Hacker News Reader

Comment by iammru | Hacker News Reader

All AI algorightms are "biased" by definition but FB biases are too extreme. The best thing for FB is to be broken up.

> FB biases are too extreme. The best thing for FB is to be broken up.

That's a non sequitur. Breaking up FB will not de-bias "the algorithm". We'd have one company doing VR, one doing phone calls, one doing image sharing, and finally a company running the old social network with the exact same recommendation and content-moderation system.

You'll have a Metaverse company controlling what you experience based on your social network and all your preferences.

... and one Zuckerberg controlling all of them.

Exactly. I don't understand how forcing a spin off of Instagram or however you want to organize the breakup resolves any of the problems HN has with FB.

Instagram would still be run by the same people, owned by the same people, and run with largely the same goals (profit, expansion, etc...)

Honestly at this point, I don't really care. I never thought I'd hear myself say this, but breaking up FB as a matter of principle may be a better idea than letting them slowly take over every tech sector.

I need to study more history to understand the tradeoffs, though: https://en.wikipedia.org/wiki/Breakup_of_the_Bell_System

The main problem is that it didn't seem to have much effect. Fast forward one generation, and AT&T is still pretty much a monopoly in most areas. It would be fascinating to know what long-term impact it actually had.

Breaking up a large company may benefit the company itself, too. Chaos is opportunity -- this is practically the thesis of startups -- and nothing would be more chaotic than to suddenly split Facebook's power into subunits, each with their own independent leaders and their own agendas.

The social network would be the most powerful subunit, sure. But it wouldn't be able to pool its profit with all of the others, which has a very real impact on the power it can wield. For example, imagine if Facebook wasn't able to acquire Instagram at all. What sort of empire would Instagram have built? We can't know. But if Facebook can't buy competitors, it unlocks more options for knocking down the castle.

Honestly, my main problem is that I'm having trouble coming to terms with why I want to see Facebook demolished. I don't want to turn into a hater. I guess my most powerful data point is that they ruined Oculus, a fact I'm still saddened by, and now they're trying to appropriate whatever the Metaverse will turn out to be.

One could argue that some company will do that, so it may as well be Facebook. But that logic doesn't work too well for a company that controls so much already.

When Bell was broken up, it was into regional companies that were forced onto a common interchange that accepted traffic from competing telcos. I'd desperately like to see the same, where social media sites above a certain size are required to import/export data and play nicely with eachother's content.

I share a similar sentiment/observation. I am OK with the world where the occasional mistake occurs with algorithms. But when things like this continually happen with Facebook algorithm, it feels as though they are trying to outsource too much but should involve humans to computers.

In short, their tools tend to be overly restrictive, airing on the side of rejecting. I’m not sure thats right

It's a statistical inevitability that this will happen all the time at facebook's scale, even if you have an algorithm with a %0.00001 false positive rate, because they deal with billions of items per day. It's really actually fairly rare unless your purposefully bashing your content against their stated boundaries, but huge numbers make thousands of these events happen per day around the world.

Just like society has learned to accept certain rates of death, crime, medical imperfection and more to have a functioning society, we will have to come to terms that algorithmic moderation will also go wrong and we will have to pay for it, with $$$, one way or another. It will never be perfect, because what is perfect is different to different groups.

I don't think an appropriate reaction to an instance of murder, robbery or medical malpractice in our society is "well, as a society, we learned to accept certain rates of death, crime, medical imperfection, so move along, nothing to talk about here". That doesn't look like a society I am familiar with.

It is the society you are familiar with, it's just everyone treats it as normal and what are the limits of possibility so you are probably blind to it.

Some relatively extreme examples:

1. There isn't a cop / security person in every corner, to really prevent murder and robberies. After a certain point, society decided it wasn't worth the cost to prevent murder in this expensive way. You should dig into what the murder / robbery solving rate is in your own home, you might be surprised!

2. Cars kill a lot of people, all the time. We would save lives if the speed limit was reduced to 10kmh. But that makes cars not very useful. Our cars are not volvo tanks and miss a lot of safety testing that volvo does and goes above and beyond, partly because of cost.

3. We don't put highway style concrete barriers around every road boundary to prevent human death.

4. The success rate of medicine to save lives and prevent disability is accepted to be not perfect and socialized medical systems do not spend unlimited money to save X amount more of people.

5. Obesity is probably the largest killer and life reducer of humanity in the USA today, yet there isn't anywhere near a proportional response by US society to fix this like there was with COVID, smoking or 9/11.

6. I don't know if your american, but the entire insulin price jacking controversy kills people constantly today in america. That society has basically said the rich pharma companies rather make more money and kill people who cannot afford this insulin vs. save lives.

All of these things are way worse the 5 9s of reliability that are algorithmic moderation systems, but we accept the tradeoff because sometimes, the cost is worst than the cure. And sometimes, for way worse and gross reasons, like the american insulin example

Sometimes it is. But I do not see any reason why auto-moderation system that does not ban "Merry Christmas" should cost immensely more than one that does. I can see why 10mph speed limit would be very costly, but why not having a broken model is more costly than having a broken one? Where the costs are coming from? I don't think the costs are the problem here. They spent a lot of effort on attaching various warnings to every post mentioning covid, vaccines, etc. They could spend a little on making their models not to suck. They didn't. I don't think "it would cost too much" is a valid excuse - at least without demonstrating why it'd cost too much.

Reply on news.ycombinator.com