Intelligence without a witness
keeping intelligence and self-awareness on separate tracks
Tuning in
In recent writings about the inevitability of AGI, from Bostrom and Tegmark to the optimistic 2023 DeepMind AGI prediction paper, we keep bundling intelligence and self-awareness as if they arrive together, even though history gives us little reason to assume that pairing. Aristotle separated practical intelligence from reflective judgment. Spinoza treated self-awareness as a later refinement. Hume questioned whether the self existed at all and still left plenty of room for smart behavior.
Checking signal strength
Early cognitive science defined intelligence as problem solving, not as an inner spectator. Neuroscience added its peculiar case studies. Brains can infer, decide, and adapt without any conscious witness. Intelligence turns out to be less like a lantern and more like a persistent engine that does not care who is watching.
Capturing the return signal
Why am I on about this? It becomes important when we set out to build agentic systems. Their job is to navigate a problem space, select among possibilities, and continue to move toward a goal. None of that requires self-awareness. It only requires the capacity to successfully act in a shifting world.
Logging the frequency
Our priority should be improving the ability of agents to reach their goals, not asking them to contemplate their existence.