One way to view the use of LLMs is by purpose or intended outcome. Some examples include: * A user has in their brain a definite statement, thought, or concept they wish to translate to words-on-the-page. They use an LLM to explore different ways to express that definite statement/thought/concept. * A user has a question they then pose to the LLM, hoping that the LLM's answer is truthful/not-truthful/funny/interestin…
For instance I got:
Chocolate Tomato Sausage Surprise where I was instructed to pour tomato sauce and mustard over some sausages before pouring chocolate over the whole dish and then cooking it in the oven.
I got a pasta dish where I was instructed to cook dry pasta and then add a sauce made from tinned pasta.
I'm sure everyone's seen the suggested mocktail made from ammonia and bleach by now.
How about a cream sauce made from mayo, tinned peaches, honey, and lollipops?
Can interest you in some "Corn Syrup Surprise" in which you pour corn syrup over a cheese, tinned meat, and aspic mixture? It's cooked in the oven.
How about some turmeric meat stew? Just boil some tinned meat in vinegar.
It spit out a spicy cupcake recipe with no flour. Half a cup of cream mixed with a quarter cup of hot sauce for the frosting though.
Meanwhile with no GPT intervention I made an unintuitive corn and grape cake today from a recipe that a human wrote. Absolutely delicious. Of course a human would know why corn and fruit pair together so well (and how to pull it off).