A novel adaptive neural network control method is proposed for the permanent magnet synchronous motor position servo system considering parameter uncertainties and load torque disturbances. First, neural networks are utilized to construct the intelligent model of the permanent magnet synchronous motor. Then, on the basis of backstepping control design and applying the characteristic of neural network basis function, an adaptive neural network dynamic surface control algorithm for position tracking is designed, which can overcome the ``explosion of complexity" problem. Finally, simulation results are given to verify the effectiveness of the designed control scheme. Compared with the traditional backstepping control scheme, the position servo system based on the neural network dynamic surface control can converge faster. The algebraic loop problem is overcome by designing novel adaptive parameter laws of the neural network weights. In addition, the proposed control algorithm can overcome the influence of the uncertain factors on the system performance with simple structure and easy implementation.