Scientific Discovery
Examinations of representation discovery
Physics Zero, Dirac, Zeta and supersymmetry are examinations the program means to run. None of them is an AIEN feature. Each one will test whether the machine can find a representation because it explains unseen observations better, after paying for its own complexity.
The rule
Never give AIEN the named answer when the experiment is intended to test discovery.
Each exam will supply generic mathematical machinery instead of the answer. A structure that emerges will count only if it earns positive Turings on held-out data. How the Turing scores.
Each exam carries its entry from the research status record, where one exists.
Physics Zero Status: Planned
Physics Zero comes last in the order of work. Like the other exams, it will follow the rule above.
It will be an examination of representation discovery: AIEN will not be given the named answer, and a structure will count only if it earns explanatory power on behavior it has not seen.
Dirac
Not on the status record yet.
There will be no Dirac primitive. The exam will provide generic capabilities and test whether the Dirac structure emerges because it earns explanatory power.
Generic capabilities it will receive
- complex values
- multi-component state
- linear operators
- matrix composition
- noncommutative composition
- commutators and anticommutators
- metric and signature
- differential operators
- constraints
- basis transformations
Zeta and spectral geometry Status: Research hypothesis
The machine will not be taught the Riemann hypothesis, Zeta Space or the critical line. The exam will test representation transitions: analytic, then operator, then spectral, then geometric or topological.
A transition will count only if it improves withheld explanatory performance after its complexity is charged.
Generic capabilities it will receive
- complex analysis
- operators
- spectra
- graphs
- topology
- group action
- noncommutative composition
- integral transforms
- symmetry constraints
Supersymmetry Status: Research hypothesis
There will be no hard-coded supersymmetry theory. The exam will ask whether the system discovers a relationship between classes of state because that relationship compresses unseen behavior.
Generic capabilities it will receive
- graded structures
- multiple state sectors
- graded commutators
- Grassmann-style machinery
- symmetry generators
- field and operator relations
When
The full design of these exams belongs to a later phase of the program, after the general mathematical machinery exists. Until then they remain designs and open questions. See what exists today.