Chart shows LLMs asked to pick a number 1-10 almost always converge on 7
gerardsans · x · 2026-11-20
A viral chart visualizes what happens when you ask models to "pick a number between 1 and 10": humans spread loosely around 5-7, true randomness is flat, but the LLM distribution spikes sharply at 7 — a neat illustration that LLM "randomness" is highly deterministic sampling, compounded by training-data bias.
More from Fun
- Newbie asks which programming language to learn first — programmers all point to ChatGPT — miniapeur · 2026-11-20
- $100K Hackathon Disqualifies #1 Product, Splits Prize Into More Winners — marclou · 2026-11-20
- "My toxic trait: explaining how his job will be automated on a first date" — sierracatalina · 2026-11-20
- Meme: what the average IBM sales rep looked like in the 1980s — thedealdirector · 2026-11-20
- 'Password Is Too Weak (Your Worst Nightmare)': the yoga grandpa meme — Feeling-Total5959 · 2026-11-20
- Anthropic falls below $90 premarket, over $135B in value set to be erased — GaryMarcus · 2026-11-20