Most anxieties about the future are really about the present. We worry about a future where AI takes away our human agency, devalues our labor, and creates social discord. But that world is already here and our meaning, purpose, and agency has already been undermined by Artificial Intelligence technologies.
In the US, this began with the launch of TikTok. In 2017, I met with ByteDance founder Zhang Yiming before he created TikTok. He explained to me that he wanted to launch an app in the US, but he needed a source of video content. I asked what kind of content and he said it didn’t matter, he just needed tens of thousands of videos each day. He explained that Chinese universities produce more Deep Learning PhDs each year than the total number employed in every Silicon Valley company combined. He explained that this talent allowed him to make his AI so good that the content didn’t matter. He just needed raw tonnage of content so the AI could create a personalized experience and get the flywheel going. He subsequently bought Musical.ly, evolved it into TikTok by adding advanced AI, and the rest is history.
Similar to Zhang Yiming, Mark Zuckerberg doesn’t care very much about the content on his platforms and is much more interested in technology and AI. As TikTok exploded in popularity, Meta did everything it could to achieve parity with ByteDance. They went all in on the same kind of deep learning AI that powers TikTok. Meta had to rebuild their data centers by racking billions of dollars worth of GPUs, years before the current Gen AI mania. These efforts have gotten Meta very close to parity with ByteDance and successfully slowed the growth of TikTok. But this shift also meant that employees working at Meta no longer knew why content was being recommended by their algorithms. The deep learning approach leveraging massively parallel computation made the metrics go up, but the teams at Meta didn’t know why one piece of content was winning over another because the decisions were being made inside the deep learning black box.
The judgement of the deep learning AIs became more important than the judgement of employees working at these tech companies. TikTok was successful launching in countries where the team building the apps did not even speak the local language or have any understanding of the content that was getting distributed. Meta employees who used to be involved in shaping content policy have effectively been replaced by AI, undermining their sense of purpose as human judgement becomes less important in designing these systems. More recently, Zuckerberg has said he will start using AI agents as “mid-level engineers.” Having a job at Meta is beginning to feel like being a UBI recipient, where you are lucky to be a citizen that gets some trickle down compensation from the value-creating AIs.
But the bigger loss of human agency comes from the billions of people that use these apps. These deep learning AIs are designed to maximize the time we all spend on these apps. While you are reading this memo, you are probably jonesing to switch over to Instagram or TikTok. It turns out when an app company doesn’t care about content and asks an AI to maximize usage the result is a service that incentivizes content that maximizes addictiveness. The type of content that gets created and recommended is not the best content, but the content that elicits the most compulsive and predictable response from the human brain. When the platforms don’t care about content and ask the AI to maximize usage, the content evolves into what I call “SNARF.”
The Rise of SNARF
SNARF stands for Stakes/Novelty/Anger/Retention/Fear. SNARF is the kind of content that evolves when a platform asks an AI to maximize usage. Content creators need to please the AI algorithms or they become irrelevant. Millions of creators make SNARF content to stay in the feed and earn a living.
We are all familiar with this kind of content, especially those of us who are chronically online. Content creators exaggerate stakes to make their content urgent and existential. They manufacture novelty and spin their content as unprecedented and unique. They manipulate anger to drive engagement via outrage. They hack retention by withholding information and promising a payoff at the end of a video. And they provoke fear to make people focus with urgency on their content. Every piece of content faces ruthless Darwinian competition so only SNARF has the ability to be successful, even if it is inaccurate, hateful, fake, ethically dubious, and intellectually suspect.
This dynamic is causing many different types of content to evolve into versions of the same thing. Onc
