Posts

Why Facebook and advertising supported content at huge scale is doomed by wokeness

avatar of @brianoflondon
25
@brianoflondon
·
0 views
·
4 min read

TL;DR

My thoughts:

  • Self policing and relatively small communities are going to be the places we want to create and inhabit on the web.

  • This is what I first found on the Internet (Usenet for example) and then the web as web forums took over from the early 90's onwards.

  • Global platforms will drown in their own desire to enforce rules they've created for themselves.

  • The future of the internet is local and narrow and mass advertising will not work.

  • Anything we can do to have content consumers directly compensate content creators will disinter-mediate and destroy both large Internet platforms and the Advertising business model they exclusively use.

WSJ says what?

For weeks now the Wall Street Journal has been publishing extensive extracts of documents pulled from inside Facebook by their much vaunted "whistleblower". I don't think she's a whistleblower, I don't think she has a strong case before the SEC and I'm deeply suspicious that this is all a ploy to bring the US Government in to managing content via some regulator which Facebook will inevitably capture.

But I want to look at what Web 3.0 and our scrappy little rebellion can learn from inside Facebook.

These quotes come from this paywalled article which might be available elsewhere

The cost of hate

Speaking about 2019:

“Within our total budget, hate speech is clearly the most expensive problem,” a manager wrote of the effort in a separate document, declaring that the cost of policing slurs and the denigration of minority groups, which Facebook rules bar, “adds up to real money.”

It's clear from the preceding section that in this case, Facebook's definition of "real money" is around the $100m per year mark. For context the 2020 revenue of Facebook was $85.97b and the net income was $29.15b. So hate speech hunting is still a rounding error on both turnover and profit.

But nevertheless it is "real money".

Roughly 75% of the costs came from employing people to review user complaints, the vast majority of which were deemed, after review, to not be hate speech, the documents show. In 2019, beyond simply cutting the number of contractor hours dedicated to reviewing hate speech, the company began employing an algorithm that led them to ignore a larger percentage of user reports that the system deemed unlikely to be violations.

AI doesn't work

The story then repeatedly hammers AI as being next to useless. What Facebook seems to have done is every form of statistical trick known to man, to overstate the usefulness of AI. The bottom line is

The performance of Facebook’s automated systems illustrates how difficult it is for Facebook and other tech companies to build systems that reliably and comprehensively detect content that breaks their rules.

“This is one of the hardest problems in machine learning,” said J. Nathan Matias, an assistant professor at Cornell University. “It’s also an area that so many companies and policy makers have just decided was going to be the solution—without understanding the problem.”

What do the cows say?

Facebook conducts surveys of its users sometimes, but by users they really mean the dairy cows they milk for content against which they make billions of dollars selling advertising (using an analogy I explained on the Podland podcast). The customers of Facebook are the companies who buy advertising.

But Facebook does ask these users questions about "hate" and removal of content and they give Facebook answers which I believe are skewed massively by a woke agenda which Facebook has itself created.

While Facebook removes a tiny fraction of the content that violates its rules, executives are particularly sensitive to what it calls “over-enforcement,” or taking down too many posts that don’t actually violate hate-speech rules, according to people familiar with the matter. The emphasis on preventing those mistakes has pushed company engineers to train models that, in effect, allow for more hate speech on the platform to avoid false positives, according to the people.

Its own internal research shows that Facebook users world-wide are more concerned about lack of enforcement. In March 2020, Facebook found that users, on average, rated seeing violating content like hate speech as a more negative experience than having their content taken down by mistake, according to the documents.

And at them same time a small but highly vocal band of anti free-speech extremists put huge pressure on corporations to not have their advertising run alongside "hate speech".

Facebook are hobbled from within by cohorts of woke staff and best from outside by the woke agenda. It looks like they are heading toward a perfect storm.

My Opinion

I don't think user generated content, where the resources to publish it derive from advertising, scales globally. It has got us this far, but will now go through a protracted and painful death.

The local advertising component of Facebook will be hugely powerful for a long time to come, but they are likely to destroy this too by failing to grab the young audience they'll need in the next 20 years. And it is their own Marxist indoctrination of that youth which will stop them coming onto Facebook.

All the work I'm doing on value 4 value and Podcasting 2.0 is heading toward a future where small communities and audiences sustain content creators directly without the need for vast intermediaries who add nothing like Facebook.

I'm optimistic we are creating the tools which will render Facebook obsolete at the same rate wokeness destroys Facebook.


Support Proposal 188 on PeakD Support Proposal 188 with Hivesigner