[duringmath]

LLMs are too damn verbose

My issue with this GPT phase(?) we're going through is the amount of reading involved.

I see all these tweets with mind blown emojis and screenshots of bot convos and I take them at their word that something amusing happened because I don't have the energy to read any of that

[askvictor]

Having been a school teacher until a year ago, it's worth considering that a decent proportion of the population is functionally illiterate (well, it's a sliding scale). This kind of verbosity is probably excluding a lot of them from using this. Similarly I wonder if Google's rank preference for longer articles (i.e. why every recipe on the internet is now prefaced with the author's life story) has unintentionally excluded large portions of the population.

[nickfromseattle]

I was surprised to learn that 54% of Americans have below a 6th grade reading level. [0]

[0] https://www.snopes.com/news/2022/08/02/us-literacy-rate/

[BizarroLand]

WTF. I can't quite say that this information makes the world make a little more sense but it does let some light in. I figured that it would be much less than that.

Looks like near-complete or complete illiteracy is ~12%, so that means 42% of the population can read words but may have difficulty understanding the context of a short story or the meaning of a phrase.

[kobalsky]

just tell them "Keep your answers below 150 characters in this conversation." at the start.

[sitkack]

It can summarize its own output, the user directs everything about the output, style, format, length, etc. Everything.

[kobalsky]

> style

I like asking to to type like a frustrated teen on the phone. it huffs and puffs and rolls its virtual eyes.

prompt: could you pick a quantum computer at the mall for me?

response: ugh, seriously? you can't just buy a quantum computer at the mall, they're like super expensive and only a few companies sell them. Plus, they require special conditions to operate.

[sitkack]

> Ugh, seriously? Like, I can't even with this. I don't know why you're making me come to the mall just to pick out a quantum computer that you're not even gonna use properly. And you had to go and choose the one that uses liquid helium? That's, like, so old school. It's like listening to classic music while everyone else is jamming to something modern and cool. Do you want to keep using your classic computer too while you're at it? rolls eyes

[comboy]

I agree. ChatGPT just cannot be succinct no matter how many times I try. But it works with GPT-3 playground, I'm able to get much better information/characters ratio there.

[prawn]

Don't you just give it a word length, or number of sentences, or "rewrite, but half that length"? Those sorts of things have worked well for me. "Make it sound more exciting" or "tone down the excitement level" as well.

[magarnicle]

Whatever answer it gives, just say "Make a haiku about that".

[simple-thoughts]

The funny part is a task language models are actually quite good at is summarization. But people lacking social interaction can’t see how generic the responses are, so they get hooked into long meaningless conversations. Then again I suppose that’s a sign these language models are more intelligent than the users.

[duringmath]

Oh the bitter irony.

Yeah article summarization is the killer app for me but then again I don't know how much I can trust the output