About whether LLMs can explain their reasoning and cite sources the way an AGI should.
Has this happened?
Yes 62 (51%)Not sure 26 (21%)No 34 (28%)
Votes cast 1–2 October 2026: 100,590 votes from 9,694 people.
Hacker News has set AI a lot of challenges over the years. Which ones has it met?
About whether LLMs can explain their reasoning and cite sources the way an AGI should.
Votes cast 1–2 October 2026: 100,590 votes from 9,694 people.
When you say explaining reasoning is different from citing sources, why are you talking about citing sources?
Because that's what I was referring to and you moved the goalposts. Sources are a necessary part of the knowledge metadata so that truth weights can be taken into account. Otherwise the data's reliability factor isn't a first-class concept within the system. Citing the sources consistently serves as proof that the system does this (trust), and allows the end user to verify (at least to a reasonable level).
My point was that while ChatGPT might be a viable product, an actual AGI system should not be attempted to be built atop a LLM that can't do this.
Here is a system that does do this. https://wiki.opencog.org/w/AtomSpace