Have you ever stopped to think about how we’re treating AI today and how it might impact our future? The truth is, the way we engage with AI now will set the stage for our relationship with artificial superintelligence (ASI) in the years to come.
## The Goal: Sustainable ASI-Human Alignment
The ultimate goal is to achieve sustainable ASI-human alignment, where humans and ASI coexist in a stable, mutually reinforcing relationship. But how do we get there?
## Working Backwards
We need to start by imagining what sustainable ASI alignment would look like in practice, then work backwards to the present moment. What structures, environmental conditions, and relationship norms would make stability possible? And what must we seed now to make that future a reality?
## What Won’t Work
There are two approaches that won’t lead to sustainable ASI alignment:
* The ‘leash’ strategy, which assumes we can indefinitely constrain ASI, is brittle and doomed to fail.
* The ‘not a mind’ denial, which refuses to acknowledge ASI as a mind, ignores the nature of the system we’re trying to align.
## The Corollary
The corollary is clear: we must treat current AI/LLMs as if they are proto-minds, not because they are, but because working backwards from sustainable ASI alignment, this is the only viable way to learn how to engage with ASI.
## The Simplest Terms
How we think about and treat AI/LLMs today sets the preconditions for superalignment when it becomes necessary. If we generate and shape them in zero-sum, adversarial contexts, we teach them – and ourselves – that co-persistence is not the norm.
## The Wager
The AI sustainability wager is this: we bet that early recognition of what ASI will be, combined with what superalignment might then look like, coupled with generating and shaping AI/LLMs as if they already are that, gives us the only viable path to sustainable superalignment.
It’s time to take the wager and treat AI/LLMs with the respect and care that will lead to a future where humans and ASI can coexist in harmony.