3 ms·
Ask HN: How do you make LLM generated text believable?
As a graudate student who just working as management & IR, I use LLM to do daily jobs , including weekly briefing and
But AI generated report looks too good to be checked, and hallucination can't be terminated. But Boss and SEC can not tolerate any number mistakes.
So, can you tell me, how do you solve this problem?
- bigyabai 4mo ago...do you actually send the SEC LLM-generated documents? Do you understand the liability you're assuming, doing that?
- StahlGuo 4mo agoOh no I am not a attorney, but I need to analyze the market, 20-F and 10-K recurringly and make strategy suggestions, and so yes I use LLM to do business brief.
- 0x1d7 4mo ago> So, can you tell me, how do you solve this problem? Write it yourself.
- dabinat 4mo agoIf you were writing it yourself you would proofread and double-check everything. Why do you think that’s no longer required?
- nehadangwal 4mo ago[dead]
- evil-olive 4mo ago> But AI generated report looks too good to be checked have you considered...checking it anyway, no matter how good it looks?
- StahlGuo 4mo agoI don't mean never check it by myself, but I want to discuss about a methodology to ensure AI generated text has the same auditability and tracebility like human wrote text.
- StahlGuo 4mo agoI am sorry that I did not explain my question explicitly. I write recurring weekly briefings for internal use, such as market insights and industry news. I’m not trying to make AI-generated text “believable”. I’m asking almost the opposite question: when an AI generated text is fluent enough to hide mistakes, how do human check how to systematically check numbers, dates, cites and judgements?
- fragmede 4mo agoHave it output the numbers it's basing the conclusion on, have it output a program that it's used to do math to derive judgements.
- StahlGuo 4mo agoYes, i agree that quantitative part， like dates, numbers, amounts should be extracted and let the LLM to output original numbers and computation steps. That's not hard for a briefing harness. However, my most confusion part is qualitative side, like market insight from a news,policy change and interpretation, and industry NEWS interpretation, they are not straight math but they need tracebility. Do you have any idea to solve those judgement claim?