this post was submitted on 03 Jun 2024
1292 points (96.4% liked)
Technology
60078 readers
3310 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I do not agree with @FiniteBanjo@lemmy.today’s take. LLMs as these are used today, at the very least, reduces the number of steps required to consume any previously documented information. So these are solving at least one problem, especially with today’s Internet where one has to navigate a cruft of irrelevant paragraphs and annoying pop ups to reach the actual nugget of information.
Having said that, since you have shared an anecdote, I would like to share a counter(?) anecdote.
Ever since our workplace allowed the use of LLM-based chatbots, I have never seen those actually help debug any undocumented error or non-traditional environments/configurations. It has always hallucinated incorrectly while I used it to debug such errors.
In fact, I am now so sceptical about the responses, that I just avoid these chatbots entirely, and debug errors using the “old school” way involving traditional search engines.
Similarly, while using it to learn new programming languages or technologies, I always got incorrect responses to indirect questions. I learn that it has incorrectly hallucinated only after verifying the response through implementation. This makes the entire purpose futile.
I do try out the latest launches and improvements as I know the responses will eventually become better. Most recently, I tried out GPT-4o when it got announced. But I still don’t find them useful for the mentioned purposes.
That's an interesting anecdote. Usually my code sorta works and I just have to debug it a little bit, and it's way faster to get to a viable starting point that starting from scratch.
Often times my issue is unknown by it when debugging though, but sometimes it helps to find stupid mistakes.
I'd probably give it a 50% success rate, but I'll take the help.
Seems like you agreed with everything I said, tho.