Now go evaluate and perform a task to get money continue your objectives of survival, as we will not provide free food or shelter. Here is a bank account to get you started. Beep boop. I mean uh, goodbye.
I bet it does have an impact on mistakes. I bet it tries harder to gaslight and hide them. Yeah, it doesn’t actually decrease the quantity of mistakes, but it does change how things are presented and makes it use words that convince you it’s actually considered things and won’t have made a mistake.
I am genuinely curious if anyone knows if that has an effect or not. I wouldn’t think so per se, but if the AI interprets it as “double check your initial analysis for errors” it would actually work maybe?
For image generation it does. They give negative prompts like “wonky”, “creepy”, and “ugly”, and the image generator evaluates how well the generated image matches those prompts, and produces images opposite those parameters.
Some poor artist in the training data not only had their work stolen to train the AI, but also had it labeled ugly and wonky.
That could also be human training as well. For example, an artist’s work would be used as a “correct” sample, and the machine told to make some other image based on the correct samples, and people would mark results with those tags.
It won’t affect the output meaningfully except by rerolling whatever training data ends up being associated with that or whatever. It may end up getting the model to “check” its work which just compares previous output to training data.
I literally have Claude send every edit to another model to check and make sure it isn’t word barfing. Every file edit is a call to another model to make sure that edit doesn’t suck.
The LLMs are really very very good at pumping out scripts that can accelerate things like machine learning where it are wrangling data and doing proof of concepts.
They are also pretty good at basic CRUD feature work which is what the majority of software devs are actually doing.
The further out of the user’s depth they go the more problematic they can be. They bake many many assumptions in and make hidden decisions that someone without domain expertise cannot easily intuit. Which means someone without experience can get into deep water and not realize and that is where a lot of the problems are.
I love the make no mistakes prompt.
Fun fact: zero impact on mistakes.
No it definitely has an impact, it’s like “don’t think about elephants” but with mistakes.
Damn I can’t stop thinking about mistakes. Am I a LLaMe?
You are mistaken.
My apologies, as a human I’m still in development, but I’ll try harder not to think about elephants next time.
Now go evaluate and perform a task to get money continue your objectives of survival, as we will not provide free food or shelter. Here is a bank account to get you started. Beep boop. I mean uh, goodbye.
I bet it does have an impact on mistakes. I bet it tries harder to gaslight and hide them. Yeah, it doesn’t actually decrease the quantity of mistakes, but it does change how things are presented and makes it use words that convince you it’s actually considered things and won’t have made a mistake.
Instead of the model just saying things plainly and being obvious so you can catch them. It makes it hide the fuck ups!
Like when you punish your child for doing anything at fucking all so they learn the only right move is to avoid, gaslight, and deflect!
I am genuinely curious if anyone knows if that has an effect or not. I wouldn’t think so per se, but if the AI interprets it as “double check your initial analysis for errors” it would actually work maybe?
For image generation it does. They give negative prompts like “wonky”, “creepy”, and “ugly”, and the image generator evaluates how well the generated image matches those prompts, and produces images opposite those parameters.
Some poor artist in the training data not only had their work stolen to train the AI, but also had it labeled ugly and wonky.
That could also be human training as well. For example, an artist’s work would be used as a “correct” sample, and the machine told to make some other image based on the correct samples, and people would mark results with those tags.
It won’t affect the output meaningfully except by rerolling whatever training data ends up being associated with that or whatever. It may end up getting the model to “check” its work which just compares previous output to training data.
Earlier LLMs it helped a bit.
Now a days the harnesses know to spawn ‘review’ agents which will catch some mistakes but not all.
You mean it will spawn agents to drive up the token costs and maybe fingers crossed catch some errors?
I’m on the 20 dollar a month z.ai plan, I’ve yet to hit the 5 hour limit. What token costs?
Claude I’d usually hit it in an hour at most lol
Keep on pumping out that mediocre work.
Correct
I literally have Claude send every edit to another model to check and make sure it isn’t word barfing. Every file edit is a call to another model to make sure that edit doesn’t suck.
Tokens++
When does it become easier to just write the thing yourself?
It really really depends on what ‘it’ is.
The LLMs are really very very good at pumping out scripts that can accelerate things like machine learning where it are wrangling data and doing proof of concepts.
They are also pretty good at basic CRUD feature work which is what the majority of software devs are actually doing.
The further out of the user’s depth they go the more problematic they can be. They bake many many assumptions in and make hidden decisions that someone without domain expertise cannot easily intuit. Which means someone without experience can get into deep water and not realize and that is where a lot of the problems are.
Lol “know”. LLMs don’t know anything. It’s an advanced autocomplete.
That’s not fair, my auto complete knows to always replace duck with duck even when I try to fix it back to duck!
Yeah 'cause faulty children won’t make mistakes. 🤣