Historical overview
The term predates 'artificial intelligence' and was used by Alan Turing in his 1950 paper 'Computing Machinery and Intelligence'. It remains the preferred umbrella term in robotics and control theory.
Scientific basis
Machine intelligence spans symbolic reasoners, statistical learners, neural networks, evolutionary systems, and neuromorphic hardware. The unifying feature is that cognition is realised in engineered substrates rather than biological neurons.
Strengths
- Deterministic reproducibility
- Direct integration with sensors and actuators
- Arbitrary scaling of compute and memory
Limitations
- No intrinsic motivation, embodiment, or evolutionary history
- Energy efficiency still orders of magnitude below biological brains
Relationship to other intelligence systems
Artificial Intelligence
AI is the dominant contemporary form of machine intelligence.
AGI
AGI is a hypothetical apex of machine intelligence.
Future implications
Neuromorphic hardware and analog AI accelerators promise machine intelligence with biological-scale energy efficiency.