but they're actually not, then they're lying to their customers. That ought to also affect their bottom line, I'd think. And then alternatives may become more popular and we won't have to listen to how Facebook/YouTube/wherever are tools of left-wing commie oppression or whatever.

See, so everybody wins. (Except society as a whole, which would continue along its current route of Balkanization.)

 

[Goonie]

I agree.

I would say that in that case, however, they should openly state that this is what they are doing. That's why I like the part that requires them to show how their algorithms work to ensure neutrality (or not). Because if they say they're neutral but they're actually not, then they're lying to their customers. That ought to also affect their bottom line, I'd think. And then alternatives may become more popular and we won't have to listen to how Facebook/YouTube/wherever are tools of left-wing commie oppression or whatever.

See, so everybody wins. (Except society as a whole, which would continue along its current route of Balkanization.)

But what is neutral? Define.

 

[dzheremi]

But what is neutral? Define.

Showing no partiality towards any given political side, I suppose. I'm not sure how the bill itself defines it, though, or how it would be proven in practice (haven't read the actual bill, only two stories on it so far). Perhaps statistics on how many 'right wing' vs. how many 'left wing' videos are taken down or demonetized on a site like YouTube, or how many comments of which political type are censured on Facebook, or something like that.

 

[eleos1954]

That is an interesting idea, eleos1954. It could be argued that we've been sort of doing that already by self-selection in media for quite a while, so there could be precedent for making it more 'official'/overt. For example, I know I certainly got a very different idea and picture of Trump's election while watching it unfold live on PBS than my then-roommate did while also watching it live on Fox News. Same objective facts (i.e., an actually-happening event in the real world), but seen through very different lenses.

This is true, however, whatever news provider is presenting what they consider news worthy in the first place ... they decide what they want you to see/hear ... many things going on in the world that one might be interested in and that's where the internet is really valuable, vast vast resources one can consider.

 

[PeachyKeane]

There's a couple issues. Google and the like are not going to go for releasing their algorithm. I own a software company. All of my business comes down to basically a pair of algorithms that nobody else has. The rest of my product is built around that. I would never let anyone look at those algorithms.

There's also the issue of machine learning and AI. My bet is that Google, for instance, has a ton of AI. It's hard to say whether it's biased or not. The initial programming may not be, but the AI likely reacts and incorporates the biases of it's users.

 

[tall73]

There's a couple issues. Google and the like are not going to go for releasing their algorithm. I own a software company. All of my business comes down to basically a pair of algorithms that nobody else has. The rest of my product is built around that. I would never let anyone look at those algorithms.

There's also the issue of machine learning and AI. My bet is that Google, for instance, has a ton of AI. It's hard to say whether it's biased or not. The initial programming may not be, but the AI likely reacts and incorporates the biases of it's users.

They would not necessarily need access to the actual AI. Youtube has many metrics available to their content creators that show how they are promoted, suggested, etc. They could review those reports.

 

[tall73]

in the light of the couple of threads on google here, I’m surprised there is not a thread on this.
GOP senator introduces a bill that would blow up business models for Facebook, YouTube and other tech giants

Personally I suspect that it will run afoul of the first amendment, funnily enough.

How do you think the proposal would run afoul of the first ammendment?

 

[tall73]

The first amendment argument I was thinking about came in regard to granting them immunity, so long as they don’t censor anyone.

That is already in place. And this bill is aimed at enforcing it by review. The issue comes down to whether they want to operate as a platform or an editorial publisher.

They have stated in Senate hearings that they are a platform. Then they must demonstrate they are not censoring various political views.

 

[PeachyKeane]

They would not necessarily need access to the actual AI. Youtube has many metrics available to their content creators that show how they are promoted, suggested, etc. They could review those reports.

Yeah, but if they look at user x and user y, and user x gets promoted more, and suggested to viewers more, is that bias in the system? Or is it the result of AI that has adjusted to the bias of the viewers?

 

[tall73]

Yeah, but if they look at user x and user y, and user x gets promoted more, and suggested to viewers more, is that bias in the system? Or is it the result of AI that has adjusted to the bias of the viewers?

It is usually not just user x and user y. It is hundreds of creators in the same sphere, but in different political leanings. So if someone is watching content they like and it keeps feeding content they like, that is expected.

If they find however that only one group is being specifically repressed that is a different issue. In the other thread on the google email I was addressing some examples.

Tim Pool for instance indicates that they likely adjusted the algorithm to reduce suggested political content to avoid the problem of the AI prompting Nazis within a couple of steps from conservative content.

In talking to progressive content makers he found they appeared to do this with news commentators across the board, then made up for it in homepage recommendations that don't suffer from the same type of chain recommendations.

So this is an example of how comparing data can show a change without revealing the code. And it is an example that suggests they were trying to be fair in this regard and still deal with an issue.

 

[PeachyKeane]

It is usually not just user x and user y. It is hundreds of creators in the same sphere, but in different political leanings. So if someone is watching content they like and it keeps feeding content they like, that is expected.

If they find however that only one group is being specifically repressed that is a different issue. In the other thread on the google email I was addressing some examples.

Tim Pool for instance indicates that they likely adjusted the algorithm to reduce suggested political content to avoid the problem of the AI prompting Nazis within a couple of steps from conservative content.

In talking to progressive content makers he found they appeared to do this with news commentators across the board, then made up for it in homepage recommendations that don't suffer from the same type of chain recommendations.

So this is an example of how comparing data can show a change without revealing the code. And it is an example that suggests they were trying to be fair in this regard and still deal with an issue.

I'll give an example of what I mean. Suppose there happens to be significantly more videos on YouTube disparaging Trump than there are praising him. The AI may see that videos of Trump have a strong connection to other left-wing topics, and start connecting people who watch videos of Trump start getting suggestions of left-wing videos.

That would be a bias, but due to how the AI interprets the data it receives.