this post was submitted on 17 Aug 2024
963 points (96.4% liked)

Technology

59217 readers
2900 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. 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
[–] [email protected] 50 points 2 months ago (3 children)

I think it might be because AI (aka LLMs) is genuinely useful when used properly.

I use AI all the time to write emails. I give the LLM the email thread along with instructions like “I can’t make it Tuesday ask if they can do Wednesday at 2pm”

The AI will write out an email that’s polite and relevant in context. Totally worth it.

I think the problem is people/companies trying to shove LLMs where they don’t make sense.

[–] [email protected] 72 points 2 months ago* (last edited 2 months ago) (3 children)

I am not a fan of this. I see it all the time at work and it's very obvious when someone has chatGPT write an email for them (it's always such a sterile and yet overcomplicated writing style). If it's a direct email to me, I tend to feel insulted that they couldn't be bothered to write those 4 paragraphs themselves - it would have taken them 2 mins. There is a definite human disconnect going on in society at the moment, and its worrying.

[–] [email protected] 34 points 2 months ago (1 children)

I agree. I actually think it's a net negative as well for friendships. As in the case of OP, I would rather get an original email from the sender saying they couldn't make it, so let's meet the next day, but instead I have to read thru several paragraphs of boilerplate and AI crap instead, which wastes my time, and I know the sender did it, so I'm mad at them for being impersonal. At some point, we're just going to have people's AI responding to each other without any person actually reading it.

We're only doing this because every company doesn't want to be left behind so they go all in. It feels like Ian Malcolm said it best in Jurassic Park

"Yeah, yeah, but your scientists were so preoccupied with whether or not they could that they didn't stop to think if they should"

[–] [email protected] 8 points 2 months ago (1 children)

In bureaucratic situations, you’re expected to have a bunch of polite boilerplate. Or at least that’s how my dad keeps telling me to write emails.

[–] [email protected] 1 points 2 months ago (1 children)

Which it seems you're missing the point of.

[–] [email protected] 2 points 2 months ago (1 children)
[–] [email protected] 1 points 2 months ago (1 children)

There's a purpose in showing that you put a little bit of thought into the email, not only for courtesy, but also because spending that attention can help you spot errors.

[–] [email protected] 1 points 2 months ago

Then why not just spend more time reading what I wrote out loud instead of spending that time solely on padding?

[–] [email protected] 1 points 2 months ago* (last edited 2 months ago) (1 children)

I can understand that. I don’t actually use chatGPT to be fair. I use a locally run open source LLM. This all being said I do think it’s important to fine tune any LLM you use to match your writing style. Else you end up with chatGPT generic style writing.

I would argue that not fine tuning a LLM to match tone and style counts as either misuse or hobbyist use.

[–] [email protected] 3 points 2 months ago* (last edited 2 months ago) (1 children)

I use a locally run open source LLM.

How? GPT4All + Llama or something else? I just started dipping my toe in locally run open source LLM.

not fine tuning a LLM to match tone and style counts as either misuse or hobbyist use

You’ve hit the nail on the head with this one. I think the other commenters are right, that a lot of people will misuse the tool, but nonetheless it is an issue with the users, not the tool itself.

[–] [email protected] 2 points 2 months ago

My main workstation runs Linux and I use Llama.cpp. I used it with mistral’s latest largest model but I have used others in the past.

I appreciate your thoughts here. Lemmy I think, in general, has an indistinguishing anti LLM bias.

[–] [email protected] -1 points 2 months ago

Agreed. People are so bad at writing that they struggle to put a few sentences together for an email. Even their prompts lack clear instructions /message. It's astounding when you think about it for a minute.

[–] [email protected] 54 points 2 months ago (1 children)

Why not just write "I can't make it Tuesday, can you do Wednesday at 2pm?"

Otherwise we just end up in this world.

[–] [email protected] 16 points 2 months ago (1 children)

You’re not wrong but at least my emails will be taken seriously by some 60 year old company exec that’s still mad his secretary stopped printing his emails for him.

[–] [email protected] 17 points 2 months ago (1 children)

You’re trying to please a boomer that’s still angry that email exists in the first place.

[–] [email protected] 6 points 2 months ago

In some cases literally yes. But at least for me I have to meet my customers where they are. If I try to force them to do things my way they just don’t use my services.

[–] [email protected] 19 points 2 months ago (1 children)

Then just write that.

I don't understand why we're having AIs verboseify simple information?

Why do many word if few word do trick.

How long until we start using LLMs to summarize messages over-verbalized by LLMs?

And offloading the accounting for context WILL bite you in the ass. If you can't remember what a discussion was about and what needs considering, you're no longer doing the thinking.

[–] [email protected] 14 points 2 months ago* (last edited 2 months ago) (2 children)

Because in my experience some business clients feel offended or upset that you aren’t being formal with them. American businesses seem to care less I noticed but outside of the USA (particularly in Germany) I noticed that formality serves better. Also the LLM uses the thread history to add context. Stuff like “I know we agreed on meeting on Tuesday at last meeting but unfortunately I can’t do that…” this stuff matters to clients.

I don’t offload because I don’t remember. I offload because it saves me time. Of course I read what is written before I send it out.

[–] [email protected] 5 points 2 months ago (2 children)

Being formal and considerate does not require being that much more verbose.

Do you really save time running messages through an LLM vs just writing them as you think of what to say?

