Mostly useful when you're only looking for the presence of words or terms in the output (including the presence of related words), rather than a coherent explanation with the quality of human-written text.
Sometimes the response is accidentally a truthful statement if interpreted as human text. The quality of a model is judged by how well-tuned they are for increasing the rate these accidents (for the lack of a better word).
[1]: EDIT: In the sense of "semantic web"; not in the sense of "actually understanding meaning" or any type of psychological sense.
I get links in my responses from Gemini. I would also not describe the response as soup, the answers are often quite specific and in the terms of my prompt instead of the inputs (developer queries are a prime example)
I'll stop calling them a soup when the part that generates a human-readable response is completely separate from the knowledge/information part; when an untrained program can respond with "I don't know" due to deliberate (/debuggable) mapping of lack of data to a minimal subset of language rules and words that are encoded in the program, rather than having "I don't know" be a series of tokens generated from the training data.
My personal experience has led me to increase my monthly spend, because the ROI is there, the UX is much improved
Hallucinations will never go away, but I put them in the same category as clicking search results to outdated or completely wrong blog posts. There's a back button
I'm keen to build an agent from scratch with copilot extension being open source and tools like BentoML that can help me build out the agentic workflows that can scale on a beefy H100 machine