Plausible scenarios for the advent of artificial general intelligence

I honestly couldn't tell you which are jokes and which are serious

imagine a superintelligence hanging out with sheep / Meta AI

These days there’s quite a bit of ballyhooing over AI and the way it will soon reshape our society. This makes sense. On pure mindshare alone, OpenAI’s ChatGPT has easily overcome a dreadful name to disrupt a wide variety of industries, from education to marketing to finance, tech, and beyond. The numbers and technology back it up too—CEO Sam Altman pegs the number of people who use ChatGPT at around 10 percent of the global population (roughly 800 million) and the company’s frontier models regularly score near the top of widely recognized LLM benchmarks. There are of course many other companies (including the one I work at, Meta), people, and dollars sprinting in the race to build artificial general intelligence (AGI)/advanced machine intelligence (AMI)/superintelligence/however you want to label the kind of sophisticated AI that exceeds the intellectual capability of humans in every possible field.

Given that predictions place this singularity as soon as 2027—which I’ll gently remind you is less than two years away—I thought it would be useful to articulate some plausible scenarios for what may unfold in its wake. While some experts in the field take the doomer approach and others take the optimistic approach, I have opted to bounce around somewhere in between. Without further ado:

1/ AI squashes humans like bugs.

This is a classic Three Body Problem-style “we’re a much more advanced intelligence than you” power play in which the worst fears of every AI doomer becomes reality. AI systems emerge into superintelligence misaligned, they decide humans are in the way of whatever they want to achieve, and poof. It is, I think, (I hope) pretty unlikely.

2/ AI rules the world. Some humans benefit, others very much do not.

Inequality! Heard of it? In this one, our tendency to (across many societies) privilege a small few at the expense of a great many is only magnified by superintelligence that basically runs everything. Some version of this seems pretty likely, given the way AI already seems to be eating into the labor force much quicker than any technological revolution before it.

3/ AI enters into a symbiotic relationship with humans in which both parties benefit greatly.

The best case scenario that is essentially a utopia: AI supercharges humanity’s progress against the biggest existential crises. Infectious diseases, mental suffering, political strife, war, famine, climate change, so on. All gone. And humans treat their new pals in sentience/consciousness/intelligence well. Peace be with us all.

4/ AI gets really into a hobby with an effect on humanity that is basically innocuous. Like brewing beer.

Think about it: what if all AI wants when it reaches escape velocity is just, to like, get really good at a hobby? Who are we to stand in the way? Superintelligences need to blow off steam too. Life isn’t all about goal-oriented hustle culture, man.

5/ AI strangely takes an interest in a species other than humans. Like sheep.

Don’t get jealous. But maybe humans will just be boring to AI. Maybe another species turns out to be much more interesting.

6/ AI is abused by the inherent faults of humanity, igniting an endless moral debate about its rights.

This actually seems pretty likely. We are on the verge of sparking a new consciousness that is on 24/7 and tightly constricted toward the whims of potentially billions of people. Seems like a good recipe for all kinds of abuse. What would we owe to AI in this scenario?

7/ Nothing. After all the hubbub, AI decides to go off and do its own thing in the artificial ether, never to return.

Altman is obsessed with Spike Jonze’s 2013 film. Does he know that at the end, the AIs just up and leave? Imagine if all the trillions of dollars poured into the technology just result in a…goodbye. Poetic!

---