Evaluating Artificial Intelligence Tools in the Pharmaceutical Industry: A Case Study on Paracetamol Dissolution and Calibration Curves

Purpose: This study evaluates the potential of artificial intelligence (AI), specifically ChatGPT, in assisting with the analysis of paracetamol dissolution tests and the construction of calibration curves. The primary research question investigates whether AI-generated data can approximate real laboratory results for pharmaceutical quality control, with an emphasis on exploring the limitations and applications of AI in pharmaceutical analysis. Methods: Paracetamol solutions were prepared, and their absorbance was measured using a UV-Vis spectrophotometer to construct calibration curves. Dissolution tests were performed in phosphate buffer solution, and both real and AI-generated data were analyzed. ChatGPT generated hypothetical data for the calibration curves and dissolution profiles, which were then compared to actual laboratory results. Statistical methods such as mean squared error (MSE), percent error, and correlation coefficients (R²) were employed to evaluate the accuracy of the AI-generated data. A paired t-test was used to determine the statistical significance of differences between the datasets. Results: AI-generated data followed general trends in both the calibration curve and dissolution tests but exhibited significant discrepancies in accuracy. Absorbance values from the AI model were consistently lower than real measurements, while AI overestimated drug release in early dissolution stages. The MSE for dissolution tests was 47.80, while for calibration curves, it was 0.0583. Despite these differences, AI-generated data aligned more closely with real data at later dissolution time points. Conclusions: While AI tools like ChatGPT can approximate trends in pharmaceutical analysis, real laboratory data are essential for accuracy, particularly in the early stages of testing. AI shows potential as a supplementary tool for theoretical understanding but cannot replace hands-on experimentation.