How to make LLMs shut up

https://www.greptile.com/blog/make-llms-shut-up

by sebg • 2 years ago

21 11 2 years ago

mydriasis• 2 years ago

Hey, isn't this that company where the CEO said

Greptile offers no work-life-balance, typical workdays start at 9am and end at 11pm, often later, and we work Saturdays, sometimes also Sundays.

ksp-atlas• 2 years ago

That would be considered illegal here, is it legal where Greptile is based or are they breaking laws? I'm not saying this makes the practice any better, but I want to know if this actually puts them at risk of punishment.

KwisatzHaderack• 2 years ago

They are based in the US (CA I believe).

I think it’s certain hourly-wage jobs that have can hour limits. (ie. physically demanding jobs in hot weather conditions require periodic water breaks, in addition to hour limits)

In other hourly-wage jobs, you can still work over 40 hours but the worker is entitled to over-time pay. My mom worked at the US Post Office and during the holiday seasons she would sometimes work 12 hours a days, but all hours above 8 were paid at twice the hourly rate. Also, note that signing up for overtime was voluntary.

Software engineers, on the other hand, are salaried. Similar to lawyers, they sometimes have grueling hours (ie sleep in the office) but it’s not illegal, afaik.

manishsharan• 2 years ago

Their technology is a wrapper around LLM ; nothing original that will yield 10x return for your time if you accepted their employment.

dang2• 2 years ago

"Eschew flamebait. Avoid generic tangents."

https://news.ycombinator.com/newsguidelines.html

polotics• 2 years ago

Well yes ok sure. But I was considering using Greptile, and looked up the quote, and now it's very clear i won't invest time or money in something non-sustainable...

OutOfHere2• 2 years ago

In other words, just instruct the LLM to do what you need. The article is about nothing we don't already know.

jhgg• 2 years ago

Except this article is about how their efforts to prompt the LLM didn't end up working and how they used embeddings / vector search to filter out comments that the LLM generated based on user feedback.

OutOfHere2• 2 years ago

If one doesn't know how to prompt the LLM, or if one uses an inferior LLM, then it's one's own fault. The prompt shown in the article failed to convey to the LLM the reason behind the ask. The LLM can behave better once it internalizes the reasoning. The low-importance comments could alternatively also have been easily filtered out using a second use of the LLM without needing any voting. The approach used in the article isn't something to celebrate.

How to make LLMs shut up

Related Stories