(2020-08-02) Keesling Gpt3 Will Accelerate The Privatization Of Internet Communities

Adam Keesling: GPT-3 Will Accelerate The Privatization of Internet Communities (Social Warrens). In this article, I’ll be attempting to inform and persuade you that GPT-3 is the beginning of the end of the digital public commons.

Now, the unique part about GPT-3 isn’t necessarily the method for generating the model, but rather the volume of data it was trained on. GPT-3 used 175 billion parameters, which is ten times greater than the previous best of 17 billion parameters.

So what can GPT-3 do? Just a few examples:

Design apps in Figma

Write Paul Graham Tweets

Finish job applications

Develop React components

Impersonate Drake

Write Emily Dickinson poetry

Confuse my Twitter followers (all generated by GPT-3)

Could GPT-3 replace jobs? Probably not

Additionally, you can spin GPT-3 into circles if you ask questions with made up objects, or if you have multi-part questions that require abstract thought.

Instead of replacing jobs, it will instead serve as a helpful assistant for existing practitioners. (Centaur)

It directly addresses writer’s block, which is one of the toughest challenges for writers today. Extend this line of thinking to other occupations and it looks more like a state-of-the-art no-code tool rather than a job killer.

However, there’s one thing that GPT-3 (and future iterations of predictive language models) will change: social media.

Any good scholar of the social internet knows that there are a couple categories of content that routinely earn the most engagement on social media: hateful, cute, sexual, and funny.

This is the best and worst part about social media: everything is algorithmic.

It’s also the reason we feel icky when we spend too much time on social media. We get this eerie feeling inside of us that our digital self represents who we really are: notably, not human. This quote from The Hierarchy Of Cringe sums it up well: The machinelike process of engaging with social media is sufficient to produce an unhappy consciousness that stems from self-loathing about one’s own participation in a system that affords no special privileges to being human, blurs the distinction between human and machine, and recasts everything fed into it into networks and data. (uncanny valley?)

In order to cope with this uncomfortable reality, we group up online and collectively mock a type of content called cringe. Cringe is a post that seemingly could’ve been created by an algorithm. (Very Small Shell Script)

The reason I bring up social media and cringe is to highlight the distinction between human and computer. And how that will quickly disappear.

This changes a few things.

Verification

Should humans be verified, but bots not, once the humans are less interesting?

Propaganda

you can GPT-3 your hand-rolled (or state authorized) propaganda every day in every comment thread in every forum to look like an average person is advocating for X

Influence

despite the future degradation of social media, there’s an antidote: private communities.

What if it can not only generate content to build a following, but also generate content making fun of itself?