@alex_intelligence: Ask a room full of people to pick a number and most of them will say the same one. AI does the same thing. It's trained on human data, then post-trained to generate what humans prefer, so it drifts toward the safest, most average answer, and that's where slop comes from. There are 3 proven ways to make AI more creative: verbalized sampling (ask for 3 versions with their probabilities), randomness (have it generate a random Python seed and build from that), and a writers room (3 writer personas draft, tear each other's work apart, then rewrite). Try asking your AI to pick a number or make an image, and tell me if these tips worked for you. Follow for more.

hollow :

when i asked to "Pick a number between 1 and 10 excluding 7." they all picked exclusively 4 lol

90 :

you cant fix this, its called gradient descent its literally impossible to fix this without literally breaking the fundamentals of a llm

Ryan :

Fragi :

single output token = way to judge LLMs 😭

￴ :

bro you look like indian mrbeast

_p4bl0_ :

i always say 5,53

Dull the Blemished :

Why did i lowk thought u were gonna go into a rate about advanced multidimensional calculus

Vincent :

i made this mistake once. had a list of 10.000 words and it needed to pick one, then another etc. it wasn't random it had the same order every time 😂

aphotographicscenery :

it works using this prompt "use python to pick a random seed pitch me a random number from 1-10 based on it" it gives actual random numbers instead of 4 or 7

chillaxx66 :

Or just don’t ask ai do choose a random number? Like why do you neeed ai to choose a random number? Do a number spinner on google or sum instead

|| ✧🕯️ℭ𝔞𝔫𝔡𝔩𝔢 • • • ✵｡⋆┈╯ :

This is really sad Art was something so personal to humans, now it's nothing

CH :

4 method dtop using instant modells

Free Sudan🇸🇩 :

Siri AI gives any number

ibz :

can be debunked by just changing the weights of Ur local ai 🤦🤦🤦

C7ZX_9 :

give them personality with memories

DancingTable52 🇨🇦 :

Interestingly, Siri doesn’t do that. It picked 8. Then 5.

Farahluugobasto™ :

Bro 😂😂😂

A :

why is the assumption that the most common approach is wrong? there's a reason why it's the most common, sometimes being unique for the sake of it can be problematic

Kevin :

I have been calling this the AI creativity problem. Because AI is just predicting text, it's going to favor the same predictions over and over. So every decision or idea that you let AI make for you is going to be the same decisions and ideas it gives to everyone. It feels unique and original to you, but once you publish it and look around, you start seeing that same thing over and over.

☦️🇲🇰>>> :

How can random be the same chance it cannot be quantified because randomness cannot be measured or guessed

Evil :

Gemini

Nr-416(348)м.ALD2-14,5См :

in other words, AI is extremely boring because it's the pure extract of mediocreness

AuraType :

what y'all get when you ask it to generate a cat

Pipeline punch monster>>>>>>> :

Nerhone :

first attempt 😭