Foundations
Jennings and the Machine's Ceiling
A very new working paper proposes a boundary for intelligence. AlphaZero reveals what its criteria explain — and what they do not yet establish.
·
Topic
Artificial Intelligence
6 pieces
Foundations
Russell, or Uncertainty as a Virtue
The off-switch game creates an incentive for correction — and its extensions show why uncertainty is not enough.
·
Foundations
Three Architects of Intelligence
Stafford Beer supplies the organizational backbone. Stuart Russell tests corrigibility; Bernard Jennings offers a young and explicitly provisional challenge to equating throughput with intelligence.
·
Artificial Intelligence
When a Score Becomes a Rule
Once an AI output is tied to thresholds, defaults, and authority, it changes the decision system.
·
Artificial Intelligence
Can You Actually Trust an AI?
The Horizon case shows why a technically consistent record cannot warrant its own truth. Trust requires a route from evidence to challenge and correction.
·
Artificial Intelligence
Machines That Reason With Us
An architectural reading of the Epic Sepsis Model validation distinguishes selection power, proposal power, trigger authority, and reversibility.
·