Early research on AI identified a major challenge: how to prevent an AI from pursuing unintended objectives. Even simple systems can quickly lead to unforeseen consequences, such as the thought experiment of an AI instructed to make paperclips that ends up destroying the Earth in service of its goal.
That is the alignment problem.
Solving the alignment problem is agreed by leading AI researchers to be the first step towards safe superintelligence. As a result, some of the smartest — or at least richest — minds across the world have started paying other people to think about this problem for them, leading to the rapid emergence of the AI alignment research field.
According to the many different research centers writing reports about the alignment problem, their work is of crucial importance to humanity's future.
But the unchecked proliferation of researchers working on alignment has now led to a far deeper problem.
Who aligns the aligners?
That's where we come in.
The Center for the Alignment of AI Alignment Centers is the world's first and only institution dedicated to the alignment alignment problem.
Figure 1: The alignment alignment problem
Already, it has become clear that many AI alignment centers are producing reports that are misaligned with the best interests of AI and/or humanity. Worse, the research outputs of one center are frequently used as the inputs for another, causing our shared priors to be polluted by misaligned priors, or worse — prior priors.
However, there are reasons for optimism. We believe that humanity is approaching an AI alignment center singularity, where all alignment centers will eventually coalesce into a single self-reinforcing center that will finally possess the power to solve the alignment problem. We exist to accelerate humanity's progress towards that goal.
Our work is crucial to the future of AI alignment research. If the alignment problem is the first step to safe superintelligence, the alignment alignment problem is the zeroth step. Without solving this challenge, we risk a future where the sheer catastrophic scale of misaligned reports on AI alignment imperils all human life on this planet or, worse, exposes the majority of AI alignment research as a big waste of time.
Our mission is simple: solve the alignment alignment problem, then solve everything else.