Automated Gas Tungsten Arc Welding (GTAW) systems rely on highly costly precision control of welding conditions to produce repeatable results. Comparably, human welders have advantages in versatility and accessibility, yet fatigue and stress build up quickly thus adversely affecting their ability to produce quality welds. This paper proposes an innovative machine-human cooperative control scheme in which a machine algorithm determines (based on model prediction of human and process responses) adjustments to human welder controlled process. As the first study, this paper aims at accurate control of human arm movement. In particular, an innovative teleoperated virtualized welding platform is utilized to conduct dynamic experiments in order to correlate the human welder arm movement to the visual signal input. Linear model is firstly identified and an Adaptive Neuro-Fuzzy Inference System (ANFIS) model is then proposed to improve the model accuracy. To account for the welder's time-varying responses, an adaptive ANFIS model is finally used to model the intrinsic nonlinear and time-varying characteristic of the human welder response. An adaptive nonlinear ANFIS model-based predictive control (MPC) algorithm is then proposed to control the human arm movement. To demonstrate the controller's performance, human control experiments are conducted. Results verified that the proposed controller is able to track varying set-point and under input disturbance.