MicrosoftSo today, I was recording an educational video to demonstrate fine-tuning for the AI-102, but I was afraid it was too expensive ($34 / hour). These new price changes are great, so I can now go back and properly record a fine-tuning video on Azure AI Studio!
I don't fully understand why newer models are not fine-tunable. It's currently my guess that it is most cost-effective to fine-tune these "completion models", and maybe newer models like GPT4 would be too costly.
I wonder how useful you can make these older models, because I imagine with fine-tuning you can make these models a lot more useful but cost-effective but I have to devise a test case.
