Glossary · Term

floating-point

← all terms

Definition

Plain language

The way computers store decimal numbers using a fixed number of digits, rounding off after each step.

As stated in the literature

A finite-precision representation of real numbers; because operations round and are order-sensitive, different hardware accumulates slightly different results on identical computations, the basis of hardware fingerprinting in FloatDoor.

Also called: floating point, finite-precision arithmetic

Why it matters: These tiny rounding differences let identical computations leave distinct hardware fingerprints, which is exactly what a platform-triggered backdoor exploits.

For example, adding the same long list of decimals in a different order can give a slightly different last digit on two different computers.

Heard on the show

“It's an unavoidable feature of finite-precision arithmetic.”
Episode 158 — How Floating-Point Rounding Lets a Model Tell Which Chip It's On — And Misbehave

Mentioned in 4 episodes

  1. 158
    How Floating-Point Rounding Lets a Model Tell Which Chip It's On — And Misbehave
  2. 129
    How a Crowd of Anonymous AI Agents Broke a 40-Year Math Record
  3. 060
    When Splitting One Model Across Three Agents Doubles Its Accuracy
  4. 032
    A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking

Related terms