[–] [email protected] 7 points 2 months ago* (last edited 2 months ago) (2 children)

It's the equivalent of when I got assigned papers with minimum word counts as a kid. Despite the fact that the prompt doesn't warrant 5000 words and it would take massive deviation off of the prompt to get anywhere close to it, people have this weird impression that more words shows more "care" than just communicating clearly. I struggled a lot with a lot of assignments (to the point of not turning some in) because all the filler they'd need to reach the word counts hurt my soul lol.

(I do tend to prefer 500+ page books, but it's because the authors I engage with the most use that space to build out better plots or develop better characters or whatever. It's not padded out.)

[–] [email protected] 2 points 2 months ago* (last edited 2 months ago) (1 children)

Is it?

I once told a teacher I'd write ten times the number of required words as long as I could pick a subject that actually warranted it. And I followed through.

The rare times I got prompts that were actually good, I would run out of paper on which to express everything I wanted expressed. (Yes, I've done writing assignments writing by hand.)

Outside academia no-one is enforcing a word-count. Which means you can just write good prose. Using a lot of words to say very little, is not good prose.

Unless you're dealing with people that don't actually read what you write and instead just look at net weight of the word-salad you threw at them, the content of the text is what matters.

Who takes offence at only a single paragraph, if it addresses their every concern and insecurity, and they are left feeling seen as they reach the final word?

Only people who don't actually read things, or have no reading comprehension, needing the same thing said three time in different ways in one message.

[–] [email protected] 1 points 2 months ago (1 children)

It's the same philosophy, yeah. That more words means more substance and more "respect" or whatever to the message.

It's not rational at all, but people genuinely don't think that way. (Unless it's a forum/social media, then 3 paragraphs is a wall of text that needs to have a 5 word TLDR, because none of it is rational).

The exaggerated version of a simple message once you have a working relationship is silly, but there are way too many times you don't get to a working relationship at all without a wall of bullshit.

[–] [email protected] 2 points 2 months ago* (last edited 2 months ago) (1 children)

Ok, but do the people you're referring to actually appreciate prose, or just skim-read through everything?

Because I'd wager they're the latter, and at that point you don't even need to try to write something good. It's fine to send them three paragraphs where the second and third ones just paraphrase the first.

[–] [email protected] 0 points 2 months ago (1 children)

Paraphrasing the first multiple times is still a big, distracting extra cognitive load, and it needs to hold up if they actually do pay attention to it. One time they notice the obvious bullshit can end a relationship. I won't use an LLM for anything like this because it's stupid, but that's why people are doing it.

[–] [email protected] 1 points 2 months ago* (last edited 2 months ago)

You're saying once they see the pointless fluff they themselves ask of people, for what it is, they'll feel insulted?

Paraphrasing yourself comes with built-in deniability. "Oh it's just something I tend to do, I don't mean anything by it, I can make an effort to stop if you like". And then boom, you get to be concise.

There is no way of bloating your prose that doesn't come off as insulting when done with people who don't appreciate volume over quality.

[–] [email protected] 0 points 2 months ago (1 children)

Also, teachers are typically smart enough to probably themselves understand the word-count problem. Which is why I was able to make deals with many of my teachers to change the assignments given such that writing something good was actually possible.

Hence why it's not the same. The people you are talking about aren't worth the effort of dealing with. A writing teacher that gives you high marks for saying nothing with a lot of words, is not a good writing teacher.

[–] [email protected] 1 points 2 months ago (1 children)

I never once had a language teacher that had even the tiniest shred of competence. It's not the norm.

[–] [email protected] 0 points 2 months ago* (last edited 2 months ago) (1 children)

Was for me. I've had teachers assistants that were intelligent and pedagogically literate. Benefits of going to school in the nordics, I guess.

But my point stands. That makes those people unworthy of the effort. You might play to those things to get ahead, but it still doesn't mean it's good communication.

And good communication should be your default behaviour, otherwise you're part of the problem.

[–] [email protected] 2 points 2 months ago

I don't play those games.

But most people do, because there's a lot of it required to succeed in a lot of industries. (Even if most recognized that it's nonsense, which they don't), everyone can't just apply for the one percent of bosses who don't do bullshit games.

[–] [email protected] 0 points 2 months ago (1 children)

The LLM responses are more verbose but not a crazy amount so. It’s mostly adding polite social padding that some people appreciate.

As for time totally. It’s faster to write “can’t go to meeting, suggest rescheduling it for Thursday.” And proofread than to write a full boomer style letter.

[–] [email protected] 2 points 2 months ago* (last edited 2 months ago) (1 children)

I feel like we might write at very different WPMs. For me, proofreading and fixing AI slop takes longer than just writing things myself.

And another difference might be that to me and everyone I work with, writing in full on "boomer" is considered an insulting waste of everyone's time.

Which it is.

[–] [email protected] 0 points 2 months ago

It’s a waste of everyone’s time for sure. It’s just good business sense to make your customers happy though.

As for typing speed perhaps ya lol. You could be faster. But I think the best approach here is using high quality locally run LLMs that don’t produce slop. For me I can count on one hand how many times I’ve had to correct things in the past month. It’s a mater of understanding how LLMs work and fine tuning. (Emphasis on the fine tuning)

[–] [email protected] 3 points 2 months ago

Abs Japan. Oh my.