Using OpenAI’s APIs is an easy way to get up and running with building AI driven applications, but what will these services cost you, and what are the implications in terms of energy usage and efficiency versus other non-AI driven options? How do you isolate the cost and energy of a specific user interaction with AI?
In this video, we’ll use the Firefox Energy Profiler to understand the energy usage of AI inference with OpenAI’s API, utilizing both the Whisper Transcription API, and ChatGPT 3.5-turbo’s completions API. We’ll then look at how much transcription and completion cost, and how they vary based on factors such as electricity cost in your area, grid intensity, and other factors.
