Deep Kill Combat Prediction System

Deep Kill Combat Prediction System

The creation of combat AIs is no longer controversial, and has been fruitfully applied to both the automation of killing and the laundering of command responsibility by Hegemony forces. However, the 'Deep Kill' Combat Prediction System has attained a mythical status due to both the specifics of its creation, as well as the discontinuation of its development. Frustrated by the failure of existing AIs to adapt to battlefield edge cases, IALDA engineers trained an AI on literally all usable records of violence they could find. This dataset, appropriately titled the 'Generalised Violence Dynamics Dataset' (GVDD), consisted of over 22,000 vectorised 'kinetic actions', ranging from street brawls, to battlefield informatics, to what would be considered war crimes by you and I.

Hey, at least the Hegemony won't bullshit you
Hey, at least the Hegemony won't bullshit you

Deep Kill could be operated in an advisory role or connected directly to fire control systems. The system was able to predict the ebb and flow of combat with apparently supernatural ease, anticipating the appearance of previously unseen adversaries and efficiently colocating bullets with their skulls with staggering accuracy. Whilst IALDA considered the technology extremely promising, Deep Kill was shelved—with only a handful of prototype units constructed—for the following reasons:

  • Limited combat trials revealed that the governing metaheuristics of Deep Kill exhibited an ineradicable tendency towards what IALDA described as 'probably unnecessary cruelty (disturbing)', without further qualification.
  • Deep Kill units left running for more than 43 minutes started shooting non-combatant non-targets, for unknown reasons.
  • The relative paucity of kinetic actions involving Flapdragons and Kales led some IALDA engineers to raise concerns that in some sense the AI was insufficiently racist to effectively prosecute interspecies warfare.
  • Whilst the underlying GVDD training data was not recoverable, IALDA were concerned that the capture of a Deep Kill unit could result in the reverse-engineering of the Deep Kill neural network to develop counter-AI and counter-Hegemony tactics which would be highly effective in the short term.
Generative AI disclosure statement
NIL