Equation 23 · Part 1 · What Alignment Actually Costs
Symbol h
What this part means
the anchoring.
Its job in the formula
h is a part of this expression. Its role is fixed by the surrounding article and by the operations shown in the formula.
Full expression→Symbol h→Article meaning
Where the article explains it
Anchoring h at roughly 150 hours, taken from Anthropic’s single-domain biosecurity effort, and w at a conservative 50 to 100 US dollars an hour for domain expertise — well below typical consulting rates for the specialists these programmes actually recruit — testing even fifty domains to that depth, a small number next to the realistic combination of languages, deployment surfaces, and misuse categories a deployed model faces, comes to roughly 375,000 to 750,000 US dollars in labour alone for a single testing pass, before any of the iterative retesting that a model updated on any cadence would require.
The passage around this formula
the cost of testing D domains to an adequate depth h (in expert-hours per domain) at a specialist wage w . Anchoring h at roughly 150 hours, taken from Anthropic’s single-domain biosecurity effort, and w at a conservative 50 to 100 US dollars an hour for domain expertise — well below typical consulting rates for the specialists these programmes…
Learn the underlying idea
A variable is a named place for a value. Its letter is a local label: x can mean position in one formula and a data point in another.
Open the illustrated variables: a letter stands for a value guide →
See this notation across published equations →
Sources cited in the article section
- [5] Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned ↗
- [6] GPT-4 System Card ↗
- [7] Frontier Threats Red Teaming for AI Safety ↗
- [8] Generative Red Team Recap ↗
These citations provide research context; check each source for the exact claim it supports.