A Collection of Good LLM Analogies

Neil Gaiman

@neilhimself

· Follow

ChatGPT doesn't give you information. It gives you information-shaped sentences.

Twelvisten

@Twelvisok

@neilhimself Some more disturbing interactions with ChatGPT the other night.. What is your position on this direct misrepresentation of your work?

Aleksandr Tiulkanov

@shadbush

· Follow

My chatbot flowchart is still making huge rounds, now with @UNESCO support. Thanks to @ylecun for instantly driving around 200K people to my initial tweet about it. The links are in this thread

hat

@revhowardarson

· Follow

a ML model is just a giant statistical Plinko board. you start with randomized pegs and drop encoded data in the top and see if it landed in the right slot. you iterate back over the pegboard such that if you put the ball back in the same place it would drop into the right slot.

Omar Khattab

@lateinteraction

· Follow

Actionable, extended analogies are underrated. Here's one: LMs are stochastic *devices* for pattern processing. Like CPUs or GPUs, but Language Processing Units (LPUs). Assembly? Higher-Level Languages? Compilers? Understanding this analogy resolves several major questions.🧵

Andrew Kadel @DrewKadel@social.coop

@FrKadel

· Follow

My daughter, who's had a degree in computer science for 25 years, posted this about ChatGPT on Facebook. It's the best description I've seen.

andrew blinn

@disconcision

· Follow

feel like these kind of pictures are good to keep in mind when thinking about LLM output as well... current tech falls into a kind of uncanny valley where some details feel so high-effort relative to most human bullshit that it can be easy to overlook higher-level incoherence

Adam J Calhoun

@neuroecology

I heard you can use AI to teach you things, so I asked midjourney to show me How Things Work Am I doing this right?

Mark Dredze

@mdredze

· Follow

For years I’ve explained to people that computers are good at math but bad at language. Now, we’ve developed a new class of incredibly powerful AI that is great at language and, somehow, bad at math.

Hyung Won Chung

@hwchung27

Hey Bard…? 🤔