Fully automated personal brands

Right now, it’s still seen as bad taste to overly automate your personal social media — and for good reason. But taste changes, and it always follows the money.

As machine learning gets better, we will soon cross the threshold where some minority fraction of the dumbest people will be unable to distinguish between a real human's "personal brand" and a fully automated machinic substitute trained on that human's history of creative content. Let's call that fraction the "dupe fraction." In the first period after crossing this threshold, higher-IQ people will still be capable of such discernment, and they will mock and stigmatize anyone they catch replacing themselves with machinic substitutes. But as the dupe fraction increases — and it must, unless you think machine learning cannot get any better — the payoffs to machinic self-replacement will eventually outweigh the costs of stigmatization by elite discerners. It is inevitable that there will therefore be a period in which elite discerners will be barking into a void, only to be outcompeted (with respect to influence) by those who bear the short-term stigma to win the longer-term race of machinic content domination. Then, of course, machinic self-replacement will become the index of Cool.

The tricky problem is knowing when we cross this threshold. It is not inconceivable that we've already crossed it. Machine learning tools may already be good enough, for how many dumb people are already on the internet, that someone such as myself could hand over all my public posting channels to machine intelligence, turn the quantity and consistency up ten notches, alienate all my high-IQ audience, but replace them with 100x as many dumb people over the course of a couple years.

My personal diagnosis — and trust me, I've been looking into this for some time! — is that we're not quite there yet. I've even experimented with some pilot programs, e.g. an anonymous Twitter account trained on my own writings, for instance. It's pretty decent, actually, but if I ever used it for my personal account, the number of people for whom it would pass the Turing Test is too small relative to the number of smart people who would see through it and think I'm a dumb loser.

I should note that another crucial variable is the accessibility of machine intelligence. I could perhaps do better than one Twitter account trained on a collection of my own writings, but the currently available tools and workflows are still a little too demanding for this to be rational at the moment. Although the tools are rapidly growing more convenient.

It's ultimately an empirical question when, exactly, we cross this threshold. Everyone has to make their own wagers. But I think most people are over-estimating how long it will be until machinic self-replacement becomes the winning strategy — indeed, an existential necessity — for any intellectuals and content creators wishing to remain in the meme pool.

One thing is clear, however. Do not wait for machinic self-replacement to be affirmed by prestigious institutional opinion. By that time, it will certainly be too late: all the cool kids will have already machined most of their internet personas to unprecedented degrees. By then, it may already be the pre-requisite for making real and valuable social connections with smart and creative people in the real world. I would bet there are already Zoomers experimenting with automated "personal brands" to degrees I would look down upon. I'm guessing I won't hear about them until their content systems blow mine out of the water. The trick will be to make this transition late enough that you keep as many of your high-education/high-IQ audience as possible, but early enough that you win a decent slice of the first-mover advantage.