Digital coaches: an alternative to expert coaches for men's fitness goals

Abstract

This study compares the impact of training programs developed by an AI-supported application and those designed by personal trainers on specific health-related physical fitness parameters. Male individuals with similar health conditions and fitness goals were randomly assigned to two groups: one that utilized personalized exercise programs generated by OpenAI's GPT-4 model (n=10; age=21.7 years; height=174.45 cm) and another that engaged in training with a personal trainer. The first group (n=10) was trained using an AI-supported application (body weight=77.16 kg), while the second group (n=10) underwent training with a personal trainer (body weight=70.60 kg). Training programsThe training programmes were generated by inputting the subjects' information and commands into the OpenAI GPT-4 model. The personal trainers were provided with the personal information of the other group in order to plan their training sessions. The training programmes for both groups were designed to enhance upper extremity muscle strength and hypertrophy, with sessions lasting 90 minutes, three times a week for 12 weeks. Significant statistical changes (p < 0.05) were observed in muscle resistance values (right arm, left arm, right shoulder, left shoulder), body weight, lean body mass, and total body fluid parameters as a result of training with a personal trainer. However, the OpenAI GPT-4 recommendation indicated that the only significant improvement observed in the training group was in left arm muscle resistance (p < 0.05), with no significant changes in other parameters (p > 0.05). It was therefore concluded that strength and hypertrophy training under the supervision of a personal trainer might be more efficient than individually conducted training supported by artificial intelligence. However, further research with larger sample sizes is needed to establish a definitive conclusion..