818
Why is no one talking about how unproductive it is to have verify every "hallucination" ChatGPT gives you?
(lemmy.world)
this post was submitted on 19 Nov 2024
818 points (98.0% liked)
People Twitter
5236 readers
1745 users here now
People tweeting stuff. We allow tweets from anyone.
RULES:
- Mark NSFW content.
- No doxxing people.
- Must be a tweet or similar
- No bullying or international politcs
- Be excellent to each other.
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
Because in a lot of applications you can bypass hallucinations.
In all these applications you can bypass hallucinations because either it's task is non-factual, or it's verifiable while promoting, or because you will be able to verify in any of the superseding tasks.
Just because it makes shit up sometimes doesn't mean it's useless. Like an idiot friend, you can still ask it for opinions or something and it will definitely start you off somewhere helpful.
All LLMs are text completion engines, no matter what fancy bells they tack on.
If your task is some kind of text completion or repetition of text provided in the prompt context LLMs perform wonderfully.
For everything else you are wading through territory you could probably do easier using other methods.
I love the people who are like "I tried to replace Wolfram Alpha with ChatGPT why is none of the math right?" And blame ChatGPT when the problem is all they really needed was a fucking calculator
Also just searching the web in general.
Google is useless for searching the web today.
Not if you want that thing that everyone is on about. Don't you want to be in with the crowd?! /s
so, basically, even a broken clock is right twice a day?
Yes, but for some tasks mistakes don't really matter, like "come up with names for my project that does X". No wrong answers here really, so an LLM is useful.
great value for all that energy it expends, indeed!
Can't agree
The energy expenditure for GPT models is basically a per-token calculation. Having it generate a list of 3-4 token responses would barely be a blip compared to having it read and respond entire articles.
There might even be a case for certain tasks with a GPT model being more energy efficient than making multiple google searches for the same. Especially considering all the backend activity google tacks on for tracking users and serving ads, complaining about someone using a GPT model for something like generating a list of words is a little like a climate activist yelling at someone for taking their car to the grocery store while standing across the street from a coal-burning power plant.
no, it's like a billion people taking their respective cars to the grocery store multiple times a day each while standing across the street from one coal-burning power plant.
each person can say they are the only one and their individual contribution is negligible. but get all those drips together and you actually have a deluge of unnecessary wastage.
Except each of those drips are subject to the same system that preferences individualized transport
This is still a perfect example, because while you're nit-picking the personal habits of individuals who are a fraction of a fraction of the total contributors to GPT model usage, huge multi-billion dollar entities are implementing it into things that have no business using it and are representative for 90% of llm queries.
Similar for castigating people for owning ICE vehicles, who are not only uniquely pressued into their use but are also less than 10% of GHG emissions in the first place.
Stop wasting your time attacking individuals using the tech for help in their daily tasks, they aren't the problem.
How is that faster than just picking a random name? Noone picks software based on name.
And yet virtually all of software has names that took some thought, creativity, and/or have some interesting history. Like the domain name of your Lemmy instance. Or Lemmy.
And people working on something generally want to be proud of their project and not name it the first thing that comes to mind, but take some time to decide on a name.
No, maybe more like, even a functional clock is wrong every 0.8 days.
https://superuser.com/questions/759730/how-much-clock-drift-is-considered-normal-for-a-non-networked-windows-7-pc
The frequency is probably way higher for most LLMs though lol