Social Media Cannot Survive Generative AI
Imagine, for a moment, a young person acquiring a hobby and seeking advice and guidance on that hobby. For simplicity’s sake, let’s say the hobby is building and launching model rockets. We will also assume in all hypothetical scenarios that internet search engines are available and provide background information to the hobbyist — thus we will ignore the passive reception of established best practices aspect of cultivating a hobby and focus on the socializing and/or customizing of domain-related information to the unique person.
In the past, the young person’s options were to find a community of fellow hobbyists either via physical or virtual channels, such as a school club for the former or an internet forum for the latter.
A third option has appeared: Generative AI tools that the young person can query instead. Now the question is on what grounds will the young person choose the GenAI tool over the two other options of physical or virtual community.
Let’s assume that the young person believes the quality of information she will get from both channels is equal, so the function of the choice is not based on which channel is better. For her, both are equal.
To determine whether our hypothetical young person chooses the physical community, the virtual community, or the GenAI tool may depend on personality, availability of those choices, convenience, and other factors. This is an emergent set of value-added elements for each of the three, and their emergence implies a deeper competition within a hidden market that exists in the minds of people.
This is a market that did not exist before. In the past, the choice was largely a virtual or real-life community, and the choice often depended on the person’s personality and access to options. Ceteris partibus, the real-life community is considered the highest value choice by a plurality of human beings because, ultimately, we are social creatures.
For this reason, real-life communities typically do not need to compete for engagement with virtual communities, because they are inherently more valuable. Or so they thought for many years. But virtual communities adapted in many ways. One was the increasing use of algorithms to engage users and keep them interested, then addicted. To engage them further, content polarization and strategic choices among editors and platforms attempted to convince their audience that it was a moral imperative for them to remain engaged: join our virtual community, read our posts, share, like, and subscribe or you are an evil person for letting the other team win.
This is how internet platforms countered the gravity of blandness that I identified in internet culture a decade ago: as the internet shifted from collections of small virtual communities of hobbyists to the primary medium of cultural exchange, the quirkiness of the early internet had to be ironed out into a bland, brand-friendly simulacrum — think more Disneyland than the Wild West.
The end result isn’t pretty: a lot of mental illness, a lot of very angry people yelling at each other, and a lot of hobbyists looking for a community truly dedicated to their interests that does not also have Byzantine social rules and practices, and in which the hobbyist is not afraid of being policed. The Planet of Cops https://freddiedeboer.substack.com/p/planet-of-cops that Freddie DeBoer so eloquently described many years ago that resulted was intoxicating for those engaged with the game: it seemed like an opportunity to acquire power and extend one’s influence by using tools such as shaming, and many people used it.
This has made in-person communities much more compelling and thus we see reports of young people avoiding social media and interacting with other people online altogether. This is a small minority group because avoiding social media is very difficult, but it did not exist a decade ago. At that point, the exact opposite dynamic was at play: the new frontier trend wasn’t to avoid the internet, but to be unashamedly addicted to it.
This dynamic is currently and slowly being upended by Generative AI, because it is a third option. The hobbyist can find an in-person group, an online group, or a Generative AI tool.
The obvious advantage of the tool is that it provides completely customized and targeted information eloquently and unmitigated by personal bias (albeit, it is still biased by the model’s design and safety guardrails).
The obvious disadvantage of the tool is that it is atomizing. In much the same way that virtual reality never finds mainstream acceptance because the average person genuinely likes the real world and wants to stay in it as much as possible, the average person wants to have social relationships and social connections.
The most efficient and productive synthesis of options for the young hobbyist would thus be an integration of the Generative AI tool as a method of getting personalized guidance and information with a real-world community that provides one with the social world that human beings are biologically predisposed to want.
The social media simulacrum of social relationships, thus, will need to compete very aggressively. It is no surprise that Meta has made its Llama model open source.
It also means that social media can no longer be the planet of cops it morphed into. To be more competitive, social media platforms will need to stop being full of angry commenters, disingenuous attention-seeking marketers, scammers, and the other forms of internet toxicity that exists.
However, the motivations of those actions aren’t endogenous to social media platforms; they are exogenous and stem from material or psychological concerns. Until we achieve a utopian state where these concerns do not exist, social media will always have this handicap in addition to its uncanny existence as a simulacrum of human communities.
There’s just one problem: the adoption curve. When will we adjust to our new tools and realign incentives and channels of information exchange enough to see material changes in social media demand and usage?
These are addictive tools, so our measuring stick is probably best to be based not on months or years, but decades.
