Quanta Foundations Lab

Physics Law Discovery: AI vs Symmetry & Conservation

Direct Dynamical Manipulation & Phase Space Runge-Kutta 4th Order Symplectic Integration
“Confronting that — what is physics, really? What is enduring about the practice of physics itself? — that does not change with AI.” — Physicist Sarah (Quanta Magazine)
Spatial Orbit / Configuration Space $(x, y)$ Active Trajectory
Click & drag body or velocity vector
Hamiltonian Phase Space $(q, p)$ & Invariant Curves Symplectic Foliation
Live Poincaré Section / $(q_x, p_x)$
Noether Invariance Residual
2.00e-6
Invariant Preserved
Empirical Extrapolation Error
14.8×
Symmetry Breakdown
Angular Momentum $L_z$
1.100
Exact $\partial \mathcal{L}/\partial \dot{\theta}$
Total Energy $E$ (Hamiltonian)
-0.395
Time Invariant

When machine learning models generate physical predictions, they interpolate over statistical patterns without asserting mathematical necessity. Below is the strict demarcation between rigorous physical theorems and empirical AI conjectures:

Category Principle / Statement Mathematical Grounding Epistemic Status
Continuous Symmetries Noether (1918): Any continuous global symmetry of the action $S = \int \mathcal{L} \, dt$ guarantees a conserved current $\partial_\mu j^\mu = 0$. Rigorous Lie algebra proof via variational calculus. PROVED THEOREM
Orbital Energy Invariant Vis-viva relation: $v^2 = \mu(2/r - 1/a)$, orbital energy $E = -G M / (2a)$ remains strictly constant under central inverse-square potentials. Analytical Kepler integration. PROVED THEOREM
Laplace-Runge-Lenz Vector Vector $\vec{A} = \vec{p} \times \vec{L} - m k \hat{r}$ is conserved due to hidden $SO(4)$ dynamic symmetry in the Kepler problem. Lie bracket commutation $[H, A_i] = 0$. PROVED THEOREM
Out-of-Distribution Shift Deep neural operators generalize accurately across unmeasured eccentricities $e > 0.9$ without explicit inductive symmetry priors. Empirical regression over sampled parameter bands. CONJECTURE
Algorithmic Complexity Minimal symbolic expression complexity (Kolmogorov/Occam heuristic) necessarily maps to the true underlying physical ontology. Heuristic simplicity preference; uncomputable in general. CONJECTURE
Enjoy this tool? Build your own with Super