Then, lastly, you need to give it suggestions.
Each time it offers you output and you’ve got added context, you’ve got added storytelling, you’ve got added expert-led insights, you’ve got added something, you’ve got made adjustments, you’re taking the ultimate model of what you’ve got written, for that part, and also you give it again to the LLM. You say, “That is the ultimate model I am going with. I’ve made X, Y, Z adjustments since you didn’t embody this, otherwise you left this out, otherwise you did this. You have gone towards our guardrails. You haven’t adopted our directions.” You are in a position to give it that suggestions.
The extra you give it that suggestions, the less errors it should make. The purpose ideally, for me, is to get to 70% high quality output. I am not aiming for 100%. I am not loopy. Seventy p.c, that is the purpose. We need to get to 70% of nice output in order that we’re making fewer and fewer adjustments to the output.
Once more, cease attempting to automate content material. I do not assume that is actually viable if you wish to keep high quality. The trick is to write down in batches, personalize the output, after which make it really feel such as you. However keep in mind, it’s a must to hold coaching these LLMs as a result of they’re machines on the finish of the day.
And that, my mates, is the way you write nice content material with LLMs. Thanks.
