warm-up · Basics
Foundations — Warm-up
8 practice questions on warm-up. Every question is written from a specific moment in a real lecture, and after you answer it links to that exact timestamp so you can check it yourself.
What this pack asks
The free questions in this pack. Answer choices and explanations appear as you play.
A leading open AI model was trained on roughly 10 terabytes of internet text. How big is the finished model?
Answer this one →The first Transformer stacked 8 attention units side by side. Why 8 rather than 12 or 16?
Answer this one →Researchers designed every step inside a chatbot. So why can't they say why it gave one specific answer?
Answer this one →A text AI is set to always pick the single most likely next word. What kind of story does it write?
Answer this one →Reading nonstop, 24/7, how long would a person need to read all the text GPT-3 was trained on?
Answer this one →"Squatch eats pizza" and "pizza eats Squatch" turn into the exact same set of numbers. How does a Transformer tell them apart?
Answer this one →Translating "don't eat the delicious pizza", a model could drop the "don't" and flip the meaning. What prevents that?
Answer this one →"The pizza came out of the oven and it tasted good." How does a Transformer decide that "it" means the pizza?
Answer this one →
Written from lectures by Andrej Karpathy, StatQuest with Josh Starmer, 3Blue1Brown. Not affiliated with or endorsed by any university, channel, creator or certification program.