This is crazy. I can now fully train models using Wolfram, in this example in a single prompt.
Using the Wolfram Alpha plugin, perform a scientific calculation (e.g., calculate the speed of light in a medium), then use the OpenAI API to fine-tune a language model for scientific explanations, and generate an explanation for the calculation result. Scientific calculation, Max tokens for explanation
These are actual requests made to the OpenAi api, you can control all settings to eliminate hallucinations. Give it a spin and let me know what you think.