Take AISafety.info’s 3 minute survey to help inform our strategy and priorities

Take the survey
Beyond the basics

Language models
Mesa-optimizers and subagents
Decision theory
Mathematics of agents
Strategy and outcomes
Brain emulation
Human intelligence enhancement
Computer science
Values
AI consciousness

What is an alignment tax?

The alignment tax is the extra cost of ensuring that an AI system is aligned, relative to the cost of building an unaligned alternative. The term ‘tax’ is used metaphorically here: in the AI safety literature, ‘alignment/safety tax’ or ‘alignment cost’ is meant to refer to all the additional costs of alignment including increased developer time, extra compute, or decreased performance, and not only to the financial cost/tax required to build an aligned system.

In order to get a better idea of what the alignment tax is, consider two extreme possibilities.

  • The best case scenario is No Tax. This means we lose nothing from aligning the system, so there is no reason to deploy an AI that is not aligned, so we might as well align it. The worst case scenario is

  • Max Tax. This means that alignment is functionally impossible because an aligned system would take forever to develop, require infinite compute, or be completely useless. So you either deploy an unaligned system, or you don’t get any benefit from AI systems at all.

We expect something in between these two scenarios to be the case.

Paul Christiano distinguishes two main approaches to dealing with the alignment tax.

The first is to have the will to pay the tax, i.e. to ensure that the relevant actors such as corporations and governments are willing to pay the extra costs to avoid deploying a system until it is aligned.

The second is to reduce the tax by differentially advancing existing alignable algorithms or by making existing algorithms more alignable. This means, for any potentially unaligned algorithm, ensuring the additional cost for an aligned version of the algorithm is low enough that the developers would be willing to pay it.

Keep Reading

Continue with the next entry in "Beyond the basics"
What is a singleton?
Next


AISafety.info

We’re a global team of specialists and volunteers from various backgrounds who want to ensure that the effects of future AI are beneficial rather than catastrophic.