this post was submitted on 03 May 2024
856 points (97.7% liked)
Technology
59427 readers
3991 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 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
The best convincing answer is the correct one. The correlation of AI answers with correct answers is fairly high. Numerous test show that. The models also significantly improved (especially paid versions) since introduction just 2 years ago.
Of course it does not mean that it could be trusted as much as Wikipedia, but it is probably better source than Facebook.
"Fairly high" is still useless (and doesn't actually quantify anything, depending on context both 1% and 99% could be 'fairly high'). As long as these models just hallucinate things, I need to double-check. Which is what I would have done without one of these things anyway.
Hallucinations are largely dealt with if you use agents. It won't be long until it gets packaged well enough that anyone can just use it. For now, it takes a little bit of effort to get a decent setup.
1% correct is never "fairly high" wtf
Also if you want a computer that you don't have to double check, you literally are expecting software to embody the concept of God. This is fucking stupid.
It's all about context. Asking a bunch of 4 year olds questions about trigonometry, 1% of answers being correct would be fairly high. 'Fairly high' basically only means 'as high as expected' or 'higher than expected'.
Hence, it is useless. If I cannot expect it to be more or less always correct, I can skip using it and just look stuff up myself.
Obviously the only contexts that would apply here are ones where you expect a correct answer. Why would we be evaluating a software that claims to be helpful against 4 year old asked to do calculus? I have to question your ability to reason for insinuating this.
So confirmed. God or nothing. Why don't you go back to quills? Computers cannot read your mind and write this message automatically, hence they are useless
That's the whole point, I don't expect correct answers. Neither from a 4 year old nor from a probabilistic language model.
And you don't expect a correct answer because it isn't 100% of the time. Some lemmings are basically just clones of Sheldon Cooper
I don't expect a correct answer because I've used these models quite a lot last year. At least half the answers were hallucinated. And it's still a common complaint about this product as well if you look at actual reviews (e.g., pretty sure Marques Brownlee mentions it).
Like most people, I have no interest in engaging in conversation with someone who gives me zero reason to.
Not that it's any of your business, but quality matters to me more than anything else, which is why I like tools that help me deliver it
Not reading anything else you write
An LLM has never generated a correct answer to any of my queries.
That seems unlikely, unless "any" means two.
Perhaps the problem is that I never bothered to ask anything trivial enough, but you'd think that two rhyming words starting with 'L" would be simple.
"AI" is a really dumb term for what we're all using currently. General LLMs are not intelligent, it's assigning priorities to tokens (words) in a database, based on what tokens were provided before it, to compare and guess the next most logical word and phrase, really really fast. Informed guesses, sure, but there's not enough parameters to consider all the factors required to identify a rhyme.
That said, honestly I'm struggling to come up with 2 rhyming L words? Lol even rhymebrain is failing me. I'm curious what you went with.
Ok, by asking you mean that you find somewhere questions that someone identified as being answered wrongly by LLM, and asking yourself.
Nope.
I’ve asked GPT4 to write specific Python programs, and more often than not it does a good job. And if the program is incorrect I can tell it about the error and it will often manage to fix it for me.
I don't believe you
I only "believe the hype" because a good developer friend of mine suggested I try copilot so I did and was impressed. It's an amazing technical achievement that helps me get my job done. It's useful every single day I use it. Does it do my job for me? No of fucking course not, I'm not a moron who expected that to begin with. It speeds up small portions of tasks and if I don't understand or agree with its solution, it's insanely easy not to use it.
People online mad about something new is all this is. There are valid concerns about this kind of tech, but I rarely see that. Ignorance on the topic prevails. Anyone calling ai "useless" in a blanket statement is necessarily ignorant and doesn't really deserve my time except to catch a quick insult for being the ignorant fool they have revealed themselves to be.
You would need to be pulling some trickery on Microsoft to get access to copilot for more than a single 30 day trial so I'm skeptical you've actually used it. Sounds like you're using other products which may be much worse. It also sounds like you work in a conservative shop. Good luck with that
Yeah you just illustrated you have no idea what copilot is like. But you were convinced you were an expert on it. Lol
K
OK
I think Meta hates your answer