3 ms·
Not that it would solve this, but how hard would it be for ChatGpt or other problems to cite the sources used in a response. Is that difficult to capture and ta
by bigmattystyles 3y ago
Not that it would solve this, but how hard would it be for ChatGpt or other problems to cite the sources used in a response. Is that difficult to capture and tag to 'knowledge' within a LLM?
It could be a best of both worlds type situation if LLMs cited sources and linked to the source itself.
Isn't that what happened with Google News's home page? I seem to recall that when Google took it away in some markets, at the behest of the news orgs, they quickly reversed course as their traffic plummeted.
- jejeyyy77 3y agonot likely with the way these models have been trained - its basically broken down into sub-words that are all mashed together into probabilities.
- qznc 3y agoThis is not possible. There is no database of sources inside an LLM. Just like the knowledge in your brain does not have sources attached. For an example, you referenced "what happened with Google News's home page". Could you give me your source? You could probably search for some suitable article for a reference, but you don't know a source from your memory.