“With fine-tuning, there’s always the issue of data that we fine-tune on and compute. We can address hallucination and bias with better data. Frequency of fine-tuning helps but it’s an expensive procedure, most of the work that happens is on the data-side. We’re always on the search for more volume of data and better annotations.” Mohammed Minhaas, DATA ENGINEER 20 21

AI Readiness Report 2024 - Page 23 AI Readiness Report 2024 Page 22 Page 24